Game Architecture and Design: A New Edition

00 3634_FM_Intro
10:14 AM
Page i
Game Architecture and Design:
A New Edition
Contents at a Glance
Game Design
First Concept 3
Core Design 35
Gameplay 59
Detailed Design 87
Game Balance 105
Look and Feel 141
Wrapping Up 171
The Future of Game Design 197
Team Building and Management
Current Methods of Team Management 227
Roles and Divisions 245
The Software Factory 263
Milestones and Deadlines 293
Procedures and “Process” 327
Troubleshooting 367
The Future of the Industry 409
Game Architecture
Current Development Methods 433
Initial Design 457
Use of Technology 511
Building Blocks 553
Initial Architecture Design 607
Development 637
The Run-Up to Release 687
Postmortem 719
The Future of Game Development 747
IV Appendixes
A Sample Game Design Documents 785
B Bibliography and References 887
Glossary 893
Index 897
00 3634_FM_Intro
10:14 AM
Page ii
00 3634_FM_Intro
10:14 AM
Page iii
Game Architecture and Design:
A New Edition
Andrew Rollings
Dave Morris
800 East 96th Street, 3rd Floor, Indianapolis, Indiana 46240
An Imprint of Pearson Education
Boston • Indianapolis • London • Munich • New York • San Francisco
00 3634_FM_Intro
10:14 AM
Page iv
Game Architecture and Design:
A New Edition
Copyright © 2004 by New Riders Publishing
All rights reserved. No part of this book shall be reproduced,
stored in a retrieval system, or transmitted by any means—
electronic, mechanical, photocopying, recording, or otherwise—
without written permission from the publisher, except for the
inclusion of brief quotations in a review.
International Standard Book Number: 0-7357-1363-4
Library of Congress Catalog Card Number: 2003111600
Printed in the United States of America
First printing: November, 2003
08 07 06 05 04
7 6 5 4 3 2 1
Interpretation of the printing code: The rightmost double-digit
number is the year of the book’s printing; the rightmost singledigit number is the number of the book’s printing. For example,
the printing code 04-1 shows that the first printing of the book
occurred in 2004.
Stephanie Wall
Production Manager
Gina Kanouse
Senior Project Editor
Kristy Hart
Copy Editor
Chrissy Andry
Senior Indexer
Cheryl Lenser
Gloria Schurick
Manufacturing Coordinator
Dan Uhrig
Interior Designer
Kim Scott
Cover Designer
Aren Howell
Media Developer
Jay Payne
All terms mentioned in this book that are known to be trademarks
or service marks have been appropriately capitalized. New Riders
Publishing cannot attest to the accuracy of this information. Use of
a term in this book should not be regarded as affecting the validity
of any trademark or service mark.
Scott Cowlin
Tammy Detrich
Hannah Onstad Latham
Warning and Disclaimer
Publicity Manager
Susan Nixon
Every effort has been made to make this book as complete and as
accurate as possible, but no warranty of fitness is implied. The
information is provided on an as-is basis. The authors and New
Riders Publishing shall have neither liability nor responsibility to
any person or entity with respect to any loss or damages arising
from the information contained in this book.
00 3634_FM_Intro
10:14 AM
Page v
This book is dedicated to the memory of Ram De Silva,
respected colleague and beloved friend.
Andrew Rollings
In loving memory of my father, Victor Morris.
Dave Morris
00 3634_FM_Intro
10:14 AM
Page vi
Game Architecture and Design: A New Edition
Table of Contents
Part I
Game Design
Chapter 1
First Concept
The Shock of the New
The Creative Road Map
Having the Idea
Shaping the Idea
Dramatic Effect
The Treatment
Taking Stock
Case Study 1.1
The One-Page Pitch
Getting it Down
Case Study 1.2
Chapter 2
Initial Treatment for Conquerors
Core Design
What Is a Game?
Cool Features
Fancy Graphics
Setting and Story
Games Aren’t Everything
Case Study 2.1 Story Versus Gameplay
00 3634_FM_Intro
10:14 AM
Page vii
Table of Contents
Chapter 3
Games Mean Gameplay
Case Study 2.2 A Missed Opportunity?
Creating the Game Spec
Case Study 2.3 Integrating Game Objectives
Case Study 2.4 An Instance of Emergence
Case Study 2.5 An Elegant Interface
Case Study 2.6 The Rules Must Serve the Features
Level Design
Case Study 2.7 Interesting Level Design
Example Game Spec
Case Study 2.8 Game Spec
The Value of Prototypes…
…And the Necessity of Documents
What Is Gameplay?
Implementing Gameplay
The Dominant Strategy Problem
Near Dominance
Case Study 3.1 Environment Plus Rules Equals Gameplay
Supporting Investments
Case Study 3.2 Unexpected Versatility
Compensating Factors
Case Study 3.3 Balancing Compensating Factors
Shadow Costs
Case Study 3.4 Shadow Costs in Age of Empires
A Final Word About Gameplay
Kinds of Interactivity
Case Study 3.5 A Different Kind of Interactivity
“Why?” Versus “What?”
00 3634_FM_Intro
10:14 AM
Page viii
Game Architecture and Design: A New Edition
Chapter 4
Chapter 5
Chapter 6
Detailed Design
The Designer’s Role
Case Study 4.1 A Development Timeline
Design Documentation
The Gameplay Spec
The Designer’s Notes
Case Study 4.2 The Need for Documenting the Spec
Using The Design Documents
Fitting Design to Development
Tiers and Testbeds
Case Study 4.3 Planning the Mini-Specs to Fit
the Architecture
Why Use Documents at All?
Game Balance
Player/Player Balance
Player/Gameplay Balance
Case Study 5.1 Is This Supposed to Be Fun?
Reward the Player
Let the Machine do the Work
Make a Game You Play With, Not Against
Case Study 5.2 The Save Game Problem
Gameplay/Gameplay Balance
Component and Attribute Balance
Case Study 5.3 Component and Attribute Balance in
Dungeon Keeper
Intransitive Game Mechanics Guarantees Balance
Case Study 5.4 Attribute Balance Using SPS
Case Study 5.5 Using Game Theory Analysis to
Achieve Balance
A Game Balance Checklist
Look and Feel
Case Study 6.1
Case Study 6.2
Sound Effects at Their Best
A Discordant Note
00 3634_FM_Intro
10:14 AM
Page ix
Table of Contents
Case Study 6.3
Case Study 6.4
Chapter 7
Chapter 8
Meshing the Interface with Look and Feel
Sometimes Less Is Less
A Toolbox of Storytelling Techniques
Case Study 6.5 An Example of a Look-and-Feel Document
Case Study 6.6 An Unexpected Development
Case Study 6.7 An Unsatisfying Conclusion
The Sum of the Parts
Wrapping Up
The Professionals
The Game Concept
Planning for Change
The Technology
The Team
Costs and Timelines
The Future
The Future of Game Design
The Necessity of Design
Don’t Be Afraid to Plan
Case Study 8.1 Design Saves Time
Why Design Is Fine
Case Study 8.2 Keep the Design up to Date
Essentials of Game Design
Is it Original?
Is it Coherent?
Is it Interactive?
Is it Interesting?
Is it Fun?
The Future of Design
Making Designs More Generic
Nonsymbolic Design
Case Study 8.3 Comparing Nonsymbolic and
Symbolic Design
00 3634_FM_Intro
10:14 AM
Page x
Game Architecture and Design: A New Edition
Part II
The Future of Games
The Next Decade
The Strengths of Software
The Crossroads of Creativity
Case Study 8.4 An Example of Mise En Scene
Games as Entertainment
The Way Forward
Team Building and Management
Chapter 9
Current Methods of Team Management
The Current Development Model
The Origins of the Industry
The Trouble with Game Developers
The Problem Developer
Excessive Long Hours Mean an Unsuccessful Project
Exceptions to the Rule
Case Study 9.1 Quake, StarCraft, and XCOM: Interceptor
Chapter 10 Roles and Divisions
Assigning Personnel
Management and Design Division
Programming Division
Art Division
Music and Miscellaneous Division
Support and Quality Assurance Division
Improving Morale and the Working Environment
Morale Boosters
Morale Booster Caveats and Warnings
Spreading the Risk
Chapter 11 The Software Factory
What Is a Software Factory?
Why Use a Software Factory?
Solving Game Development Issues
Case Study 11.1 The Effects of Losing Key Personnel
Case Study 11.2 Code Reuse
00 3634_FM_Intro
10:14 AM
Page xi
Table of Contents
Organizing a Software Factory
A Structural Overview
Group Responsibilities and Interactions
Case Study 11.3 Ineffective Problem Handling in Action
Case Study 11.4 Effective Problem Handling in Action
Case Study 11.5 The Benefits Of Tool Reuse
Applying the Software Factory Structure and Methodology
Getting off the Ground
Knowing When to Use Each Team—a Parallel Development
Rotating and Reassigning Team Members
Case Study 11.6 The Indispensables
The Suitability of a Software Factory
Smaller Teams
The Final Word
Chapter 12 Milestones and Deadlines
How Milestones Currently Work
Case Study 12.1 What Fuzzy Milestones Can Do to
a Project
Fuzzy Milestones
Milestones and Mini-Milestones
When to Use Milestones
Making Your Milestones Accurate
Case Study 12.2 The Costs of Canceling Projects
Checkpoint 1.0 General Requirements Gathering
Checkpoint 1.1 Technological Requirements Gathering
Checkpoint 1.2 Resource Requirements Gathering
Checkpoint 2.0 General Feasibility Study
Checkpoint 2.1 Technological Feasibility Study
Checkpoint 2.2 Resource Availability Study
Checkpoint 3.0 Draft Architecture Specification
Checkpoint 3.1 Project Initialization
The Next Steps
Defining Milestones
Bad Milestones
Good Milestones
Case Study 12.3 A Real-Life Milestone
Research Versus Deadlines
Evaluation of Milestones
00 3634_FM_Intro
10:14 AM
Page xii
Game Architecture and Design: A New Edition
Chapter 13 Procedures and “Process”
Testing in General
Case Study 13.1
Process Gone Mad
Procedures: Where to Use Them?
The Design Phase
The Development Phase
The Testing Phase
Source Control and Code Reviews: A Synergy
Case Study 13.2 Source Control? We Don’t Need
No Steenkin’ Source Control!
What Should Source Control Be Used For?
The Importance of Information Transmission
Proactive and Reactive Information Transmission
Chapter 14 Troubleshooting
Design and Architecture Problems
Case Study 14.1 The Case of the Deaf Manager
Schedule Threats
Case Study 14.2 Applied Schedule Readjustment
Organizational Problems
Contractor Problems
Personnel Problems
Development Problems
Process Problems
Chapter 15 The Future of the Industry
The State of the Industry
The First Era
The Second Era
The Third Era
Violence in Games
The New Model Developers
Case Study 15.1 It’s Hard for Developers
The Online Revolution
Delivering Games Online
Playing Games Online
00 3634_FM_Intro
10:14 AM
Page xiii
Table of Contents
Part III
Game Architecture
Chapter 16 Current Development Methods
The History of Development Techniques
The Rise and Fall of the Original Game Idea?
The Development Environment
The Present Day
Chapter 17 Initial Design
The Beginning
Case Study 17.1
Abstraction in Quake II
Hardware Abstraction
Graphics Hardware Abstraction
Sound Hardware Abstraction
Other Hardware Considerations
“Not Built Here” Can Be Better
The Twilight Zone
The Problem Domain
What Is a Game? (Revisited)
Thinking in Tokens
Tokenization of Pong
Tokenization of Pac-Man
State Transitions and Properties
Case Study 17.2 The Inflexibility Trap
Chapter 18 Use of Technology
The State of the Art
The Rise and Fall of the 3D Engine
The Perception of Technology
Case Study 18.1 A First Impression
Blue-Sky Research
Research Types
Case Study 18.2 Losing Sight of the Ball
Case Study 18.3 Tetris: A Caveat
Case Study 18.4 Outcast: Good Use of Technology
Keeping a Journal
Reinventing the Wheel
Use of Object Technology
The Pros and Cons of Abstraction
00 3634_FM_Intro
10:14 AM
Page xiv
Game Architecture and Design: A New Edition
Chapter 19 Building Blocks
Reusability in Software
Code Reuse
Case Study 19.1 Reuse of Engines
Design Reuse: Patterns
Game-Specific Patterns
Chapter 20 Initial Architecture Design
The Birth of an Architecture
Architectural Concepts
The Tier System
Tier Zero: The Prototype
Case Study 20.1 A Database-Driven Approach
Tier One and Beyond
Architecture Design
Applying the Tier-Based Approach to Architecture Design
Case Study 20.2 Discussing the Architecture of Warbots
Architecture Orthogonality
Chapter 21 Development
The Development Process
Code Quality
Coding Standards
Coding Priorities
Debugging and Module Completion
Types of Bugs
Case Study 21.1 Class A Bugs or Not?
The Seven Golden Gambits
Case Study 21.2 Reusable Architecture
Design First
Catch Mistakes as You Go Along
00 3634_FM_Intro
10:14 AM
Page xv
Table of Contents
Limit R&D
Know When to Draw the Line
The Three Lead Balloons
Bad Management
Feature Creep
Coder Insularity
Chapter 22 The Run-Up to Release
Late Evaluation
Final Analysis
Is the Game Pp to Scratch?
Case Study 22.1 A Self-Inflicted Disaster
Case Study 22.2 A Recovery Plan
Case Study 22.3 Licensing Hell
Case Study 22.4 Last-Minute Madness
Late Localization
Case Study 22.5
Case Study 22.6
Giving the Game Away
Keep Something Back
Case Study 22.7
How Did They Miss These!?
Focus Groups
The Web Site
Getting Ready for the Gold Master
Chapter 23 Postmortem
Case Study 23.1
Team Dynamics
Case Study 23.2
Case Study 23.3
A Tale of Two Projects
It’s All Gone Horribly Wrong!
Misjudging the Climate
Software Planning
Case Study 23.4 Oubliette
00 3634_FM_Intro
10:14 AM
Page xvi
Game Architecture and Design: A New Edition
Business Aspects
Case Study 23.5
Secure Your Revenue Stream
The Postmortem Postmortem
Chapter 24 The Future of Game Development
Part IV
Development in Context
Future Development
Case Study 24.1 Marketing Means Targeting
Case Study 24.2 Development Without Strategy
Small Is Beautiful Too
Building the Team of the Future
New Directions in Development
The Holistic Approach
“Jurassic Park” Software
Immanent and Transcendent Worlds
The Shape of Things to Come?
Sample Game Design Documents
Detailed Design Discussions
1. Balls! Introduction
2. Overview of Gameplay
3. Platforms
4. Time Scales
5. Why Puzzle Games Aren’t as Good as They Used to Be
6. Puzzle Game Appeal
7. Why Balls! Would Be Good
8. Game Design: User Interface Elements
00 3634_FM_Intro
10:14 AM
Page xvii
Table of Contents
9. Physics of Balls!
10. Blocks
11. Special Case Block-Block Collisions
12. Playing the Game
13. Further Embellishments
Initial Treatments and Sample Designs
Racketeers: Gang Warfare in the Roaring Twenties
1. Overview
2. Game Objectives
3. Graphics
4. Playing a Game
5. Character Types
Non-Gang Members
6. Personality
7. Orders
8. Combat
9. The Game World
10. Joints
11. Messages
12. Tutorial Campaign
13. Target Platform
1. Introduction
2. Game Elements
3. How Does it Play?
Technical Specifications
Technical Specification: Fully 3D Plug-In Graphics Module
for Balls!
Code Review Form
Test Scripts
Bibliography and References
00 3634_FM_Intro
10:14 AM
Page xviii
xviii Game Architecture and Design: A New Edition
About the Authors
Andrew Rollings has a B.S. in Physics from Imperial College,
London, and Bristol University. He has worked since 1995 as a
technical and design consultant spanning many industries.
Andrew lives in Auburn, Alabama, and can be contacted at
Dave Morris has worked as a designer and creative consultant on PC
and console games for several major publishers, most notably Eidos.
His strategy game Warrior Kings reached number six in the United
Kingdom PC charts. He has done creative development and scriptwriting on television shows for Endemol, Pearson, TV2 Norway, and the
BBC. He has also written more than a dozen novels, gamebooks, and
movie novelizations, and in 1991 he was the UK’s top-selling author. He is currently
writing the screenplay for the film version of the classic adventure game The Seventh
Guest. Dave lives in London, England, and can be contacted at
00 3634_FM_Intro
10:14 AM
Page xix
A work of this kind, drawing on our combined experiences over many years in the
games industry, owes a debt of gratitude to all the people we have worked with. The
impossibility of acknowledging everyone in person does not mean that we fail to value
every contribution, suggestion, or conversation that has helped us to refine these ideas.
So, let us start by thanking all who have been our colleagues on any development
project, great or small.
It is possible to single out a few individuals among the many. Roz Morris, though no
gamer, proofread the manuscript and made many valuable suggestions to improve its
clarity based on her professional expertise as a journalist and writer. As she is married
to one of the authors, it goes without saying that she also contributed a very great deal
of moral support.
Sam Kerbeck, that rare combination of gentleman and genius, gave us the benefit of
his technical advice, and we are indebted to him for ably clarifying many of the more
abstruse issues of architecture and coding. As Co-Founder and CEO of Turn3D, he has
provided us with state-of-the-art realtime graphics, and as a colleague of long standing,
he has also given his valued friendship over many years.
Ian Turnbull, former Development Director at Eidos Interactive, now Commercial
Director of Black Cactus Games, contributed enormously with his wise counsel regarding the economic realities of the industry. Without his guidance, this book would be
merely a theoretical work. It is Ian’s down-to-earth clear-headedness that reminded us
to make it more than that: a practical handbook for developers.
We would also like to thank Steve Foster, who has been extraordinarily patient over
the course of many speculative discussions, often stretching long into the night, concerning the future directions and methodology of game development. His contribution
has been much more than merely academic, however. When problem projects have
weighed us down, it has been Steve’s cheerful encouragement that has given us the
resolve to keep going.
Special thanks are due also to Leo Hartas, Tim Harford, Matt Kelland, Dave Lloyd, Tim
Gummer, Jamie Thomson, and David Bailey.
00 3634_FM_Intro
10:14 AM
Page xx
Game Architecture and Design: A New Edition
The fact that you are holding this book at all is due to the sterling efforts of the folks at
New Riders—in particular Stephanie Wall, who was also head honcho on the original
edition from Coriolis Press. Despite having had to chivvy us along once before, she
was willing to put herself through it all over again. We would like to thank Kristy Hart,
our editor, who is nothing short of a saint for her patience in tolerating broken promises and overlooked deadlines. And thanks also to our agent Jawahara Saidullah of
Waterside Productions, for making the whole thing happen in the first place.
Andrew would also like to thank his wife, Stephanie Park, for her continued encouragement and tolerance for his budding writing career.
00 3634_FM_Intro
10:14 AM
Page xxi
Tell Us What You Think
Tell Us What You Think
As the reader of this book, you are the most important critic and commentator. We
value your opinion and want to know what we’re doing right, what we could do better,
what areas you’d like to see us publish in, and any other words of wisdom you’re willing to pass our way.
As the Publisher for New Riders Publishing, I welcome your comments. You can fax,
email, or write me directly to let me know what you did or didn’t like about this
book—as well as what we can do to make our books stronger. When you write, please
be sure to include this book’s title, ISBN, and author, as well as your name and phone
or fax number. I will carefully review your comments and share them with the author
and editors who worked on the book.
Please note that I cannot help you with technical problems related to the topic of this book,
and that due to the high volume of email I receive, I might not be able to reply to every
Stephanie Wall
New Riders Publishing
800 East 96th Street, 3rd Floor
Indianapolis, IN 46240 USA
00 3634_FM_Intro
10:14 AM
Page xxii
00 3634_FM_Intro
10:14 AM
Page xxiii
Andrew Rollings’s Introduction to this New Edition
I must confess to being more than a little surprised at the success of Game Architecture
and Design. When we originally pitched the idea, back in 1999, we sent off proposals to
about ten different publishers. Only Coriolis, and more specifically, Stephanie Wall at
New Riders Publishing, got back to us. I bet those others are kicking themselves now.
I am especially pleased that, despite the implosion of Coriolis and the subsequent legal
adventures involved in ensuring the rights of this book reverting to us, Stephanie Wall
is still handling the book, but this time at New Riders. I’m very sure that after five
years of having to deal with me as a reluctant author, she’s more than fed up with me
by now.
So here we are with the second edition of Game Architecture & Design. Things have
changed in the four intervening years, though not as much as we’d like. We’ve come a
long way since 1999, but there’s still a long way to go. I’m sure we’ll get there…eventually.
I hope that this new edition of the book continues to serve as a useful reference for the
aspiring and professional game developer in the same way as the first.
—Andrew Rollings
Auburn, Alabama, July 2003
00 3634_FM_Intro
10:14 AM
Page xxiv
xxiv Game Architecture and Design: A New Edition
Dave Morris’s Introduction to this New Edition
The temptation when revising one’s work of several years past is often to rewrite
history a little. There are always those embarrassing predictions that come back to
haunt you. And yet with a few keystrokes, it’s possible to seem to a new generation of
readers as if we were always infallible. What an enticing position to be in.
In fact we have left most of our forecasts from the twentieth century intact. That’s
because in many cases—for example, the rise of middleware—we turned out to be correct. Frankly, like everyone else we enjoy being able to say, “We told you so!”
In cases where we were wrong, we move on and try to learn from the mistakes. In a
way, that’s at the heart of our design philosophy. Having a methodology can’t always
prevent you making a mistake, but it makes darned sure you don’t make the same
mistake twice.
The case studies are culled from our mutual experiences and those of colleagues.
People ask if these case studies could really be true. No, in fact not. The real truth in
almost every case was much worse!
But the encouraging thing is that the games industry is changing. Four years ago, our
rallying cry for a formal development methodology rang like a lone voice in the wilderness. Nowadays games development is becoming a much more structured process. And
publishers have a better understanding of what the process entails. In another four
years, a developer coming across the first edition of Game Architecture & Design will be
astounded that development could ever have been so ramshackle. We are happy to
think that, in however small a part, we helped contribute to this evolution of the
Even better, as the production process becomes better understood and more streamlined, it consumes less of the developers’ time. The extra creative energy this frees up
can now be devoted to the game content itself. We are starting to see the first signs
that games really are moving from being the equivalent of silent movie one-reelers.
They are acquiring depth, beauty, and emotion.
00 3634_FM_Intro
10:14 AM
Page xxv
Tolstoy wrote, “Art is not a handicraft, it is the transmission of feeling the artist has
experienced.” Great art doesn’t simply entertain you—it does that, granted, but it does
more. Art changes your life. In the next decade, we will see videogaming’s Birth of a
Nation and the Citizen Kane of the console generation.
Bliss it is in this dawn to be alive!
—Dave Morris
London, England, July 2003
00 3634_FM_Intro
10:14 AM
Page xxvi
xxvi Game Architecture and Design: A New Edition
Introduction to the First Edition
The philosophy behind this book is simple, stark, and absolutely true. If you are failing
to plan, then you are planning to fail.
Of course, games are unique. Game development constantly throws up unexpected
issues. Coping with accelerating technology is like holding onto the tiger’s tail. Worse,
sometimes the client revises their requirements halfway through.
But these are not reasons to forego planning. A good design provides a goal to aim for
that will guide you when changes are unavoidable. Frequently, the design anticipates at
least the domain of future changes. A full project plan establishes a framework for
change. Planning does not end where development begins. Rather, you sustain the plan
through any changes that need to be made so that, although the target may shift, you
never lose sight of where you’re going.
So, yes, games are unique. For that reason, they require a unique kind of planning.
That is the methodology that we have set out in this book. To illustrate these points,
we make copious use of case studies. These are based on common circumstances in
the industry, but are fictional—any similarity to real company or product names is
unintentional, except where explicitly stated otherwise. (In addition, we have referenced several trademarked games and replicated trademarked images such as Pac-Man,
owned by Namco. These references are included to enhance the instructional and
visual delivery of game concepts specific to this book and should not be construed as
a challenge to such trademarks.)
Who should read this book? The short answer is everyone who is, or intends to be,
involved in game development. All members of a team can benefit from understanding
each other’s area of responsibility. However, each section addresses issues specific to
one part of the development team. We recommend that you begin by reading the
section of primary relevance to your own expertise.
00 3634_FM_Intro
10:14 AM
Page xxvii
Introduction xxvii
Part I: Game Design
The book treats pure game design separately from architecture and formal planning.
The game design constitutes a feature-based description of the end product that can
be used as a shared creative vision by the whole team. As development progresses and
change becomes obligatory, it is the designer’s task to evolve the game design also so
that the intent of the project remains clear.
We reject the assertion that gameplay is entirely unpredictable and thus cannot be
designed. In this section, well-understood techniques from game theory, mathematics,
and creative writing are applied to the design process in order to elaborate a development model based on successive iterations of the product.
It is clear that many games contain a spark of gameplay brilliance that could have been
nurtured into a flame if the designers better understood the basic issues of gameplay.
Part I shows you how to achieve this in your own designs.
Part II: Team Building and Management
The book advocates formal process because we have found it to work. Many developers are wary of formal process because they fear it will lead to bureaucracy and overmanagement. In fact, the reverse is true.
Consider an analogy. In the Japanese martial arts, much emphasis during training is
placed on constant repetition of predefined sequences of movement, called kata. The
student may wonder how a formal routine of countering blows to the head, chest, and
abdomen can possibly be of use in the infinite combinations of moves that can occur
in real-life. And then, one day, somebody takes a swing at you and your arm comes up
to block. You didn’t even need to think about it.
Similarly, we espouse the application of formal process precisely because it lessens the
need for management. Instead the emphasis is on honing everyone’s skills as part of a
team, with the developers themselves taking responsibility through self-management.
Thus, Part II details simple, common-sense procedures that are easy to adopt and will
soon become second nature. The payoff will be seen in increased efficiency, reliability,
and team morale.
00 3634_FM_Intro
10:14 AM
Page xxviii
xxviii Game Architecture and Design: A New Edition
Part III: Game Architecture
The architectural plan of the project is the point of contact that draws together the
conceptual, artistic or gameplay factors with the technical requirements. Envisage the
design as an ideal view of what the game should be. The architecture maps out how to
converge reality with that ideal view.
A perfect architecture should aim to achieve all of the following:
Separating the project into completely encapsulated modules allows each to be
independently tested, modified, or replaced without impacting the rest of the system.
Reinventing the wheel every time makes no sense. Modules should be designed to be
extensible and portable so that they can be plugged into other projects.
A high degree of reliability is best attained by building architecture free of module
interdependencies. The final goal is a meta-architecture capable of building games
that are always able to function in unexpected circumstances—in short, that are crashproof. Though such a goal is considerably beyond the scope of this book, the subject
is treated in an introductory manner by means of object-oriented design patterns.
The project plan is derived directly from the architecture, yielding a schedule that
allows realistic measurement of progress.
Projects fail because of poorly thought-out architecture that fails to assist in creating
the intended game or, worse, constrains development in an unsuitable format. We
show how design shades into architecture, which shades into the technical design and
documentation, which in turn shades into code—as a continuous cohesive process
always directed towards the (evolving) vision described in the game design.
00 3634_FM_Intro
10:14 AM
Page xxix
Introduction xxix
Part IV: Appendixes
Here we use real-life projects as case studies to illustrate the techniques of the book.
We show how to begin with a feature-based description of the desired end product—
the vision document that we call the gameplay design. When this is mapped to a logical
abstraction of the game environment then you have the software plan, which describes
discrete modules of the game and how they will interface. Finally, the details of implementation are added to create the technical design of the project. Then the coding
Summing Up
Game developers are the most enthusiastic, creative and motivated workers in the software industry. But they have been ill served by the development models in place today.
Too often, the development teams are left baling water when it would make more sense
to fix the leak instead.
Many development houses have finally seen the need for change, but few know what
changes are needed. In this book we set out a new development model for game software. Beginning with the core concept, the book covers all you need to know to organize a team, plan your project, and commence development with confidence.
Rest assured, these are not abstract theories. We have applied our methodology in
practice with great success. This development model is one that works. Our aim is to
tell you how to fix that leak so that you can get on with the important business of creating games. Good luck!
00 3634_FM_Intro
10:14 AM
Page xxx
Game Architecture and Design: A New Edition
This book follows a few typographical conventions:
A new term is set in italics the first time it is introduced.
Program text, functions, variables, and other “computer language” are set in a
fixed-pitch font—for example, insert example from book here.
Code lines that do not fit within the margins of the printed page are continued
on the next line and are preceded by a code continuation character ➥.
01 3634_Part 1
10:20 AM
Page 1
Part I
Game Design
Chapter 1
First Concept
Chapter 2
Core Design
Chapter 3
Chapter 4
Detailed Design
Chapter 5
Game Balance
Chapter 6
Look and Feel
Chapter 7
Wrapping Up
Chapter 8
The Future of Game Design
01 3634_Part 1
10:20 AM
Page 2
02 3634_CH01
10:21 AM
Page 3
Chapter 1
• Types of games
First Concept
• Storylines
• Originality
• Feasibility
his chapter focuses on how to assemble the basic gameplay treatment, which is really just a document of no
more than five or six pages. It is a proto-manifesto to
communicate your concept for the game, and, most of all,
your feeling of what the game would be like to play.
Eventually, the treatment will grow into a full specification
for the game, but writing this initial treatment will help you
solidify your thoughts. You can return to the treatment later,
whenever you feel in danger of getting lost in the details of
the design, to get back on track with your original idea. Even
better, the treatment is something you can use to sell the
game to others—the publishers, programmers, and artists
who will be developing the game.
To begin, you’ll learn ways to find, shape, and refine your
concepts. We’ll view a sample treatment and examine why
the game took the form it did.
The Shock of the New
The designer’s job is to create something new. Everybody
thinks his or her job is difficult, but this one truly is. The
poet from the Book of Ecclesiastes says, “Under the sun there
is no new thing.” If the statement is true, isn’t it the kiss of
death to creativity? Why bother to try to create anything if it
isn’t going to be original?
02 3634_CH01
10:21 AM
Page 4
Chapter 1
Of course, it’s not all bad news. Occasionally there is the opportunity to rework one of
the old ideas in completely new ways, and we are lucky to live in an age that is rife
with these opportunities. The advent of computers has allowed us to formulate,
calculate, and express our ideas in new ways. And the field of computer gaming has
provided us with a new means of recreation, taking all the old ideas and obsessions
that have occupied mankind since long before Ecclesiastes and giving them a new spin.
Artists of every generation have drawn on the ideas and works that have come before
them. Your own creativity is a result of the sources that you borrow from, the unique
mix of ideas that synthesize in your own work.
And there’s something else that’s encouraging to the creative artist. Ideas aren’t like
seams of ore; you don’t deplete them and then they’re gone. Rather, they’re like living
things. Ideas that are overused can become inactive for a while. They regenerate.
So when you’re searching around for a first concept, take the path less traveled. Start
by looking for inspiration where others haven’t been for a while. Bad game concepts
are mere plagiarism: “I’m going to do Command & Conquer but with more light-armor
vehicles.” Good game concepts bring something fresh: “It’s a Frankenstein strategy
game where you plunder graves and battlegrounds for spare parts.”
Some ideas may be just too plain wacky to work, which doesn’t necessarily mean that
they are bad ideas. Maybe their time has not yet come, either because the technology
isn’t there (Wolfenstein came first, but Doom did all the business) or because the market
has not yet been created. Would we have been ready for Pac-Man two years earlier?
The Creative Road Map
There are three kinds of skill involved in creating any new work. These are creativity,
craft, and technique. They are quite different from each other, but they overlap. Take
painting. The first step is to decide what to paint. Then you plan how you’re going to
do it, probably by sketching out a rough in charcoal. The third stage is the actual
02 3634_CH01
10:21 AM
Page 5
First Concept
In creating a game, the stages are: concept, structure, and design. For example,
suppose you wake up one morning with the idea of doing a game based on War &
Peace. That’s the concept. Next you decide that because war lends itself to action—and
hence interaction—you will focus the gameplay itself on the periods of conflict. The
cutscenes will be flashbacks to a time before the war. Hence the cutscenes won’t fit
chronologically in place with the game levels, their purpose being not to advance to
“plot” of the game but rather to give context to the action in the game levels by gradually revealing the characters’ backstories. That’s your structure. Lastly, you start planning out the core game mechanics—what the player sees on screen, the actions he or
she can take, the objective, and the “play intent” of the game. And that’s the design.
In most art forms, the third kind of creative ability is relatively common. Many painters,
given a plan to work from and regular direction, can execute the details of the work.
This is how the great artists of the Renaissance were able to create such prodigious
works—they had a staff of brush technicians and paint mixers working under them.
The ability to structure a work is less common—but it can be learned. A host of books
on how to write bestsellers or hit screenplays all exploit the fact that you can teach
somebody how to structure. Hollywood movies tend almost exclusively to use the
three-act redemptive, or restorative, story structure because it is a formula that works.
It might not make a great movie, but it’s like a table that you store in your garage. As
long as the legs are the right length and the surface is planed smooth, it may not win
any design awards, but it’ll do for stacking your tools on.
The first skill is raw creativity. Almost by definition, it is extremely rare. It’s also not
something you can learn, although you can hone creative ability if you have it to begin
with. There are many things that can go wrong with game development, but perhaps the
most common and avoidable pitfall is a lack of creative vision. The successful games
companies are the ones that recognize that their business is entertainment, not software:
“Bruno Bonnell, Atari’s chairman and chief executive, maintains that the [games]
business—worth some $30 billion (£18bn) worldwide at retail prices—is breaking free
of its hard-core audience and penetrating the mainstream market. Accordingly, he says:
‘We need technical and entertainment values at the same quality level as in other mainstream media like movies and TV.’”
02 3634_CH01
10:21 AM
Page 6
Chapter 1
“This is not an altogether welcome development. ‘I like it and I fear it,’ he says. It has
already inverted the basic cost structure of his business. ‘It used to be two-thirds of the
budget for technology and one-third for the creative,’ he says. ‘It is now the reverse,
and in future it will be a quarter for coding and three-quarters for the entertainment
—Christopher Parkes writing in the Financial Times, July 2, 2003
As competition grows among the leading games developers, the semi-technical skills
of structuring and designing gameplay are becoming less important than the raw
creativity required to come up with world-class characters, settings, and stories.
The future belongs to the shaman. The visionary. The dreamer.
Having the Idea
How many industries can claim to deal in daydreams?
Dreams are where every game begins. Before the code, before the software plan, before
the concept artwork, even before the first document, the game starts life as a spark in
the designer’s imagination, and the idea is the single most persistent entity in the game
development cycle. It can evolve and develop as the game progresses, but it was there
from the start. It is the seed from which the game grows.
Just as programmers are often warned not to rush into coding (as we’ll see later in the
book), designers must guard against rushing to get their ideas down on paper. When
that daydream comes, give it time. Resist the urge to go straight to the word processor.
We firmly believe ideas have a gestation period, the time your subconscious takes to
mull over the concept and hand it up to you in its raw form. Writing anything down
too soon warps the process because it allows your critical and analytical faculties to
come into play, which can kill the idea before it’s been fully formed. When you start to
hear quotes from the treatment document in your head and when you’re scribbling pictures to show what the game screens will be like—then it’s time to start.
So indulge your subconscious mind and kindle your creative spark. Author Stephen
King refers to these subconscious thought processes as “the boys in the basement.”
Allow yourself to daydream, but, when you have finished daydreaming, get ready for
02 3634_CH01
10:21 AM
Page 7
First Concept
some serious effort. Edison said that genius is 1% inspiration and 99% perspiration.
Well, enjoy this 1%; everything after is pure hard work.
We’ll deal successively with the four phases of the creative process:
Inspiration—Where to get ideas
Synthesis—Combining ideas
Resonance—Creating synergy from ideas
Convergence—Finishing the concept
Early film directors were electrified with the possibilities of their new medium. Casting
about for ideas to put up on the screen, they drew their inspiration from anywhere and
everywhere. German Expressionism, a creative lineage that can be traced right down to
Tim Burton today, mixed elements of fairy tales, theater, and Freudian psychology.
Kung Fu movies are Chinese opera without singing. Star Wars is music hall drama by
way of Saturday morning serials.
While directors are happy to take ideas from everywhere, games designers have not
been quite so adventurous. Except for a few notable exceptions, their inspiration comes
always from the same safe sources. Only so many shoot-’em-ups can be inspired by
Aliens and still seem even remotely fresh. Computer Role Playing Games (CRPGs) based
on the 1970s game mechanics of Dungeons & Dragons are dated. Tabletop role-playing
has progressed so far since then. It’s time to look to new pastures for fresh ideas.
Nearly every game released today seems to be related to the others in the market.
Looking at a single genre, you don’t so much see a rich tapestry of diversity as the
same game repeated in slightly different forms. A herd mentality dictates that only
“safe” games are released. For safe, read unoriginal: “It’s like Medal of Honor, only it’s
set in the Korean War.” And yet, is that really the safe approach? The polymath designers like Will Wright, Shigeru Miyamoto, and Peter Molyneux draw on a very broad
range of sources for their inspiration. Their game concepts are always fresh—and
always successful. There’s a lesson to be learned.
02 3634_CH01
10:21 AM
Page 8
Chapter 1
Creative motherlodes are waiting to be mined. So start with a brainstorming session in
which anything goes. Leaf through a Victorian encyclopedia. Go to an opera. Browse
obscure sites on the Web. Watch a movie with subtitles. Sit in on a random lecture
at your local university. Go scuba diving in Florida or hill-walking in the Pyrenees.
The movie director, Ang Lee, draws inspiration from his sand and rock gardens. Do
anything that will help you get rid of the stale thinking.
Originality can present itself in many aspects of a game: the gameplay, the story, the
setting, the characters, the interface, or in the technology. If you can bring a freshness
to most of these, you can make a great game. If you manage to make all of them new
and exciting, you’ve probably just spawned a new genre!
Stephen King says that the spark of originality comes when you put familiar things
together in an unexpected way. A vampire game? Don’t put your vampire in a cyberpunk city (that’s old) or in a Transylvanian castle (that’s ancient). How about a pirate
setting? There aren’t any bats flitting about Gothic battlements, but the vampire can
transform into a shark instead, as well as into rolling banks of sea fog. By day, rather
than sleeping in coffins, the vampire has to return to rotting hulks on the sea bed. Or,
for something more exotic, you could rework the vampire theme using the style and
tropes of Indonesian shadow puppet theater, a whole mythology that is ripe for the
Make sure, though, that the idea has some public resonance. For example, Grim
Fandango used the Mexican “El Dia de los Muertos” or “The Day of the Dead” as its
theme, and it worked effectively because its mythology is known worldwide.
It isn’t enough to have an idea—or even to have lots of ideas. You have to make them
work. Otherwise, your game may trundle out of the metaphoric hangar and taxi down
the runway, but it’s never going to get airborne.
An old saying claims that new concepts are produced by “chimeras bombinating in a
vacuum.” In ancient myth, a chimera was a monster that had body parts from several
different animals. Think of that as the raw ideas you’re throwing into your game. To
ensure that the hybrid isn’t stillborn—and, more importantly, that it produces something new—you need to pick the right mix of parts.
02 3634_CH01
10:21 AM
Page 9
First Concept
Let’s go back to vampires for a moment. Say you’re fixated on the idea of a vampire
game, and for other reasons (possibly commercial), you want an outer-space setting (a
starship, for example).
One possible scenario for such a game would be to have the starship visit a backward,
superstitious planet and take delivery of a stack of wooden boxes. Later, with the starship under way, a box opens and a cloaked figure emerges. Why wouldn’t you do that?
Because it’s boring! Why bother to transplant vampires into space if you’re just going
to rehash the old clichés? Instead, you need to think about how you’re going to synthesize a game out of the two concepts.
Vampires sleep in coffins, but that doesn’t have to literally mean a wooden box. What if
the crew of the starship goes into a “cold sleep” when they don’t have duties to perform? The vampire could then be a member of the crew, drawing frozen blood samples
from his slumbering shipmates whenever he can. As an adventure game, it’s not entirely original (shades of Alien), but it isn’t exactly a cliché either.
Everyone also knows that vampires come out only at night. Does that mean a certain
distance from a nearby star, or will you say that it is always night in space? How about
in hyperspace? Does that count as night or day? Perhaps the vampires can’t stay close
to bright stars for too long, which immediately may give you some thoughts on how to
develop the game balance if it’s going to be a strategy game in which the vampires are
one of the player races. In the context of a survival horror game, on the other hand,
the same idea might lead you to decide that the vampires can’t enter the engineering
bay (too much UV from the warp core), and that this fact will function as a potential
clue to the vampire’s identity.
Look at the ideas in the mix. Ask yourself what each idea can contribute to the others
so that the emergent game makes full use of all of them.
In an outstanding episode of his Sandman comic book series, Neil Gaiman has his
heroine visit Paris at the time of the French Revolution in search of the severed head of
Orpheus, the legendary Greek hero. Decapitation gives Gaiman his plot link. One
memorable scene in the story takes place in a room piled high with heads. Most of the
heads have come from the guillotines; only Orpheus’ can speak.
02 3634_CH01
10:21 AM
Page 10
Chapter 1
The concept is disturbing in itself, but what gives this story its power is the resonance
between the elements Gaiman has chosen. The Sandman comic often features the dream
world: Paris during the Reign of Terror is a place out of nightmare. The story takes place
when the executions are out of control, a hysterical bloodbath. Civilization is being
destroyed by insane violence, and a flashback reminds us of Orpheus’ fate. (He was torn
apart by madwomen.) The story’s title is Thermidor, the Revolutionary Convention’s new
name for the month of July. The Convention’s aim was to expunge classical traditions
and make the world anew. But, as Gaiman shows, it is never as easy as that.
Resonance is a way of making the whole greater than the sum of the parts. This is the
definition of synergy, and it is an effective means of making the story and the subject
matter significant to the game players. It applies to game design as much as to any other
creative work. Adventure games, built around a clear story line, will benefit most.
Begin by deciding your theme. In Warcraft Adventures, for example, the hero was to be
a young orc brought up after the victory of the human and elfish alliance. Cut off from
their homeland, the orcs are a subjugated race confined to reservations. The hero’s
quest is to regain face and win back the honor lost in defeat. In the circumstances and
even the tribal costume of the orcs, the game’s designers appear to be evoking comparison with American Indians. You may question the social commentary and indeed the
appropriateness of swiping from real human cultures to depict a nonhuman species,
but the resonance is undeniably effective.
You will also find that people create their own resonance if the game is powerful
enough. Take a look at the Half Life Alternative Background Story Web site: This site is where people have contributed to
creating a rich tapestry of fictional stories, filling in details on the background of the
world(s) that Half Life is based in and around.
Up to this point, we’ve been encouraging you to ignore your critical and analytical
instincts. You simply can’t give birth to new ideas while worrying about whether or not
they’re any good.
02 3634_CH01
10:21 AM
Page 11
First Concept
But now we’re passing out of the artistic stage and into the craft of games design.
Craftsmanship demands a critical sense. You have to be able to judge if your ideas are
going to work together to make a good game. Good critical judgment can come only
from experience. Look at every game that’s remotely like your new concept. Do those
games work? If not, why not?
Obviously, it’s far better to anticipate flaws so you can correct them now. Once the
whole programming and art team is assembled, changes in design are expensive and
paid for by the hour. An accurate critical sense can save thousands of dollars further
down the line.
Shaping the Idea
According to the common model of drama, there are five elements: style, plot, character, setting, and theme. These same elements have been combined over the last twenty
centuries in theater, opera, novels, films, and television. And they still work today.
Dramatic Effect
It might seem strange to claim that computer games can be thought of as a dramatic
form. Doesn’t interactivity mean that you can’t simply apply the traditional analysis?
The player is in control of the game, after all. But we would say that an interactive
game is really no different from a work of more “classical art.” For example, if you
read the epic poem, The Iliad, you construct your own unique narrative, which
certainly differs from what Homer had in mind. Even puzzle games like Tetris in one
sense tell a story and are thus “dramatic” to an extent.
All good computer games must entertain, and most gain a great deal of their entertainment value from drama. The fact that the player has a greater control of the drama
than in any other kind of entertainment is a difference in degree, not form. The underlying rules of drama therefore still apply, because they are aesthetic rules dictated by
the human mind. So it’s instructive to look at the dramatic elements of a game, which
we will do in the following few sections.
02 3634_CH01
10:21 AM
Page 12
Chapter 1
How do you define a genre? Is the first-person shoot-’em-up a genre all its own? Did
Dune 2 really spawn a new genre when it gave rise to all those realtime strategy games?
No a genre is much broader than that. In literature, “fantasy” is a genre; Tolkien
rip-offs—ubiquitous though they are—are just a variant in style. It’s similar with
games. The genres are identifiable by the goal of the designer, whether that was to
frighten you, engage your intellect, or delight you with the beauty of phantasmagoric
scenes. We’re doubtful that taxonomy is of much value in these cases, but we could
say the broad genres are:
Action—Lots of frantic button pushing
Adventure—The story matters
Strategy—Nontrivial choices
Simulation—Optimization exercises
Puzzle—Hard analytic thinking
Toys—Software you just have fun with
Educational—Learning by doing
Obviously, this is not an exhaustive list and some of the list members can be combined
with others. For example, Alone in the Dark is primarily an adventure game that combines elements of action and puzzle games. It would nowadays be categorized as “survival horror,” which is treated as a genre in its own right but is in fact a formalized
instance of the adventure genre. Games, like films, can and do straddle multiple genres.
The equivalent of what Aristotle termed style is the way the game is executed. At heart,
both Deus Ex and Tomb Raider are action-adventure games, but their styles make us
perceive them as different genres. And role-playing games (RPGs), we would argue,
aren’t a genre at all; they are all fundamentally Dungeons & Dragons clones, adding an
element of simulation (existence in a fantasy world) to an action-adventure format.
02 3634_CH01
10:21 AM
Page 13
First Concept
The worst kinds of game are those (usually adventure games) in which it’s obvious
that the designer thought he was writing a film or a novel. It’s the frustrated-author
syndrome, and you know you’re bearing the brunt of it when your choices in the game
have shrunk to one linear path, or when you find yourself scrolling though neverending cutscene dialogues waiting to get to the action. These designers are entertaining
nobody but themselves.
Choose-your-own adventure gamebooks, an early low-tech form of adventure game,
aimed for a minimum of three main routes through the narrative, with smaller branches along each main route and several ways to solve each problem encountered along
the way. Compare that with a modern computer adventure like Grim Fandango. It has
one route through the story, and only one correct solution to every problem. You may
enjoy it for the artwork, music, even the story (all of which are breathtaking), but this
is in effect is just a film that you’re having to solve puzzles to view. Every time you
solve a puzzle, you are rewarded by being shown a little bit more of the movie.
Interactive fiction is possible, but the end result is not a game as such, and this is
the wrong kind of interactivity to include anyway. (There will be more on this in
Chapters 3, “Gameplay,” and 8, “The Future of Game Design”).
There is plot in any game, but for the most part it is created by the player himself. It is
the player, not the game’s designer, who is the author of the game’s events. The game is
a tool for allowing the players to create stories.
You can see this in any game. Ask an Age of Empires enthusiast to describe some of his
games. Most likely, he’ll start telling you about the two priests he sent in to convert a
key enemy tower while his archers created a diversion. Or how his scout led a lion
right into the enemy camp. These are stories. Sharing them with each other is not so
different from Neolithic hunters describing the day’s adventures around a campfire,
or Homer recounting the Trojan War.
There is often a backstory—the scene and setting that is particularly important in
adventure games. The designer must bear in mind that the players are impatient to
begin playing. They haven’t paid $40 to sit through interminable Full Motion Videos
(FMVs). They want to get their hands on the game. A good designer makes his ego
transparent in the game.
02 3634_CH01
10:21 AM
Page 14
Chapter 1
Also it’s important to square the need for scene setting with the player’s unfamiliarity
with your game world at the start. Notice how the best games minimize how much the
player has to wade through. In Half Life, it’s your first day at work, which neatly
explains away why you know so little. Ecstatica makes its setup with startling simplicity: a fairy-tale village, the bridge collapses behind you, and the game begins. Resort to
cliché if you have to; just get the story started. As long as the gameplay is original, no
one will fault you for skimming through the setup.
Publishers and designers have different reasons for favoring games with identifiable
characters. From the publisher’s perspective, a character is a merchandisable commodity.
Once you have a Lara Croft, the fact that she originated in a game is irrelevant. She
belongs to the marketing department, who can leverage her everywhere.
The designer’s motives are different. A character enhances the story (that the player is
creating), and this applies to the player’s own characters as well as the ones you write
into the game.
You think SimCity doesn’t feature a character? Wrong. The character is the player, who
is a god—albeit of a very restricted world. You think the god in question could be any
character, that the designer doesn’t have any say in it? Wrong again. The choices that
the game presents to the player define the kinds of character that the player can be.
The god of Populous is not a kindly old guy with a beard!
Think of the leafy glades of Warcraft, the barren war-ravaged landscapes of Command
& Conquer, the dark gothic catacombs of Dungeon Keeper. Who lives there? What are
they like? If your players are asking themselves these questions, then you’ve got them
How about puzzle games like Tetris? Do they have a setting? Of course they do. It is a
formalized universe governed by a few logical rules: the landscape of the reasoning
02 3634_CH01
10:21 AM
Page 15
First Concept
The theme of a dramatic work is the philosophical idea that the author is trying to
express. You can think of it as the defining question of the work. Can love triumph?
Is murder ever justified? Are dreams real?
We’ve already said that the true author of a game narrative is (or should be) the player.
Level design can steer the player towards the favored themes of the designer, but it’s
like leading the proverbial horse to water. You can’t force the player to think your way.
If a game is to have themes, they should come as a mix-‘n’-match bag that leaves the
player with freedom of choice. This doesn’t apply only to adventure games. Starcraft’s
different species and the levels of each species’ campaign tend to slant towards a
theme. By one Freudian reading, the Terrans are the ego, the Protoss are the superego,
and the Zerg are the id. It doesn’t matter if you see the theme. It doesn’t even matter if
the designers thought of it themselves. But it shapes the way you think about the
species you’ve chosen, and so it shapes the way you feel about the game.
The Treatment
So, your game has wheels, but can you drive it around the block? The only answer is
to put it to the test, and the first test of your concept comes now, when it’s time to put
it all down on paper. When you start to do this, you’ll see the previous gaps in your
thinking. Maybe you hadn’t considered the multiplayer issue, or how complex the
interface is going to get. Writing it down exposes these points, and it can come as a
short, sharp shock after the heady indulgence of your flight of fancy. After you’ve
written the treatment, you’ll be able to see if this game has potential, or whether you
should maybe give up on it right now.
When you’re writing the first treatment, you can cut loose and put in all those great
ideas you’ve had. At this stage, you don’t need to worry about technical constraints.
In fact, you should deliberately avoid thinking about them. The devil is in the details,
and this is supposed to be a vision statement for the game. Remember the saying,
“Programmers can’t see the wood for the trees; designers can’t see the trees for the
wood.” And this is exactly how it should be.
02 3634_CH01
10:21 AM
Page 16
Chapter 1
What points should the treatment cover? We don’t believe a checklist is useful creatively. It would be nice if creativity were a straightforward process that could be
learned, but (in spite of all those “How to write a bestseller” books) we have to tell you
it just isn’t so. Again, the best advice is to let your instincts steer you. The points you
find yourself covering in the treatment are the ones that make your concept special.
The details may get pushed into the background for the time being, but it doesn’t matter. They can wait until you do the full spec.
So, you have written your treatment? Hardly like working at all, was it? Enjoy that bit
because it won’t get that breezy again until you’re playtesting the beta version.
Taking Stock
You’ll recall we previously said that we didn’t like checklists as part of the creative phase.
But now it’s time to get critical, and there are some points you’ll need to consider:
Now turn another eye on your beloved concept, an eye of cold, objective criticism.
Deconstruct the concept. Pick it apart. Think about what’s in there—the ancestors of
your chimera—and what motivated you to put them there. Here are some pointers to
get started:
What is the game genre(s)?
Which existing games most closely resemble your concept?
In what way are you proposing to do things the same as in those games?
In what ways are you going to do things differently?
02 3634_CH01
10:21 AM
Page 17
First Concept
Ask yourself if your game is different enough. Particularly if it’s an entry into an
already crowded genre (such as first-person shooters), it will need a good strong list of
unique selling points (USP list). But just because a feature is original and unique is not
sufficient reason to include it. Take another look at the USP list.
Will this feature be fun? (This is always rule #1.)
Will this feature lead to good gameplay?
Why hasn’t anybody used this feature before? (Is it really that you were first to
think of it?)
Is it realistic to expect the team to be able to implement this feature?
Will the feature be workable? (A player can handle only so much with three
mouse buttons or a console controller.)
These factors begin to touch upon the concept of feasibility, which we’ll discuss in a
moment, but, for now, we’ll just concentrate on the issues involving gameplay.
After working up the treatment, you will typically present your concept to both the
client (publisher) and the development team. Both must buy into it if the project is to
succeed. If you thought you’d critiqued your own brainchild thoroughly, just wait until
you hear the comments from your team. You must be ready to explain and defend your
ideas. It’s essentially akin to the modern scientific method: advance a theory and see if
anyone can demolish it.
Case Study 1.1 shows a one-page treatment that illustrates the key points you need to
be covering at this stage. Note that the one-sheet begins with a hook, hopefully to grab
people’s attention. (In this instance the opening resonates with the game’s underlying
theme: the passing of great glories.) It then elaborates the theme and intended playing
experience and follows that with a brief outline of how the game design should deliver
that experience. This should be enough to decide whether it is worth proceeding to a
more complete treatment.
02 3634_CH01
10:21 AM
Page 18
Chapter 1
Case Study 1.1 The One-Page Pitch
An old man in a tunic is being guided by a boy. The old man’s eyes are closed. We’re
in extreme close-up—we see the boy’s fingers holding the man’s gnarled hand, a
close-up on scuffed sandals, the old man’s cane tapping the packed earth floor as he
goes. The background is in darkness, but we get the sense of a huge interior space,
perhaps a hall.
They reach the center of the hall. The old man steadies himself, leans on his stick.
The boy scurries off. The man stands there, head bowed, gathering his energy. His
back straightens. He lifts his head.
We’re close on his face. Think of Richard Harris. The old man takes a breath and
starts to speak. He has a voice of surprising power:
He opens his eyes. They are milky white, clouded over. Blind. But he sees with his
mind’s eye.
“Goddess, fill my lungs with breath. Give me the words to tell of Achilles’ fury.
Murderous and doomed, it was a fury that cost the Achaeans so many men and sent
brave souls into the underworld leaving bloody limbs for dogs and birds to pick
apart. Thus was the will of Zeus….”
As the old man speaks, our view switches behind him in a wider shot and as the
camera rises up we see a hundreds of men seated at long wooden tables, lamplight
picking out details in the gloom of this huge hall. They listen in utter silence as we
cut to—
A beach in dazzling sunlight. In close-up as crystal-clear water gently laps the shore
and then the prow of a longboat drives into the sand, we hear shouts and a clamor
of jangling war-gear. A sandaled foot jumps down from the boat, leaving a strong
imprint as the warrior strides ashore and we pan up to see—
Achilles, greatest of the Greek heroes, standing tall against the brilliant blue sky.
There is a proud smile on his face as he surveys the hinterland. He has no need of
02 3634_CH01
10:21 AM
Page 19
First Concept
armor; the gods have made him invulnerable to harm. He has just arrived and
already he is eager for battle.
The game is Troy. It’s a wargame—but not like any wargame that has gone before.
Players will take the role of various legendary heroes like Achilles, Odysseus, and
Ajax. Each hero is accompanied by a war-band of non-player characters whose
morale and fighting ability will reflect how well the player is doing.
This is an epic story. It’s also elegiac. The world is never again going to see heroes of
this caliber. They are favored by the gods, are almost gods themselves. Men like the
blind poet, Homer, will sing of their exploits for thousands of years. But the war is
destined to end in the deaths of most of the heroes and the destruction and pillaging of the beautiful city of Troy. It is a tragic time. The end of an era.
We intend the game to reflect this. Just as the original Iliad poem interweaves
themes of rage and melancholy, glory and waste, this is the first videogame that will
convey both the excitement and the tragedy of war.
How is this achieved? We said before that the actions of the player heroes affect the
rank-and-file non-player soldiers. (Think of Dynasty Warriors, for example, but with
more varied AI among the soldiers.) A war-band that has lost its hero will start to
degenerate into a rabble. They will lose their tight formations and rigid discipline. They
will start to skulk away from the hard fighting. When an enemy is struck down, instead
of moving onto the next foe, they’ll wait around to loot the body. After the death of
honor, these men will become the carrion dogs that Homer spoke of in the intro.
Hence, as the game progresses and heroes fall on both sides, the war moves from
being a glorious adventure to a dirty, desperate, vicious struggle for survival. The
gods, who are initially willing to provide aid, advice, and magical weapons, withdraw
from involvement as more heroes die. Even the graphics depict the dying of the
light. To begin with, bright colors evoke a time of glory. As the heroes are whittled
down, the images become more gritty, the colors flatter and desaturated. It is as if a
Technicolor epic like Spartacus were gradually turning into Saving Private Ryan….
02 3634_CH01
10:21 AM
Page 20
Chapter 1
Try not to get touchy. Creative people very rarely enjoy negative criticism, but if you
can’t marshal arguments in support of the creative decisions you’ve made, how do you
expect your team to buy in? Even more importantly, you need to understand that the
development process has now begun. Simply gathering the team and discussing the
treatment has already set up a “thought testbed” that is the first stage in proving (or
junking) your idea.
We’ve assumed thus far that you’re designing your game in an ivory tower. You can
think up a concept, play around with it, work it up into a treatment, and never stop to
consider the practical limits on development. If only it were that simple!
The authors of this book have found that it’s useful to have a portfolio of ideas to
draw on. We have about a couple dozen game concepts that we’ve taken to the first
treatment stage. It’s handy to have a file filled with ready-made ideas sometimes (and
not least when you’re looking for a job). The odds are that most of the ideas in your
portfolio will never find their niche. But that doesn’t mean they represent wasted effort.
It’s always good practice to try out different ideas. Even when a concept does generate
interest, other factors will soon change it. The full design specifications will be very
different from those first treatments. The factors that will enforce those changes are
covered in the following sections.
A game is released that uses many of the things that you thought made your game
unique. If you forge ahead, it’s going to look like you’re copying the other game. Find
a way to change direction: add humor, or change the game to make it more simple or
complex. Anything to move off the other guy’s patch and plant your own flag.
Your R&D group can’t deliver that 3D engine or the character Artificial Intelligence
(AI) on which your game depends. Plan for this in advance. Your design should have
alternatives. If it’s impossible to develop a game without technology that exists from
the start, that game should never be given the green light.
02 3634_CH01
10:21 AM
Page 21
First Concept
Gameplay emerges from rules. The most experienced designers have a feel for how
their game will play, but this isn’t an exact science. As we will discuss later in this
book, development must begin with a prototype that proceeds through successive
iterations towards the finished game. Each stage of the iteration will reveal features
that ought to work but don’t, and other features that nobody expected at all. Know
what you want to examine at each iteration, but don’t presuppose the results or you’re
heading for disaster.
If your concept has made it this far and you’re not ready to abandon it yet, then well
done. It could be that you’re on to a winner. It might even be (Ecclesiastes notwithstanding) that you really have devised something new under the sun!
Chapter 2, “Core Design,” will tell you how to start turning your design into a fullyfledged game.
Getting it Down
Case Study 1.2 is an example of a treatment based on our game, Conquerors, developed
with the game and TV production company, Creative Attic, which we presented in a
series of meetings from May to July 2000 to the BBC (British Broadcasting
Corporation, a UK network funded by special tax on the ownership of television sets).
We leave it to you to decide whether it deserved to pass the evaluation and criticism
stages needed to get a green light—as the BBC producers clearly felt it did.
Note that this is no longer a simple selling document, like Case Study 1.1. We’re now
getting into the initial stages of the design process.
Figure 1.1 depicts two Conquerors characters (called “myrmidons”) in battle.
The design and 3D models are by Leo Hartas (
02 3634_CH01
10:21 AM
Page 22
Chapter 1
Figure 1.1 Mean genes make for murderous myrmidons in Creative Attic’s Conquerors.
Case Study 1.2 Initial Treatment for Conquerors
A sport featuring virtual stars
A TV show where content is created by the audience
A league that anyone can enter
Sit back and watch, or sit forward and join in—it’s your choice
Players breed a stable of nonhuman gladiators called “myrmidons.” Each myrmidon
is a named character with unique attributes based on his underlying “genetics.”
Each has his own strengths and weaknesses. It’s up to the player to teach him how
to fight.
Training prepares your myrmidons for battle. Myrmidons learn to employ the
maneuvers and tactics that you teach them. When you set them against another
player’s team, the myrmidons’ AI takes over and the battle is fought without intervention by the trainers.
02 3634_CH01
10:21 AM
Page 23
First Concept
Each week, the top contests in the league are rendered in high-quality graphics and
broadcast on television, complete with human commentators just like any other
televised sport.
Individual gladiators will become Internet/TV stars and can be hired to other players temporarily or sold outright via the official Conquerors League server. You can
play the Conquerors videogame on local networks with friends or enter the official
league online and pit your team against the best in the world. The broadcast content
is created as a shared participatory process, and you can enjoy the TV show as a
spectator whether or not you choose to enter gladiators in the league.
The great potential of an interactive entertainment product like Conquerors is that
it takes in the whole spectrum of involvement, from “lean forward” videogamers to
“lean back” television viewers.
Conquerors is a virtual sport that viewers can watch on TV and can actively participate in via the Internet. It is a virtual sport that gives everyone in the audience a
shot at being owner and manager of a world champion team.
The Conquerors software package will be bought at retail or downloaded for a fee.
The Conquerors software allows players to breed, mutate, and train teams of nonhuman myrmidons. Training teaches the myrmidons to survive in different environments and to use their natural weaponry (claws, horns, and so on) in battles to the
death. You can pit your gladiators against AI-run opponents or against other players’
teams over a local network or internet connection.
Players can also submit their trained teams to the Conquerors server where they will
be placed in competition against other teams from all over the world. On the
Conquerors site, players can post challenges, discuss training strategies, buy or
exchange useful genes, auction their champion gladiators, and so on.
Each week, the premier division battles will be rendered in high-quality graphics and
broadcast on conventional television. Ideally, viewers with an Xbox or Playstation2
will be able to move seamlessly between the TV show and their training stable.
02 3634_CH01
10:21 AM
Page 24
Chapter 1
Hence, Conquerors is a theater-sports show like WWF except that the athletes are
tamagochi-like virtual creatures. And at the same time it’s an open participatory
show like Robot Wars, except that anyone in the world can download the core product and enter their own personal team for competition.
In summary, Conquerors viewers/users will be able to do any or all of:
Watch the regular non-interactive show on TV (high-quality rendered 3D
Create their own Conquerors teams on PC or console
Train their teams offline or online
Play offline, on PC, or game consoles
Submit their teams for official league battles online
Trade gladiators and training tips online
Place wagers on matches
View any match on demand on the Internet
The Game
In essence, Conquerors is a TV show based on an Internet videogame in the same
way that Popstars was a TV show based on a manufactured band. That is, a product
is created (in this case a videogame) that generates a fanbase and feeds back into
the show itself.
At the same time, Conquerors is a tactical videogame with an online dimension in
which players can put forward their own four-man gladiatorial teams to do battle in
various terrain “arenas.” Individual gladiators (myrmidons) have their own strengths
and their own ways of applying those strengths in combat as part of a team.
Players post challenges online. Being registered on the server’s league table costs a
subscription fee, and there might be an additional fee per bout. It’s also possible to
play local network bouts for free, of course, the downside being that the results of
such bouts don’t become part of the official league. Subscribers to the Conquerors
02 3634_CH01
10:21 AM
Page 25
First Concept
League can download any bout and watch how it went on a pay-to-view basis, using
their local software to reconstruct the bout from the match data. The top bouts will
be rendered with a higher quality graphics engine for broadcast on TV.
The gameplay itself has two main features:
1. Instead of directly setting their team’s initial stats, players mutate the myrmidons using a set of genetic algorithms. Every mutation yields a new generation
of myrmidons each with a unique set of attributes. You continue the mutation
process until the myrmidon has the attributes you’re looking for.
2. You don’t directly control your myrmidons during an on-line battle. Instead,
you carry out training bouts beforehand that involve body-hopping between
your myrmidons as you tell them what to do. As you do this, each myrmidon
is learning the tactics you want him to use. A stealthy scout-character might
learn to always shoot and run away when a tough warrior is coming for him,
for instance. This learned behavior is what the myrmidons will use when fighting a real contest on the server.
Creating a Myrmidon
In CRPGs you often start off by allocating characteristic points (either by choice or
randomly). Conquerors is a bit like that except that the characteristics are derived
from an underlying character gene-string which is mutated to give a wide range of
myrmidons—some with wings, claws, scales, gills, horns, even wheels.
Genetics specify the myrmidon’s potential, but it’s in “growing” the myrmidon that
his potential is realized. The growth process builds bone, muscle, and hide—all at a
cost in points called Chrome. Spending more Chrome makes the myrmidon bigger,
generally increasing his speed, strength, hit points, armour, and so on. But you only
have so much Chrome to spend in breeding your whole team, so making one of
them gigantic will mean another might have to be small and sneaky.
Myrmidon abilities are also limited by the underlying physics of the game environment. There are always trade-offs. If you want a fast myrmidon, you can’t have him
heavily armored; if you want him to fly, then he can’t be too big, and so on.
02 3634_CH01
10:21 AM
Page 26
Chapter 1
The procedure for generating your stable of myrmidons is described here:
1. The parent
A random “parent” is generated and shown on screen. This individual might
have long or short legs, wings or fins, claws or fingers, horns or huge ears—all
based on a string of numbers that are read as genes.
2. Mutation
A new generation of myrmidons is created by mutating the parent chromosome (asexual reproduction). Two types of mutation can occur. First, numerical values of genes may change by a small amount in either direction. That’s
the most common form of mutation. Secondly, the Gene Reader (which
decides the locations on the chromosome that relate to each body part) may
slip for one of the body parts—so that the creature’s left arm is no longer read
from the same sequence as its right arm, for instance. The probability of this
second type of “embryological” mutation is small and is likely to be pre-set to
certain sequences that will correspond to different “species” types.
3. Selection
A number of mutants from the original parent are displayed on screen. The
number doesn’t matter, except that it needs to be sufficient to give you reasonable choice over the direction the mutation is tending—say half a dozen. You
now select one of these to be the parent for the next generation.
4. A Myrmidon is born
Continue with successive generations until you have an individual that you
want to add to your stable. Before proceeding, save this individual’s genetic
code—you might later want to clone him.
Now, what you’ve got displayed on screen is the nascent myrmidon’s basic morphology—his shape, but not his size. The body template still has to be scaled up to yield
the adult myrmidon. Just as in real life, building bone and muscle costs energy. This
energy is provided in the form of Chrome, which you feed to the myrmidon. Bone
02 3634_CH01
10:21 AM
Page 27
First Concept
has a given cost in Chrome per unit, muscle another cost, and so the amount of
Chrome you allocate to the myrmidon will determine how big he grows.
Myrmidons can also have special abilities including webs, poison attacks, chameleon
skin, sonar, and so on. All of these also have a Chrome cost.
In filling your stable, you could produce just the basic four myrmidons needed to
make a team. Or you might prefer to have more to choose from, so as to be able to
hand pick each team for the terrain they’ll be fighting in. But you only have a limited
stock of Chrome to feed all your myrmidons, so greater versatility has its cost in that
each myrmidon can’t be as individually tough.
We want to encourage versatility rather than terrain specialization, as this makes it
easier for any two players to field teams against each other. Therefore you do get
more Chrome if you’re generating more than four myrmidons—but not in direct
proportion. So if you get 100 Chrome points to generate four myrmidons, you might
get 120 to generate five, 135 to generate six, and so on.
Although Chrome establishes an upper limit to how tough your team can be to start
with, two teams can never have exactly equal toughness. How would you define
absolute toughness anyway? Differing abilities can’t be so easily compared. To take
an extreme example, suppose you breed a group of armored myrmidons—so massively armored that they only move around at a crawl. Confronted by a team that
had even a single member with armor-piercing potential, your guys would be
doomed. The Chrome you’re given to spend at the start thus sets an upper limit on
any team’s initial toughness, but many other factors (specifically, the game physics
and the tactics you use) will play their parts in the final outcome.
Game theory is all about trade-offs—such-and-such a feature makes a myrmidon
stronger, but it also makes him slower. Consider two myrmidons with the same lower
body morphology and size. One is given little spindly arms, the other huge biceps
and massive claws. Obviously the second myrmidon has the greater offensive capability. But since arm strength doesn’t affect locomotive power, and that extra bulk only
has the same leg strength to move it around, he’ll also be the slower of the two.
02 3634_CH01
10:21 AM
Page 28
Chapter 1
The virtual physics system is there so that players can get an intuitive grasp on the
game mechanics. Flying requires a lot of energy, for example, so the ideal flying
myrmidon will be very fit and probably unarmored. Also wing lift goes up with L2
while mass goes up with L3, so the bigger you make the myrmidon to start with, the
less viable he’ll be as a flyer. A big hefty guy will have to win a lot of extra Chrome
(building up his muscle efficiency) before he can fly for more than short distances.
The game physics specifically relates to dynamics; that is factors such as the following:
Thrust (leg strength) minus resistance of the medium, divided by body mass.
Maximum speed
Depends on stride and leg flexibility (or equivalent for swimming and flying), traction, resistance of the medium, and the power delivered by the myrmidon’s metabolism (for example, heart/lung system).
Determined by the Myrmidon’s strength, mass, and his ability to get a grip on the
medium he’s moving through—hoofed or clawed feet are better than toes, for
All muscle mass consumes energy even at rest. In use, the muscle consumes extra
energy equal to force times the distance it is applied through. As energy stored in
the body is used up, fatigue sets in and restricts the muscle’s ability to exert maximum force. The myrmidon’s stamina (based on his heart/lung genetics) is a measure of how quickly lost energy is replenished.
A myrmidon’s main goal in life is to collide his armaments with the enemy’s vulnerable bits! Visual acuity and manual dexterity decide if he is successful or not.
Damage Potential
A question of weapon mass, speed, and sharpness matched against the target’s
armor and body toughness.
02 3634_CH01
10:21 AM
Page 29
First Concept
Special Abilities
Special abilities will be modeled generically in the design. For example, there will
be entangling attacks that all have the same gameplay effect of immobilizing the
target for a time. Graphically, various entangling attacks could be shown as a web,
a net, a viscous spray, a paralyzing volley of sparks, and so on, even though they
are functionally the same at root class, only with differences in duration, defenses,
and counters.
The game is set in a true 3D environment, which forms the main screen view.
There’ll also be a small top-down window you could open to get a view of the
whole arena. Different arenas would vary in size, but on average they’d be maybe 9
or 16 times bigger than the scene shown on the main screen. (Smaller arenas favor
brute force; larger arenas favor stealth, sniping, and ambush tactics.)
Myrmidons could rendered by morphing between the basic body types to reflect
the specific mix of attributes for that individual (long arms, horns, webbed feet, or
whatever). We would need a set of body-part anims for each leg length, arm
length, and so on, and these would be assembled at runtime to create a set of procedural animations unique to that myrmidon. (If it sounds a huge task, remember
that we can constrain the phenotype ranges for a force-to-fit solution.)
An alternative system would be to generate each myrmidon as a 3D model and animate these with a full IK (inverse kinematics) system. This has the advantage of
directly incorporating the physics within the game environment (as distinct from
precalculating it and applying the effects). The downside is whatever time we’d
need to allocate for developing an IK system.
How the Game Is Played
The game has two phases: practice bouts, which allow a player to pit a team of his
own against one run by the computer, and contests, where you put forward a team
to fight against another player’s myrmidons.
02 3634_CH01
10:21 AM
Page 30
Chapter 1
Practice Bouts
Practice bouts are single-player combats played out in realtime in randomly generated arenas. You control a team of four myrmidons drawn from your stable, pitted
against a computer-run team. The prevailing terrain might be jungle, desert, forest,
swamp or hills and, obviously, this affects how well your myrmidons can use their
Practice bouts are single-player games in their own right, but the main point of
them is to train your myrmidons in tactical use of their abilities. In a practice bout,
you will body-hop between the myrmidons.
The interface presents you with a range of options specific to that myrmidon and his
weaponry. You might have bred a charioteer with javelin attacks, a lance, and a
smokescreen ability, say. Maybe you tend to use him to ride in close discharging
javelins and then wheel off, using his smokescreen to avoid close combat. Only when
in rocky or marshy terrain, when he can’t get away quickly, do you favor using the
lance. Over time your charioteer’s AI will learn these preferences, so that when he
goes into the arena for a real bout he can apply the moves you’ve taught him.
Contests occur between teams put forward by two players. This can be by direct
connection, of course, but the official Conquerors leagues will be run by games centers accessed over the Internet.
On the day of a bout, you will need to adjust your team’s exercise regimen throughout the day, tamagochi-style. Myrmidons with high stamina will be enhanced if they
are told to spend the day working out. Others are better left to rest and gather their
strength. The individual myrmidons’ psychological factors will also have a bearing—
some are serene, others will have nervous energy that needs burning off, and so on.
It’s envisaged that expert players may change their team’s regimen several times in
the hours before the bout (probably via cell phone), and this can have a significant
effect on the team’s performance.
Any potential latency problems are avoided because you do not have continual
direct control of your myrmidons as you do in practice bouts. Instead, your
02 3634_CH01
10:21 AM
Page 31
First Concept
Myrmidons will reference the attack priorities and tactics you’ve taught them to
come up with a game plan of their own.
There is no compulsion for players to take part in league contests. Players who
aren’t involved in a bout can still pay to view it, and (just like in real-life sports) it’s
possible that most revenue will be generated this way. Many players may prefer to
practice at home and never participate in real online battles. This is just the same as
the guy who kicks a ball around in the park with his pals on Sunday afternoon and
then goes home to watch the football on TV. The fact that Conquerors can be played
in a variety of ways to suit the individual player or viewer is what will give it true
mass market appeal.
A Call to Arms
Players name their teams and post challenges online, possibly specifying their
choice of battleground or other terms. (“The Spine Suckers will take on anybody in
the Jungle Arena. one hundred Chrome says we’ll send you home in body bags.”)
Fighting it out for real requires both teams to be sent to the official server. The players would pay a fee for this (although maybe the first couple of challenges would be
free) and anyone who wanted to watch would also pay to download the bout.
You won’t actually get to find out a rival team’s Chrome value before fighting them,
but you could study their previous bouts to make sure you weren’t outclassed—and
to see the tactics they used. Also there would be no shortage of advice from other
Conquerors players on the Arena Newsboard.
Chrome won in a bout is distributed among all your team. Unlike the Chrome you
spend when generating a myrmidon, this doesn’t make the individuals grow any
more. Instead it increases the toughness of bone and hide, the efficiency of muscle,
and the efficacy of special abilities, to superhuman limits. (So a myrmidon who was
once too heavy to use those wings you started him with may eventually be able to
get airborne, and so on.)
02 3634_CH01
10:21 AM
Page 32
Chapter 1
Money Transactions
As well as Chrome, contestants could stake money on the outcome of a bout. Money
will be used to trade equipment, healing services, training from expert players, and
so on. You can use it to buy anything in game that you can convince another player
to sell. You might even be able to buy a champion myrmidon from another team.
Myrmidons will be identified by a code number on the server so they can’t be endlessly duplicated. If you sell a veteran myrmidon to someone else, then you can’t
continue to play him in contest on the league—though you could still use him during practice bouts or in LAN challenges. (However, you could of course sell a myrmidon’s genetic code any number of times, allowing the player who bought it to
start a new myrmidon with the same abilities as your veteran began with.)
Myrmidons as Virtual Stars
The key elements that will captivate an audience are the ways that we will personalize the myrmidons. This will not require complex AI; it is simply a question of having certain distinct “personality types” that will manifest in various ways. (For security reasons, this section was omitted from the version of the treatment requested by
the BBC.)
Why Teams?
The question may be raised why feature a team of myrmidons. Surely it would be
easier simply to have one-on-one battles?
The answer is that one-on-one battles would lead to a very trivial game. You can
liken optimum species in an ecosystem to attractor points in an n-dimensional
space, where the dimensions are different attributes. When you are competing one
individual against another, the solution for an optimum individual will be fairly
trivial. For example, legendary boxing trainer Cus D’Amato recognized that the
winning heavyweight in a match was almost always the stronger of the two. Anyone
who saw Mike Tyson in his heyday will know that Mr D’Amato’s theory was correct—and that it led to a number of very boring matches, as Iron Mike
usually dispatched his opponent within one minute or less.
02 3634_CH01
10:21 AM
Page 33
First Concept
At best, in a one-on-one Conquerors match, we would occasionally see intransitive
relationships such as those examined in Chapter 5, “Game Balance.” (Eg, Aegis
beats Feral beats Sppedy beats Aegis.) More likely there would just be one optimum
solution and the TV show would become a tedious process as players homed in on
the attractor point representing that specific set of winning attributes. Whereas, to
see an interesting variety of winning myrmidons, tactics have to be a major factor
in the game, and this necessitates teams. We have to hope (as development of
Conquerors will require collaboration between ourselves and a TV broadcaster) that
the broadcaster will understand this crucial point and wouldn’t end up wasting
development time on a fundamentally flawed design.
The biggest tasks to be faced in developing the game will be (1) the myrmidon AI
and (2) the artwork for different genetic patterns.
The AI will need to correlate multiple inputs (terrain, proximity of opponents,
current wounds, fatigue, type and range of weaponry of self and opponent, and so
on) to intuit a strategy (attack, defend, support, or evade). This is effectively a
pattern-recognition system that is defined by the semantic categories we specify as
the myrmidon’s senses. The way that the output strategy is then enacted would be
through tactical scripts, a bit like preplanned plays in football. This is not, in fact,
a difficult AI task; it is merely a question of limiting what the myrmidons can look
for (inputs) and what actions they might take (outputs).
The artwork could be handled one of two ways: either by developing a morphing
program that generates a custom-built 3D model for each chromosomal makeup or
by building a whole range of models (probably in the form of separate body parts)
that are fitted together as needed. The first is more computationally complex, the
latter more labor-intensive.
02 3634_CH01
10:21 AM
Page 34
Chapter 1
We envisage myrmidons as being quite alien—more like some kind of reptile/insect
hybrid than humanoid fighters. The main reason is exotic visual appeal, but there is
also the advantage that animation is easier to get looking right for a nonhuman
creature, and by not seeming at all human there will be lower expectation of the AI.
Players will be glad their trained critters can learn anything at all, rather than
grumbling that they don’t act as intelligently as human fighters.
It needs to be emphasized that the game is not the main product; it is the hook for
capturing larger audiences and revenue than the Internet would currently allow.
The idea is that viewers enter the “funnel” as casual spectators, get interested in
placing bets, then maybe start buying and selling myrmidons (which they can do
without owning the Conquerors software), and then the true aficionados buy the
product, subscribe, and become our next generation of content providers.
By comparison, a show like Robot Wars has high barriers to entry in terms of
time, cost, resources, and ability. But all anyone will need to create a team is the
Conquerors software—available in its simplest version free on the Internet. Looking
at other online games, Everquest has more than 400,000 regular players putting
in 20+ hours per week. And Everquest’s followers are just the narrow point of the
“funnel”—the active participants—without the publicity boost or audience size
possible with a TV show.
Conquerors is truly a new genre—a virtual sport that owes as much to football and
tamagochi as to traditional computer games. Players can choose the extent of their
participation, from solo practice bouts through friendly competition on LAN (the
“amateur league”), right up to full participation in one of the official internet
competitions (the “professional league”). The pleasure of spectating will be as
great as taking part, making this a game with true mass-market potential.
03 3634_CH02
10:26 AM
Page 35
Chapter 2
• Working up the concept
Core Design
• Adding gameplay
• Developing the game
• Prototyping
n this chapter, we’re going to show how to turn a raw idea
into a first-pass working document. By now, you have your
game concept—which means that you already know the
environment in which the player will be making choices,
whether that be a starship, a dream world, or the British
empire. You will even have a good idea what those choices
will involve—searching for dilithium, hiding from pursuers,
or sending armies off to war. Now it’s time to pin down the
details. What are the spells, weapons, units, or tactics that
will feature in the game?
Throughout this chapter, we will be using as an example,
one of Dave Morris’s own designs, Warrior Kings, a medieval
realtime strategy game that was developed at Eidos and Black
Cactus Games.
What Is a Game?
If we want to start putting in some gameplay, the question
“What is a game?” can serve as a good starting point.
But first, let’s consider what a game is not:
A bunch of cool features
A lot of fancy graphics
A series of challenging puzzles
An intriguing setting and story
03 3634_CH02
10:26 AM
Page 36
Chapter 2
Cool Features
Cool features are just fine; in fact, they are a necessity. However, cool features do
not, of themselves, make the game. You know those brainstorming sessions with your
developers that end with everybody saying, “Wouldn’t it be great if you could give a
whole bunch of swordsmen an order and every swordsman would do something a
tiny bit different?”
Even if all those great little ideas didn’t take forever to implement, there’s a point at
which cramming in extra features just starts to damage the elegance of the gameplay.
This tendency to add unnecessary features, commonly known as gold plating, is always
the result of somebody, at some point, deciding those features would be cool. They may
well be cool, and having them might help the game—just know when to draw the line.
Fancy Graphics
Games need fancy graphics, just as a blockbuster movie needs special effects—but
neither will save the product if the core creativity and quality is lacking. The fact is
that in today’s market, not having fancy graphics in your game is commercial suicide
effectively because games are a chart-driven industry. All such industries are strongly
affected by the reviewers—who tend, of course, to be hardcore aficionados with top-ofthe-range machines and who therefore expect to see impressive technology on show.
The danger with fancy graphics (as with cool features) is that they can distract the
development team’s attention from putting any depth into the game. The movie
industry is littered with examples of movies that cost $100 million and up but failed
to spend anything of that on getting a good script. Cinemagoers can see through that,
and there is growing evidence that gamers are starting to also.
We would certainly never turn down any fancy graphics that the technology can
provide, but the game has to be able to work without them.
03 3634_CH02
10:26 AM
Page 37
Core Design
The following quote from an industry master summarizes this quandary quite nicely.
“We need graphics. We need a good interface, we need visual clarity for our information
to come across, and we need graphics to do this. But when a designer is asked how
his game is really going to make a difference, I hope he has an answer that talks about
gameplay, fun and creativity—as opposed to an answer that simply focuses on how
good it looks.”
—Sid Meier, quoted in Edge magazine, September 1997
All games have puzzles. From determining where to build your castle extension to figuring out how to kill a wave of tricky aliens, games can universally be described as
containing sets of linked puzzles.
You may or may not like puzzles. Heuristic puzzles can be diverting, but often that’s
just the problem: They divert attention from an interesting story or game. On the other
hand, some people like them—although a “pure” puzzle game does not exist in the
wild, as they are usually merged with another genre to boost the interest level. (A good
example of this would be Tetris.) Either way, puzzles are not gameplay in themselves.
Puzzles are specific problems. Game design is about creating a system that will spawn
generic problems.
Setting and Story
Again, who could bring themselves to drown this kitten? A good setting encourages
immersion—what Coleridge called “the willing suspension of disbelief.” And a good
story draws the player in and impels him through the action. But, again, it will add
up to nothing if the gameplay isn’t there.
Another thing to avoid is using your game as a vehicle to tell a story—another
symptom of Frustrated Author Syndrome, and a very bad motive for wanting to be
a game designer. Case Study 2.1 provides an example of story versus gameplay.
03 3634_CH02
10:26 AM
Page 38
Chapter 2
Games Aren’t Everything
None of the factors we just looked at are sufficient—in and of themselves—for gameplay. All they can do is enhance a game or, in the case of puzzles especially, dovetail
into the kind of choices that define a game.
But here’s a confession: We don’t like the term “game” anyway. What’s so special about
games? Consider your idea in a new light. Does it have to be game? It’s like we said in
the previous chapter. Rule Number One isn’t “Make sure it’s a game.” It’s “Make sure
it’s fun!”
Many products are ruined by trying to force the elements of gameplay (or, worse,
pseudo-gameplay). Designers so often think they have to do this that it has become a
kind of dogma. They’re called games, aren’t they? So surely they have to have gameplay
in there, whether it enhances the product or not? Kids’ software is the most glaring
example of this. Look at the way children’s products are forever exhorting them with
pseudo-gameplay like “See how many blossoms you can collect while steering Roger
Raccoon’s boat across the river and avoiding the rapids.” The tasks are put in for lazy
reasons, and they get in the way of just playing.
Games are something that computers can do very well. It is not the only thing.
Interactivity is what computers are good at, and although games are interactive, so
are many other things—and that include many kinds of entertainment that nobody
has used the computer for yet. There’s nothing sacrosanct about gameplay, as shown
in Case Study 2.2.
03 3634_CH02
10:26 AM
Page 39
Core Design
Case Study 2.1 Story Versus Gameplay
Baldur’s Gate is an adventure game with quite a linear structure. Going the wrong
way too early will certainly get you killed. Therefore much of the playing experience involves finding this out the hard way and then starting again with a saved
game. Sometimes you meet people who have been sent to kill you or have other
nefarious goals. You are given various dialogue responses, allowing you either to
admit who you are or to lie. But it rarely makes any difference. Whatever you say,
you will usually get the encounter the designers intended.
Although there is a lot of story here, there’s little gameplay. The experience is
much more like reading a slightly interactive novel than playing a game. The
question is how would this story stand up as a novel? Or a film? Both of these
other mediums are ideal for storytelling. On the other hand, computers are good
for interactivity.
If we end up creating an inferior version of something that can be done better in
another form, we’re not making best use of the computer’s strengths.
Games Mean Gameplay
Assuming that your idea is not actually the stepping-off point for some entirely new
genre—that it will, in fact, work better as a game than as anything else—what will
now be uppermost in your mind is how to make it a good game.
A good game is one that you can win by doing the unexpected and making it work.
This—sometimes known as the surprise and delight factor—is almost a definition of
gameplay. We’ll be looking at others and examining them in detail in the next chapter.
However, it will do as our working definition for now. Phrased another way, gameplay
encourages the player to employ strategies. This does not mean that all games are
strategy games (in the sense of belonging to the strategy genre); it just means that all
well-designed games, from Tetris to Half-Life, require the development of strategies to
play them effectively.
03 3634_CH02
10:26 AM
Page 40
Chapter 2
Case Study 2.2 A Missed Opportunity?
Grim Fandango is an extremely impressive product of the quality we would expect
from LucasArts. The visuals are gorgeous, the music and vocal talent are faultless,
the setting is genuinely and breathtakingly original, and the dialogue and storyline
are better than most movies. You think we’re going to say the puzzles are bad?
Not a bit of it. The puzzles are fun and fit perfectly with the overall style. And
what’s more, the interface is unobtrusive and helps ensure that solving the puzzles
doesn’t break the flow of the story.
But here’s a question. Why did the designers insist on Grim Fandango being an
adventure game? If your main enjoyment comes from solving puzzles, you’ll
happily work your way though it and probably appreciate having a funny and dramatic story line to carry you through. But what about all the people who could
have enjoyed the story but not the game? For every hard-core gamer, there are
many more PC owners who would rather not have a story continually interrupted
by having to solve puzzles before you can find out what happens next.
As we said, Grim Fandango is of exceptionally high quality—easily good enough
that you could just sit through a rolling demo of the entire game, with every
puzzle solved for you, and enjoy it as perfectly well as an animated movie. The
point being that it could have been sold both as an interactive adventure and as a
sit-back-and-watch kind of story. And it could have had an option to let you think
about the puzzles, but with the addition of an “I give up!” key that would have
the hero solve the ones that stumped you. It could even have had a “Gimme a
clue!” key so the hero could help you out
With hardly any extra development effort, Grim Fandango could have been a
product that empowered the consumer with free choice in how to use it. It didn’t
have to be only a game.
This brings us back to the question at the start of this chapter. What is a game? If you
look at all existing games, you can see that the aim is to achieve one or more of these
03 3634_CH02
10:26 AM
Page 41
Core Design
Collect something (point-scoring games, and so on)
Gain territory (the classic wargame from Go onwards)
Get somewhere first (a race, either literally or figuratively—for example,
an arms race)
Remove a series of obstacles (by beating opponents, finding keys to open new
areas, and so on)
Discovery (exploration or problem-solving—very rarely ends in themselves)
Eliminate other players
Racing and conquest-type games both involve visible objectives. You can immediately
see at any point how well you’re doing. Collection games often don’t involve visible
objectives, which is why the designer often finds some way to give in-game rewards
for the points scored, as the following examples reveal:
In most role-playing games (RPGs), you earn experience points to spend on
improving your character’s skills, attributes, and spells.
In strategy games, you gather resources to spend on new units and upgrades.
In adventure games, you collect items to use in later puzzles.
Few games involve just one type of objective. Age of Empires, for example, has some
elements of a race: You want to be the first to snatch the scarcer resources, and you can
even play the game as a race to build a ”Wonder Of The World.” Age of Empires also
has two levels of collection: the obvious one involving the resources, and an abstract
victory-points system tacked on behind the scenes. And of course there’s the further
objective (the primary one in most scenarios) of wiping out the other players.
When you are designing multiple objectives into your game, make sure that they interact as shown in Case Study 2.3. Otherwise, you are really designing two or more games
in parallel and the parts will not combine to become a whole.
We’re saving a detailed analysis of the specifics of gameplay for the next chapter. For
now, just aim for clarity. You are about to start documenting your vision of the game.
Ask yourself the following questions:
03 3634_CH02
10:26 AM
Page 42
Chapter 2
What are the aims of the game?
How will players achieve those aims?
What will the game be like to play?
What are the key features that will create that gameplay?
Creating The Game Spec
In the last chapter, we talked about preparing treatments. Every treatment can be
different. The treatment is an outline concept, and at most a sketch of the game
design. Consequently, it can afford to focus on the game’s unique features and gloss
over—or even completely ignore—the details.
Not so for the game specification. Why are you writing it, after all? It isn’t to sell the
game; you have the treatment to do that. Instead, your purpose now is to describe how
the game will work and to communicate how you see that end result being accomplished. Because of this, you must address certain points in every game spec you write.
We’ll run through an inventory of these points, with the five most-important ones
Level design
While preparing the game spec, be sure to document your reasoning. (Documentation
might be rather a grand word for it at this stage, actually. We’re really talking about
those diagrams and calculations that end up scrawled on the backs of envelopes.)
This will form the basis of the designer’s notes, which we’ll discuss in Chapter 4,
“Detailed Design.”
03 3634_CH02
10:27 AM
Page 43
Core Design
Case Study 2.3 Integrating Game Objectives
A few years ago, one of us was asked to suggest ways to improve the gameplay
of a Japanese PlayStation title called Nessa no Hoshi. The original game involved
exploration of an alien desert environment, interspersed with Street Fighter-like
battle sequences.
The problem was that there were two quite distinct games there. In exploration
mode, the player was finding things like hints, maps, and keys that enabled
further exploration. In battle mode, the rewards were things that helped in later
battles: weapons and recovery item. It didn’t feel like it held together.
The answer was to integrate the two modes of play. Thus, the rewards earned in
exploration mode would often be useful in battle mode, and vice versa.
Features are what make your game different from any other game, and this is one reason why features are a good place to start. Another is that a feature-based description
of the game will endure throughout the development process, whereas it is very likely
that most of the rules you write in an attempt to create those features will have to
change further down the line.
This is because features are emergent from rules. Although the concept of emergence is
something else that we’ll cover in detail in the next chapter, emergent factors are those
that arise from the interaction between several rules. Emergence is not simply a case of
a feature being not immediately obvious—that, after all, would be merely a subjective
assessment. In precise terms, a feature is emergent if new categories must be used to
describe the feature that are not required to describe the underlying rules. Emergence,
we might say more simply, is that which makes the whole greater than the sum of the
parts. Case Study 2.4 provides an instance of emergence (as shown in Figure 2.1).
Types Of Features
Broadly, there are three categories of features. The first two are valuable, while the third
takes development time but adds little (if anything) to the game itself:
03 3634_CH02
10:27 AM
Page 44
Chapter 2
Some features are vital to make the game work properly, In Warrior Kings, troops
can be placed in formation, and the gameplay comes from knowing when to
group units in formation and when not to (and which formations to use if you
do). Without formations, you would have lost a whole dimension of choice.
These are integral features.
Some features enhance your enjoyment of the game but have no effect on the
way the game is played. These features convey the look and feel of the game and
help draw you into the story. An example would be a game like StarCraft that
customizes the interface to reflect the alien species you are playing. These types
of features are called chrome.
Some features are gameplay substitutes. They don’t enhance the game in any way,
merely giving you another exactly equivalent choice—which is no choice at all.
An example would be a computer role-playing game (CRPG) that lists different
costs for a slice of cheese, a turnip, and a loaf of bread, even though all have the
same food value. If a game designer keeps trying to include features like that,
it means he probably needs to get out more.
Figure 2.1 Emergence.
03 3634_CH02
10:27 AM
Page 45
Core Design
Case Study 2.4 An Instance of Emergence
In Populous: The Beginning, when enemy warriors run up to one of your preacher
characters, they will stop at a short distance as the preacher starts chanting. For a
while, they sit listening. If the process is not interrupted, they will then convert to
the preacher’s side.
That’s one rule. Another is that if enemy warriors have started fighting, a preacher
cannot convert them.
What this means in practice is that you can leave your preachers in a perimeter
defense around your settlement. The first wave of enemy warriors will be converted. This is automatic, and the player is free to get on with other things. But
now consider the second wave of enemy warriors. They come running up, and
the first thing they encounter is not the preacher’s conversion radius but the attack
radius of the last lot of warriors to be converted. A fight breaks out between the
groups of warriors, and the preachers cannot intervene to convert the second wave.
The feature that emerges is that you cannot leave your defenses unmanaged, From
time to time, you need to scroll around your settlement and move newly converted
warriors behind the preachers.
We have no idea whether this was what the designer intended or not. Very possibly
it was. However, it is a good example of emergence because it is not immediately
obvious that those two rules would give rise to this feature.
Your treatment described the game features. Your aim in writing the spec is to describe
how those features will create gameplay.
We already discussed gameplay briefly, and we will revisit it later in the book, so we
won’t spend too long on it here. Your description of anticipated gameplay in the game
spec serves three purposes:
03 3634_CH02
10:27 AM
Page 46
Chapter 2
It explains to the developers how the game is supposed to work.
It provides a vision statement that can be referred to throughout development.
It helps you focus on which features are integral and which are chrome.
Let’s look at an example from the Warrior Kings design. The aim was to include some
kind of supply rules. As a long-time wargamer, it has always irked Dave that supply
lines aren’t an issue in games like Age of Empires or Warcraft. You can draw on your
resource stocks, anywhere and instantly. This seems a bit of a cheat. It means that players get no reward for using strategies that are valid in the real world, like consolidating
their territory or using chevauchée, the tactic of ravaging your own crops and farms to
prevent an invading army from foraging any food. This certainly doesn’t help to foster
that willing suspension of disbelief. More significantly, if there aren’t any supply rules,
there won’t be any sieges, which is a bit of a drawback in a medieval wargame!
Dave thought about what he really wanted from his supply rules. If the design tried
to model everything from real-world warfare, the player would get bogged down in
logistical details that would be no fun at all. Who wants to have their army grind to a
halt or (worse) start dying of starvation because you didn’t have time to attend to your
supply lines?
An early intention was to delegate logistics to an AI assistant. Our philosophy is that
any no-brainer elements of the game—any choice that the player will always make,
that is—can and should be left to the AI. (The corollary, of course, is that these are
the things that it’s easiest to program the AI for.)
But supply lines aren’t that simple. Without going into a detailed analysis, it’s clear that
the routes you secure for supplying your army are not obvious. You may send an army
by land but arrange to supply it from the sea, for example.
Dave eventually reasoned that it wasn’t necessary to model every aspect of supply. The
game just needed something that the player could handle quickly. Obviously, in reality,
a medieval army commander couldn’t ignore supply lines, but, to create good gameplay, Dave needed a set of rules whereby it was the player’s choice whether to ignore
supply matters. If troops were continually running out of arrows, for example, that
would just be a headache for the player.
03 3634_CH02
10:27 AM
Page 47
Core Design
The answer was not to punish a player for failing to supply his troops, but to reward
him as long as he succeeds in supplying them. Therefore a supply wagon unit was
added to the design that could be supported by semiautonomous quartermasters.
Once the player had assigned supply wagons to his army, he didn’t have to keep issuing them with orders; they would follow the army around until out of supplies or
reassigned. And what reward would the player get for tending to his supply lines?
Simple: instead of dying when they run out of food, army units automatically recover
lost hit points as long as they are supplied with food.
So, not only does this mean that there are supply lines in Warrior Kings, it also means
there is a gameplay choice to be made. If a player thinks he can push through and
defeat his opponent in the first attack, he might not bother with supply wagons. But,
if he thinks victory will take longer, he will want his units to be able to heal up in the
field, and so he’ll need to secure his supply lines. The opponent now has the option
to counterattack directly or to harass the invader’s supply wagons, hoping to win by
attrition because his own supply lines are much shorter. And, indeed, he can use
chevauchée, burning out his own farms and then sitting it out behind his castle walls
in the classic medieval style.
Always remember what the interface is for. It isn’t just there to look pretty; its primary
function is to help the player play the game.
“I find the interface is one of the hardest aspects of game design. It should be intuitive
and icons should be kept to a minimum. I have never got an interface right (the) first
time. It’s one of those things that should be tested and tested until everyone is happy
with it.”
—Peter Molyneux, quoted in Develop magazine, May 1998
Good interface design involves answering the question, “How do I make sure the
player isn’t having to work against the system?” And really good interface design is
both an art and a craft, for it tries to address the question, “How do I make the player
forget there are any restrictions on his control of the game?” Case Study 2.5 is an
example of an elegant interface.
03 3634_CH02
10:27 AM
Page 48
Chapter 2
Case Study 2.5
An Elegant Interface
Dungeon Keeper (Bullfrog) is a complicated sim/stategy game with a lot happening
at once. The player manages a dungeon full of monsters, all of which are wondering around in realtime. Because each monster has his own agenda (sleep, eat,
study, train, collect his pay, and squabble with other monsters), the game could
very easily have become overwhelming.
What rescues it is a very powerful and user-friendly interface. The player designates tasks (tunnel here, build this type of room) and merely grabs and drops the
creatures he wants to do them. A judicious slap now and again serves to chivvy
the little devils on. Moreover, although there are many different monster types, a
sidebar allows you to see what all of them are up to at any time. You can use the
sidebar to select an individual monster, and by using the sidebar it’s also easy to
grab any number when you need them in a hurry.
Although it is reported that Peter Molyneux, the designer, did not like the user
interface (as he believed it was still too complex), it has to be said that it was a
very efficient tool for handling a large number of simultaneous complex tasks.
The feature-based description allows everyone to share a vision of the game that you
are aiming for. At this stage, though, you can only make a guess at the rules that will
actually create that game for you. This is why we stress time and again throughout
this book that game development must be iterative. You will be continually adding or
amending rules, very often only to find that the rule that you’ve added interacts in
some unexpected way with those already in place. So a little qualitative reasoning at
this stage will pay off later.
Trying to guess at emergent behavior is in principle very complex, but in practice
it’s often quite simple. For example, if characters in an online CRPG can become
stronger by killing other characters, then it doesn’t take a genius to see you will get
03 3634_CH02
10:27 AM
Page 49
Core Design
Now, suppose you’re not getting the feature you want. You’re designing a CRPG and
you were aiming for the kind of world described in Lord of the Rings, but instead you’ve
got people waiting to kill new players for the experience points. You need to look at
why this behavior is being fostered by your game rules. Adding extremely powerful
AI-operated police, for example, is not really the most sensible solution. A shortage of
police in the real world is not the primary factor in the murder-rate equation. It is a
factor, but a much more significant one is the fact that the vast majority of us are not
willing to kill our fellow citizens.
So, instead, you might reason that altruism exists in real life because society places us
in a situation that does not reduce to a zero-sum game. In other words, all members of
a society can benefit from cooperation. So you might look for rules that will create that
kind of payoff. The most obvious real-world equivalent is that new players could go
around in groups, which directly reflects the synergy of numbers. A single first-level
character is an hors d’oeurve for the player-killer, but 20 are a posse. However, players of
CRPGs don’t really want to join large groups. Because their paradigm is the solitary hero
(Conan, Indiana Jones, and James Bond), you need some rule that give individuals a benefit for committing to a group rather than being an outlaw. One way is to give playercharacters more personal power is by giving them more allies—not unreasonable in a
fantasy world where each individual has a personal link to his deity and temple, say. And
this is in fact the kind of solution being proposed by the designers of Asheron’s Call,
Microsoft’s online CRPG. Case Study 2.6 is an example of rules serving a game’s features.
Case Study 2.6 The Rules Must Serve the Features
In Warrior Kings, Dave wanted to make formations useful in a couple of ways. First,
they were a visible depiction of the AI behavior protocols, or standing orders, of
those troops. But that was merely a handy way to make sure the game helped the
player—an extension of the interface rather than a feature of the gameplay, in the
same way that archery units in games like Warcraft and Age of Empires will open fire
on the enemy without having to be told, because there is never a case when you
wouldn’t want them to.
But Dave also felt that units grouped in formation should be more robust than units
on their own. They would be slower and less maneuverable in formation,
so this was the main trade-off.
03 3634_CH02
10:27 AM
Page 50
Chapter 2
His first-pass design solution was to share damage through a formation. This meant
that, when one member of the formation was hit, part of that damage (a proportion
depending on the formation type) would be split between all other units in the
formation. This probably would have worked in a turn-based game, but Dave didn’t
even need to reach the testbed stage to see that it wasn’t ideal in realtime, where the
underlying game mechanics are a lot less visible. It would simply look odd.
So, his next idea was that the size and type of the formation would increase the
toughness of all its members. When they take damage, a certain amount can just be
ignored. This makes formations even more robust than Dave had originally intended: a 20-strong battalion of pikemen in orb formation is a match for 100 barbarians.
That means there needs to be compensating factors—slowing the formation movement rates, maybe, or increasing the time taken to form up. But it works better
because the player can straightaway see the size and type of his formation, so it’s
easy to get an intuitive grasp of how good that is, and having units shrug off
damage rather than share it around is credible in the context of the game.
Level Design
Level design affects the core gameplay; it is not just tagged on afterwards. Level design
contributes greatly to the style, background, and story line of the game. Still more importantly, the way the levels are constructed can either enhance the inherent gameplay or
detract from it. Although the lead designer is rarely the person who oversees the day-today design of the levels, he needs to address level-design issues in the game spec itself.
What level design should not be used for is to cover deficiencies in the gameplay. As
we’ll see in the next chapter, good gameplay consists of choices that are non-trivial.
Choices should never just be a question of recognizing that X is always better than Y,
and so therefore you should always do X. A level that says, “You can’t build bridges;
find another way,” begs the question, “Why are bridges in the game at all, then?”
Case Study 2.7 is an example of level design.
03 3634_CH02
10:27 AM
Page 51
Core Design
Case Study 2.7 Interesting Level Design
The game being developed is Arena, a first-person action game with fantasy roleplaying elements. Oliver, the Chief Level Designer, is discussing his ideas with the
game’s Lead Designer, Stelios.
“The fireball spell is turning out to be quite powerful,” says Oliver. “In playtesting
we’ve been using them a lot.”
Stelios frowns. “That’s worrying. Maybe we should put in lots of fire-resistant creatures. No, scratch that—it’d just be a force-to-fit. I should make fire spells take
longer to recharge, maybe, or have some other drawback like blowback, so that
there’s a game choice there….”
Oliver has seen that daydreaming look in Stelios’ eyes before. Anxious to get the
discussion back to specifics, he presses on: “Well, I have to get next week’s level
done. We thought we’d make it that the player just can’t use the fireball spell on
this level; it isn’t available. Like all those levels in Dungeon Keeper where you can’t
build some types of room.”
This has the desired effect of breaking Stelios’ reverie, at least. “That’s no good! The
fireball spell is supposed to be an integral part of the game. If the game is designed
properly, it can hardly be improved by cutting something out!”
“Okay, then,” Oliver says quickly, having learned to be diplomatic when discussing
gameplay issues, “would it be interesting if this level only had limited oxygen? The
player could still use the fireball spells, but he’d be using up the oxygen every time
he did.”
“But there’s already a lifebreath spell, which the player could use to survive without
oxygen even though he’s not underwater.” Stelios starts to get that thoughtful look
again. “Actually, that’s something we could leave the player to figure out for himself,
isn’t it? And if you make some of the foes on this level completely flameproof, but
they need oxygen to breathe too….”
“…Then the player could put on lifebreath and then use fireballs to suffocate
them,” agrees Oliver. “Fiendish! This could be an interesting level after all.”
03 3634_CH02
10:27 AM
Page 52
Chapter 2
Understanding the Game
The ways in which levels can complement the core game design should be described in
your game spec. Remember that the spec will be used by the level designers, who may not
yet have an accurate picture of how you expect the game to play. Describe how you envisage a typical level so that they don’t jump to the wrong conclusions. In particular, point
out the ways in which your game differs from a real-life equivalent. Otherwise, you may
find the level designers creating levels that ought to work, not levels that actually do.
For example, suppose you had just been given the original game spec for Age of
Empires and you were asked to design a level on paper. Never having seen the game,
you might start with two cities on either side of a mountain range. A narrow pass is
the only way between the two civilizations “That pass will be strategically important,”
you might think. But no; there are no lines of supply in Age of Empires. Conquest
therefore more closely resembles infection than invasion. That is, if I can get just one
villager through that pass into enemy territory, I can build a suite of army buildings on
his side of the mountains. I don’t need to guard the pass because my supplies can be
accessed by any of my villagers anywhere on the map.
This is not to say that Age of Empires is not an interesting game (in fact it’s one of our
favorites), merely that the kinds of scenarios that would be interesting in real-world
warfare won’t necessarily be the same in a world that plays by different rules. Make
sure your level designers understand those rules, and their likely repercussions, by
explicitly discussing them in the game spec.
Nonlinear Level Design
It’s our belief that the best level design is not linear and instead allows for multiple
approaches. Linear level design requires that the player must tackle problem A, then
problem B, and so on. The way to solve each problem might still be interesting, but the
linear structure precludes the possibility of strategic thinking, whereby the order in
which you tackle the problems is an interesting gameplay element in its own right.
A good game should allow for tactics (short-term decisions, like which gun to use) and
strategy (long-term decisions, like what path to take for victory). Following the thesis
that level design should provide an opportunity to showcase the game’s merits and not
close off valid choices, a nonlinear approach is the best way to achieve this.
03 3634_CH02
10:27 AM
Page 53
Core Design
However, designing interesting nonlinear levels isn’t easy. It requires a good underlying
game system, and it requires the designer to “let go.” He has to trust that the game
itself is more interesting than the puzzles he might devise using it. This is an act of
faith that is only justified if the design contains good gameplay, which is something
we shall be looking at in the next chapter.
Example Game Spec
To round off this chapter, we’re going to look at the bare bones of the Warrior Kings
game spec in Case Study 2.8.
Case Study 2.8 Game Spec
The following is a summarized template for creating a game spec.
Warrior Kings is a realtime strategy game set in the Middle Ages. It is
not overly historical: this is “fun medieval” not “real medieval.” Players build cities
and must manage an agricultural and geographically widespread economy while
waging war on each other.
Games are expected to take longer than Age of Empires, say, because of the need to
build up an army of multiple troop types and keep it supplied in the field in order
to win total victory.
The game will support both solo and multiplay. Up to six players can compete in freeform games as well as in scenarios with specific objectives, such
as maximizing piety (religious merit), being first to build a Star Chamber, and so
on. Any player may be human or computer.
If time allows, there may be other races with different mixes of units. But such differences will be quite minimal, with only one or two units being unique to each
race, like Warcraft 2 rather than Starcraft.
The first WK release will definitely feature a European medieval race. Other possible
races are Saracens, Byzantines, and Mongols.
03 3634_CH02
10:27 AM
Page 54
Chapter 2
The main screen will be a full 3D view of a medieval landscape dotted with trees, river, farm, etc. Levels will be much larger than the view on the
main screen. The camera can be tilted and rotated without restriction, as well as
scrolled anywhere on the landscape. Some kind of “fog of war” effect will prevent
a player seeing what’s gong on in any place where he doesn’t have characters.
Look and Feel
Art style is based principally on late-medieval sources such as Brueghel. Characters
will be polygon-based, not sprites: tiny industrious figures tilling the soil, hacking
back forests, erecting great cathedrals. Buildings will be based on those of the 14th
and 15th centuries.
Films to look at are The Hunchback Of Notre Dame, The Adventures Of Baron
Munchausen, and The Seventh Seal.
Musical sources: Carmina Burana, “Siegfried’s Funeral March,” Gregorian chant,
and Tudor lute music.
Books: Strange Landscape, A World Lit Only By Fire, The Paston Letters.
Commands are issued via mouse-selected icons. Selecting a unit and
then right-clicking on something will “intuit” the command. For example, selecting a military unit and right-clicking an enemy unit is interpreted as Attack. There
will also be keyboard shortcuts.
A side-of-screen interface layout is used as in Command & Conquer. This shows statistics for the selected unit(s), available command icons, and a circular minimap of
the entire level.
The area in view on the main screen is highlighted on the mini-map as a trapezoidal
window. Rotating the main view causes this window to rotate, but does not rotate
the mini-map itself. (This is to avoid disorientating the player.)
Most levels begin with several manor houses scattered across the landscape, each with one or more farms. Each player controls several manors. At the
start of the game, peasants are already working the farms; they can be reassigned
or left at those tasks.
Start Up
03 3634_CH02
10:27 AM
Page 55
Core Design
Each player also has at least one palace where his resources are located. Resources
must be brought to a palace before they get added to a player’s stocks.
The first game choices involve spawning more peasants or townsmen, assigning
them to different tasks, collecting resources, building new structures, and exploring
the landscape.
The object of the game varies according to the scenario. It will be some
combination of the following:
Destroy all enemy characters
Destroy or capture enemy buildings
Maximize your own resources
Steal or damage enemy resources
Achieve a custom objective (in special scenarios)
Entities in the game are structures, characters, world objects, and other
Characters are civilian or military.
Military units are pikeman, archer, knight, and so on.
Features are resources, supply lines, battalions, acts of God, and so on.
Many (not all) key military units can be grouped into battalions. Battalions can be
placed in formation, which defines behavior protocols and gives some formationspecific combat advantages, usually at a cost in speed or maneuverability, and so on.
The five formation types are column, orb, line, scattered, and wedge.
Units that can be grouped into formation are pikemen, archers, knights, lancers,
mercenaries, and sergeants.
03 3634_CH02
10:27 AM
Page 56
Chapter 2
Allowed formations for pikemen are column, orb, line, and scattered. Allowed formations for archers are…and so on.
Units in scattered formation retreat if taking proportionately more damage than
they are inflicting on their target. (This is intended to lead to skirmishing, so the
calculation must be done a little ahead of time, to give the unit time to turn and
move away. This will require tweaking.)
Placing units in formation assigns them a standard “behavior protocol,”
which the formation layout allows the player to see at a glance when viewing his
army. Units in scattered formation will retreat if seriously threatened (although
missile troops will occasionally stop, turn, and loose a volley at their pursuers).
Units in wedge will always attack with persistence if an enemy comes within
charge range, and so on. (See “Features” and “Rules.”)
It’s anticipated that the use of formations will allow players to set up reasonably
reliable battle plans, preventing the usual problem in realtime wargames where
your mix of units arrive at the wrong time because you didn’t click the mouse button fast enough. Warrior Kings is designed to appeal to the adult market (15 years
up), and hand-eye coordination should not be too decisive a factor. Strategic thinking is what counts.
Military units in the field cannot recover hit points unless supplied with food. (See
“Features.”) This encourages players to consolidate to protect their lines of supply
before expanding. Conversely, the player who succeeds with a daring raid on enemy
supply lines may be able to win in spite of poor odds. The terrain on any given level
will determine how easy this may be, with mountain passes being of key strategic
Levels with plenty of level, open ground will favor the use of units in
formation. Levels where the terrain contains many hills, chasm, and stretches of
difficult ground will militate against massed formations, instead encouraging the
use of more mobile and/or versatile units in small groups.
Level Design
03 3634_CH02
10:27 AM
Page 57
Core Design
Most levels will not be exclusively one type of terrain. However, levels must be
predominantly level or gently rolling terrain with infrequent cliffs if the feature of
troop formations is to be worthwhile. (In the extremely bumpy terrain of Populous
3, formations would not be worth including because they would never be used.)
Cavalry tend to be faster than infantry but are more seriously affected by difficult
terrain such as marshes, heavy snow, broken heathland, and so on. A level with
swathes of such terrain will force the player to think about how to use his cavalry.
Only a PC Version is planned. Minimum spec will be
Pentium 200 with a graphics card…and so on.
Technical Requirements
The client’s Marketing department has sales figures and projections for
this genre. It’s expected that the appeal of Warrior Kings will be akin to the thoughtful strategic gameplay of Alpha Centauri or Settlers rather than the furiously fast
action of StarCraft. The game’s unique selling points (USPs) are…and so on.
This summary of the game spec for Warrior Kings should give you a template to
follow as you design your latest blockbuster.
Obviously, there isn’t space to put everything here. The complete design would half
fill this book! But it’s useful to consider this as a checklist of the things you will need
to cover.
The Value Of Prototypes…
Along with the game spec, we recommend producing any kind of prototypes that you
can use as proof of concept. For Warrior Kings, we sketched out a simple board game
to test the campaign structure that would link a sequence of levels.
Board-game prototypes are fine for something like that. It was a design decision to
keep the campaign system simple, in fact, because it isn’t the main point of the game.
We didn’t want players having to deal with continually moving resources and armies
around their provinces, when the game they bought was supposed to be a realtime
wargame. So a board-game version was a good way to make sure the campaign system
didn’t grow out of control.
03 3634_CH02
10:27 AM
Page 58
Chapter 2
Board-game prototypes of the game itself are getting harder. Realtime, 3D, and artificial
life are all kind of hard to model with paper and pencil. Increasingly, however, there
are powerful commercial software development kits (SDKs) that you can use to
mock-up your game and even take it into full development. Among several very good
products, we will mention Virtools (, which we find useful for
putting together our own prototype designs.
…And the Necessity of Documents
We believe that prototypes are only adjuncts to the game spec document; they cannot
replace it.
Documentation explains what you are trying to achieve and how you might get there,
whereas a software prototype merely demonstrates the end result. The code architecture may be very different from what the designer imagines when writing the game
spec (if, indeed, he thinks about it at all). And, while game rules and features are the
province of the designer, the implementation of those rules and features is best left to
your software planner.
04 3634_CH03
10:28 AM
Page 59
Chapter 3
• Gameplay means
interesting choices
• The use of strategies
• Gameplay is just a
subset of interactivity
ow we’re going to take a closer look at some of the things
we mentioned in the last chapter. In particular, we’ll be
examining gameplay—what it means, how to achieve it, and
how to enhance it.
Strictly speaking, game theory is a branch of economics
in which systems governed by rules are mathematically
analyzed to determine the payoffs of various end points.
The strategies required to reach specific endpoints are
collectively termed gameplay.
This is not quite as limited a definition as it sounds. Morality
can be considered as an economic question—the costs and
payoffs of an altruistic strategy, for example, being quite
different if you are within 500 feet of the top of Everest as
opposed to an English village. Getting to an important interview after you overslept could involve gameplay. In the
movie, A Beautiful Mind, Russell Crowe (playing John Nash,
a pioneer of game theory) applies gameplay to the choice of
which of three women he should try to chat up.
However, for all its many applications, gameplay is only one
element in the composition of modern games. In GTA Vice
City, it isn’t some careful mathematical assessment of payoffs
that makes you choose which kind of car to drive or which
radio station to listen to. You do it because it’s fun.
Hence there is nothing special about gameplay. A software
product doesn’t have to involve gameplay to be entertaining.
Famously, SimCity has been described by its creator as more
• Different kinds of
04 3634_CH03
10:28 AM
Page 60
Chapter 3
of a toy than a game. And the foremost aim of simulation war games is to recreate the
reality on which they’re modeled. If you have a pure Conquistador sim, it may not be a
good well-balanced game at all (it may not be possible for the Aztec player to win) but
it could still be fun to play.
Only precedent makes us expect gameplay in entertainment software. No one demands
the same thing of a movie or a book. Plenty of things are enjoyable that aren’t games.
Who’s going to say, “Hey, that Hamlet is a great drama but where’s the gameplay?”
So then, rule number one is still that your product should be fun. That’s where the
entertainment part comes in. To be a “good” software product implies it will also be
interactive. That’s what will make it deliver something that you could not deliver in
any other medium. But gameplay? That’s a matter of choice.
Later in this chapter we’ll discuss some other kinds of entertainment software that are
interactive (certainly) and fun (hopefully) but aren’t necessarily games at all. This
chapter focuses on gameplay because it can be objectively analyzed. That’s all. You can
learn how to design a good game system the same way you can learn to lay bricks or
mix cocktails. But be aware that there is a lot more than game design to creating great
entertainment software. The bigger part of it is an art form.
What Is Gameplay?
Imagine you are playing a role-playing game. In this game, you play a group of adventurers—say a knight, a priest, a dwarf, and a thief. During an encounter, you typically
want your fighters—that’s the dwarf and the knight—at the front of the group while
your thief snipes from one side with arrows. Your priest, who is vulnerable, stays at
the back to cast spells.
Now, suppose the priest has two kinds of spells, each of which cost him the same
number of magic points. One spell injures the enemy (we’ll call those “E-Bolts”), and
the other heals injuries to your own group (we’ll call those “Band-Aids”). Which
should he cast during a fight?
First off, suppose the E-Bolts do as much damage as a sword blow and that the BandAids heal the same amount. When facing opponents who are equally matched with
04 3634_CH03
10:28 AM
Page 61
your own fighters, E-Bolts and Band-Aids are equally useful. Against a single foe using
several weapons (an insect creature with a sword in each of six hands, for example)
you are obviously better off using E-Bolts. When facing a horde of opponents with
individually weak attacks, such as a pack of rats, you are probably going to use BandAids. The point is that in each case you can easily decide which spell is better.
However, suppose instead that Band-Aids still only affect a single target, but E-Bolts
are area-effect spells that damage all the enemies in a given radius. And suppose also
that E-Bolts don’t do quite as much damage as before, but the target’s armor no longer
makes a difference. Now which is the best to use?
There’s no easy answer. It depends on lots of things. That makes it an interesting
choice. And that’s what gameplay is all about.
Implementing Gameplay
Sid Meier said, “A game is a series of interesting choices.” To be worthwhile, gameplay
choices must be non-trivial. Each strategy that the player considers using must have an
upside and a downside. If there is only an upside, the AI should take care of it automatically. If there’s only a downside, no one will ever use that strategy so why bother
including it in the game?
Remember that we are only talking about game theory here. You might put in something that’s just intended to be fun. A ray gun that plays a little tune whenever you
shoot something, for example. But even so, if that ray gun doesn’t have some gameplay
value—if it doesn’t have some upside, even if that upside is only a wide variety of catchy
tunes—then pretty soon players will stop using it, which means all that development
effort was spent just so players would say, “That’s cool—now hand me the BFG.”
A decision has gameplay value when it has an upside and a downside and when the
overall payoff depends on other factors. Usually this will be something along the lines
of “What if the other player uses neutron torpedoes?” or “What if I’m too close to the
sun when the warp drive kicks in?” So you choose the option that seems best for you
in the circumstances. It might be that you can see a better tactic but you simply don’t
have the manual dexterity to carry it off. That’s still a game choice—albeit one that is
likely to have hardened strategy gamers grinding their teeth.
04 3634_CH03
10:28 AM
Page 62
Chapter 3
Also it’s important to note that the game must be a series of interesting choices. Each
decision affects the next. The value of using the nail gun as opposed to the mining
laser depends on whether your previous decision was to run to the center of the room
or skulk in the corner. That depends on whether you set the cargo bay doors to blow
or not, and so on. Hence a game allows strategy. Indeed, a well-designed game cannot
be won without strategy. And strategy manifests itself as a series of interesting choices.
We’ve come full circle.
All this requires the game to display complexity. This doesn’t mean, however, that the
rules themselves must be complex. In the last chapter we looked at the concept of
emergence—complexity arising from simple rules. A neat example of emergence that
we like is the termite’s nest. Termites build towers of dried mud with cooling vanes
inside that keep the nest at the precise temperature needed to incubate the eggs. But
no termite has a blueprint of this wonderful structure in its head; it only has a few
simple rules to follow about where to put each bit of mud.
Emergence results from rules (behaviors) interacting with other rules or with the
environment. Resist the temptation to design in too many rules. The best games follow
the less-is-more principal. In the rest of this chapter, we’ll discuss how very often the
simplest rule sets are the ones that lead to the most interesting gameplay.
The Dominant Strategy Problem
How often have you flicked through gaming magazines and seen articles promising
“10 killer tactics” or claiming to disclose the ultimate character, weapon, or maneuver
in your favorite game?
If those articles are on the level (as they too often are), what they are doing is taking
advantage of flaws in the game design. A well-designed game shouldn’t contain an
option that is never worth using. So it certainly shouldn’t have a best maneuver,
or character, or weapon—otherwise, what was the point of the other maneuvers,
characters and weapons?
To borrow some terminology from game theory, an option that is never worth using
under any circumstances is a dominated strategy. An option that is so good it’s never
worth doing anything else is a dominant strategy.
Dominance means some of the options that you designed into your game—and maybe
even all but one of the options—are useless. If they’re useless, it means that pretty quickly
04 3634_CH03
10:28 AM
Page 63
players will drop them and won’t use them again. Not only is that bad game design, it was
a waste of development time as well. The only people happy to see dominance in a game
are the magazine columnists—because it makes thinking of strategy tips a lot easier.
Summing up, then: A dominated option is worthless. You wasted your time putting it
in your game. A dominant option is worse. It means that all the other options are
Near Dominance
Looking for dominance warns you if options will rarely be used. It’s worth taking some
time to look for near-dominance as well. A near-dominated option is one that is only
useful in very narrow circumstances. Conversely, a near-dominant option is one that
players will use most of the time.
Let’s consider an example of a near-dominated option: You put a special weapon in
your game, a stun gun that is only worth using against one foe, the raptors. Since the
stun gun has no effect on other foes, you won’t see it used except on the raptor level.
This is obvious, but still it’s valuable information. It means you can ask yourself these
1. Did I want the stun gun to be a more common feature?
Should I think of other applications?
Is the one-off use a positive feature?
What should I do to make the most of that?
2. If the stun gun is going to be used only once during the game, how much
development time is it worth spending on it?
Should we spend a lot of effort on this feature?
Is it worth putting in some special effects just for the raptor encounter?
3. What’s the best way of building the stun gun into the interface, as compared with
weapons that are more commonly used?
A game with lasting appeal must remain interesting even when players know the
tricks. Near-dominance isn’t as disastrous as dominance. A careless player could overlook the near-dominated option, and opponents can exploit that. But near dominance
04 3634_CH03
10:28 AM
Page 64
Chapter 3
does show which options will get used most, and logically those are the ones on which
you’ll want to lavish the best graphics.
Avoiding Trivial Choices
Good gameplay is achieved when the player faces problems that have non-trivial
solutions. But how do you avoid designing in trivial choices?
For this example we’ll be talking about units in a wargame, but they could just as easily be different maneuvers, weapons, spells, or whatever. The principal is the same. Say
that that the three units are warriors, barbarians, and archers. Figure 3.1 shows one
possible way that the combat relationship between them might work.
Figure 3.1 Transitive combat relationship.
This is a transitive relationship. Warriors are best, then barbarians, then archers. Now
consider the intransitive version in Figure 3.2.
Figure 3.2 Intransitive combat relationship.
04 3634_CH03
10:28 AM
Page 65
Warriors still beat barbarians, and barbarians beat archers, but now archers can beat
warriors. It’s a little bit more interesting. In fact, it ought to remind you of something
familiar. That’s right, it’s Paper-Scissors-Stone. The example in this case comes from
Dave and Barry Murray’s 1984 game The Ancient Art of War. In that game, the warriors
were stronger than barbarians at close quarters, but they were slow. The barbarians,
on the other hand, were fast enough to close with the archers without taking much
damage from their arrows.
There’s more to be learned from Paper-Scissors-Stone, and we’ll analyze it rigorously in
Chapter 5, “Game Balance.” For now, though, it’s worth emphasizing that a non-transitive
relation like Paper Scissors-Stone is only the minimum condition to get interesting choices.
It ensures a dynamic equilibrium, which keeps the game from stagnating, but it’s really
just first base.
To illustrate this, suppose you’re putting a Paper-Scissors-Stone kind of set-up into
your game and you decide to hardwire it. That is, A beats B, B beats C, C beats A—in
each case according to some look-up table that applies in all circumstances. If you did
it that way, you would need to decide the values for the look-up table—how much
better is each unit than the next one round the chain?
Figure 3.3 shows one option, in which a single warrior can beat any number of
Figure 3.3 Absolute superiority.
04 3634_CH03
10:28 AM
Page 66
Chapter 3
And Figure 3.4 shows the opposite extreme. Now, the warrior is just a tiny bit better
than the barbarian, so 99 warriors are an even match for 100 barbarians:
Figure 3.4 Marginal superiority.
Which of these is better? Don’t waste time thinking about it; it’s a trick question!
Neither of them leads to interesting gameplay. In the first case (assuming all units have
the same cost) armies will have to consist of an equal mix of As, Bs, and Cs; all other
combinations will lose. In the second case, a player loses very little by fielding the
wrong unit and so the unit interaction becomes virtually redundant.
So you can’t hardwire in a single rule and hope for interesting gameplay. It just won’t
happen. What’s needed is a dynamic relationship between different strategies so that in
some cases A is much better than B, and in others there’s little to choose between
them. And sometimes maybe the worm can turn, and B can beat A.
Paper-Scissors-Stone is a first step in avoiding trivial gameplay. It means there isn’t a
single optimum strategy that wins every time. However, there should be an optimum
strategy for each situation; “situations” in a well-designed game being complex because
they include not only terrain, altitude, weather, time—the obvious things—but also
whatever your opponents are doing. Case Study 3.1 provides an example of gameplay.
04 3634_CH03
10:28 AM
Page 67
Case Study 3.1 Environment Plus Rules Equals Gameplay
At the Battle of Hastings in 1066 AD, King Harold of England had the dual problems of an exhausted army (having just force-marched the length of the country
after another battle) and an enemy with a better mix of troop types. Duke William
of Normandy had more and better archers, as well as cavalry in the form of his
Norman knights. Harold, by contrast, was obliged to rely on infantry. The mainstay
of his army, his housecarls with their massive axes, were formidable fighters but
lacked the mobility for an offensive battle.
Knowing this, Harold drew up his troops on Senlac Hill, near Hastings, and
awaited the Norman onslaught.
Under normal conditions, the Norman archers could have skirmished against the
unsupported infantry and won eventual victory. Their volleys would pick off
targets and, whenever the lumbering infantry in their heavy mail hauberks came
too close, the archers could always run away.
That was the classic theory of early medieval warfare. On this occasion, Harold
had found a way to refute the theory. Shooting uphill, the archers’ volleys proved
less effective. Harold’s men were able to stand their ground, their kite shields
interlocked to form a defensive wall around them. Even so, initially some of the
arrows found their mark. Now and then one of the English soldiers would fall.
The housecarls yearned to charge down the hill, but Harold knew that the enemy
archers would counter that by retreating, and then would have the advantage. He
ordered his men to stand and, as the day wore on, the Norman archers grew tired.
Their arrows carried less force. They fell harmlessly among the housecarls, like twigs.
Harold was wise in warfare, but Duke William too had some tricks. Ordinarily,
William would not have ordered his cavalry against the English shield wall. Horses
will not charge onto a solid line of infantry (as long as the infantrymen know this
and have the courage to stand firm). Harold’s army did not lack for courage,
William knew; they had the opposite problem. He sent his knights in wave after
wave against the English line. Each time, the knights would wheel around at the
last moment, or take a few lance-thrusts against the shields of the English footmen, and then ride off.
04 3634_CH03
10:28 AM
Page 68
Chapter 3
The English housecarls had suffered arrows raining down on them all day. They
had little patience for a long wait and liked much better to be in the thick of battle.
Now they had to watch Norman knights turn their backs and flee. Many broke
from the line and charged.
Off the hill, the conventional theory of combined arms prevailed. Archers could
inflict injury on infantry who lacked the mobility to close with them. If the
infantry charged in an effort to catch the archers, enemy knights could ride in
amongst them and wreak havoc.
If things had been different, could Harold have won? Who knows? The point is
that he understood that there are ways to change the balance between different
troop types. A good commander isn’t the one with the best army; he is the one who
knows best how to use it.
Ensuring Interesting Choices
Elite was one of the seminal games of the early ’80s. The aim of the game was to accumulate wealth in the form of credits by trading between planets. When a player had
earned 1000 credits, he could trade in his pulse laser and get a beam laser and 400
credits in exchange. The 400 credits kept the player in a good position to trade, and
the beam laser was far superior to the pulse laser. Therefore, the player would regard
the decision to upgrade as a no-brainer. The curious part of the choice was that you
had to have 1000 credits before you could spend 600 on a beam laser. Had the beam
laser been available as soon as you had 600 credits, an interesting choice would have
been created: whether to upgrade right now (but have no credits left to trade with),
or carry on trading armed with a substandard weapon.
In the context of a great game, it was a small enough flaw, but a highly illustrative one.
A difficult choice had been taken out of the hands of the player.
Here are the kinds of choices that gameplay can involve:
An option that should sometimes be taken, and sometimes not, depending on
other factors
An option whose timing is critical and depends on the context
04 3634_CH03
10:28 AM
Page 69
An option that makes little difference whether you take it or not
An option that is always worth taking
An option that is never worth taking at any point in the game
The first two are evidently the most interesting in a gameplay sense. The third kind
(choices that don’t make much difference) are perfectly valid but they are chrome and
should be recognized as such. Options always worth taking should be handled by the
AI. For instance, in most third-person action games (Tomb Raider, Enter the Matrix, and
so on) the AI auto-targets the nearest enemy because, nine times out of ten, that’s the
one the player will need to shoot first. Lastly, options that are never worth taking will
be fun once only (if that) and thus ought to be seriously considered before you waste
developers’ time on them.
A Toolbox of Interesting Choices
Interesting choices are those that require good judgement on the part of the player,
which means that the correct choice must vary with the circumstances. Your aim as
designer is to ensure that those circumstances will never stagnate to the point that
there is only one right way to win.
There is no method for finding the best and most intriguing choices. That’s where the
creative process comes in. (We said at the outset that it’s a black art.) Even so, you can
follow a number of tips to ensure the gameplay choices aren’t trivial.
Strategic Choices
Strategic choices are those that affect the course of the game over the medium or long
term. (This is distinct from tactical choices, which apply right now.) Strategic choices
will have a knock-on effect on the player’s range of tactical choices later, so they are a
prime means of enabling good gameplay.
StarCraft is a typically well-designed game from Blizzard. One of the choices in the
game is whether you should first upgrade the range of your Marines’ rifles, or upgrade
the damage the rifles do, or research the StimPack which will allow the Marines to fire
faster. This decision requires some thought. If you are expecting a heavily armored foe
such as a Protoss Zealot, you will probably want the extra damage—whereas lots of
fast-moving opponents like Zerglings call for different tactics. Other factors that will
04 3634_CH03
10:28 AM
Page 70
Chapter 3
influence your choice are how many Marines you have, what the surrounding terrain is
like, whether you have bunkers built, and whether you anticipate an offensive or
defensive campaign.
Warzone 2100 includes a beautiful example of strategic decision making. The player
needs to build manufacturing plants to spawn his war machines. Optionally, he can
build these manufacturing plants in the current level—in which case they deliver units
quickly but only for use in that level—or he can build them back at his main base,
which means that new units arrive more slowly but the plant can be used to provide
units in later levels too.
Sometimes one option will be called for, sometimes another. A player who selects
options according to a set plan will not do as well as one who adapts to circumstances.
The lesson is that, to create good gameplay, different choices should lead to different
kinds of payoff. This way, you reduce the risk that choices will be trivial, and increase
the scope for good judgment.
Supporting Investments
Often a game has not only a primary objective but also secondary aims that have to be
attended to before you can reach the final goal. For example, the primary goal might
be to destroy the enemy, but to do that you might need to build farms to produce food
to spawn peasants to trade to make money to recruit soldiers and so on.
Some expenditure directly achieves your eventual goal. In a wargame, an example
might be buying a mercenary. Other types of expenditure indirectly work towards your
goal, and these are called supporting investments.
Primary supporting investments operate at one remove. For instance, improving
weapons technology to make your mercenaries tougher, or building an extra barracks
to house mercenaries.
Other supporting investments work more indirectly. For instance, building a smithy so
that weapons can be upgraded, or researching construction techniques so that barracks
can be built more quickly.
04 3634_CH03
10:28 AM
Page 71
Note that researching construction techniques helps at both the second and third
remove: It helps you build barracks to attract mercenaries, and it helps you build
smithies to research technology to improve mercenaries. This kind of complexity is
excellent, because it makes the decision to research less simple and more interesting.
By including decisions that operate at different levels, you create the need for players
to think strategically. The payoff for choosing different investments will alter according
to what other players do.
A useful rule-of-thumb for anticipating gameplay is to ask what is the best and worst
thing about each of the player’s options. For instance:
This maneuver does the most damage, but it’s the slowest.
This maneuver is the fastest, but it leaves me defenseless.
This maneuver gives the best defense, but it does little damage.
And then there’s a unique kind of choice:
This maneuver is never the best or the worst, but it’s the most versatile.
So, a useful question to ask yourself when designing a weapon or strategy for your game
is “When, if ever, is this the best option for the player?” Most choices that you put into
the game should be the best in some way. And one of these can be the choice that works
only moderately well, but in many different ways: the jack-of-all-trades option.
The more unpredictable the game environment, the bigger the payoff for having versatility of choice. Beginners in particular will benefit from versatile options in a game, as
it means there’s something they can do while working their way up the learning curve.
But versatile options are handy for expert players too. When fighting an expert opponent, you must expect the unexpected, and choosing the versatile maneuver or unit
may buy time to put together a more considered response.
One obvious kind of versatility is speed. The fast moving character or unit can quickly
go where it’s needed. So, normally, you won’t want the fastest units to also be the best
in other ways.
04 3634_CH03
10:28 AM
Page 72
Chapter 3
Also, the value of a fast-moving unit depends on the game environment. On the battlefields of the 14th century, a knight was deemed to be worth 100 foot soldiers. That
wasn’t because knights were each individually as tough as 100 men, but rather because
in a terrain of hedgerows, ditches, ploughed fields, and heathland, the knight had more
chance of being at the right place at the right time.
There are many other ways to make an option versatile. If a beam weapon can be used
to mine asteroids as well as to destroy incoming nuclear missiles, then that versatility
can make up for a disadvantage elsewhere. Of course, if there is no compensating disadvantage, there’s no interesting choice. Be careful not to make the versatile choice
dominant over all others. Also, be aware that the versatility of a choice may not be
obvious even to you as designer. In the last chapter, we saw how the designer of the
fantasy game Arena hadn’t originally anticipated the way players might use the fireball
spells. Case Study 3.2 provides an example of unexpected versatility.
Case Study 3.2 Unexpected Versatility
In 1997, Dave Morris was working with Jamie Thomson on Abraxas, an on-line
CRPG for a major UK publisher. One of the things we wanted was for the game
world to have its own laws for the players to discover and exploit. As Jamie put it,
“We might suddenly find that a player in Ohio has invented the Abraxas equivalent
of steam engine.”
One of the factors we were considering was combustion. The temperature scale
ranged from -30 (equivalent to absolute zero) up to +1000, with 0 being comfortable
room temperature. Materials had an ignition temperature, a burning temperature
(within the flame) and a radiative temperature (adjacent to the flame). So oil would
ignite at +6, the temperature of its flame was +4, and the flame modified the ambient
temperature by +1. (The range of that effect depended on the size of the flame.)
In one playtesting session, players were transporting barrels of phlogiston, a
magical substance that burns with a cold flame. They got caught in a blizzard
and the temperature dropped to -8, which was definitely life threatening. Now, the
properties of phlogiston were that it ignited at +5, burned at -5 and had a radiative
temperature of -1. How could they use it to save themselves? A phlogiston fire
04 3634_CH03
10:28 AM
Page 73
couldn’t warm them because it would merely reduce the ambient temperature by a
further -1, taking it down to -9.
Then somebody realized that the temperature within the flame was constant irrespective of the environment and because phlogiston burned at a negative temperature, it wouldn’t ignite them or their belongings. So they started a fire and survived
the night by sitting inside the flames at a cold but tolerable -5. They had found a
versatile solution that we as designers had never expected.
You can measure versatility by looking at the switching costs in the game. This is
how much it costs a player to change his mind about the strategy he’s using. An
example in an espionage game might be if you recruit a spy and later realize you
need an assassin instead. The switching cost is however much you wasted on the
wrong character, assuming for the sake of argument that the spy is not usable
elsewhere. So, say that both cost $1 million. When deciding which to buy, at first
you’d think, “If I buy the spy and I need the assassin, I’ll end up paying $2 million.
If I choose right, it costs me just the $1 million. On the other hand, suppose I buy
both now—I only need one, so I’ll definitely have wasted $1 million.”
Now suppose there is another character, the ninja, who can function as either spy
or assassin. How much should the ninja cost? It depends how unpredictable the
game is. In this example, if the game were completely predictable, the player would
know in advance which character to recruit and so versatility is of no value—the
ninja should cost $1 million just like the others. In a completely unpredictable
environment, the average cost would be $1.5 million ($1 million if I choose right,
$2 million if I choose wrong), which is what a good gambler would pay you for the
ninja. Since the truth will lie between those extremes, the versatile unit should cost
more than $1 million but less than $1.5 million.
Versatility is more prized in an uncertain environment. No multiplayer game is
completely predictable, since you can never know what the other player(s) will do.
Even in a relatively predictable game, some levels are more uncertain than others.
All of which makes the choice between specialization or versatility an interesting
one because it all depends on the circumstances.
04 3634_CH03
10:28 AM
Page 74
Chapter 3
Compensating Factors
Consider a strategy game that includes an aerial unit capable of crossing any terrain.
The aerial unit is the best at getting places because no other unit can cross all terrain
types—ships can’t go on land, mountaineers can’t cross oceans, camel riders are
slowed by forest, and so on. But this unit can go anywhere. How can we as designers
balance that?
We could make it slow.
We could make it weak (for example, easily destroyed).
We could give it low surveillance range (rather unlikely).
We could make it expensive to buy.
The last of those isn’t so good, though, because it doesn’t oblige the player to use the
unit in a interesting way. He either buys the unit or he doesn’t, and once he’s made that
choice he’s committed. The choice whether to buy might be strategically interesting,
but it won’t lead to clever tactics.
These are all compensating factors. They ensure that any game choice has something
in its favor and something weighing against it.
Compensating factors only work when it is clear to the player what they are. Maybe he
will decide that the powerful unit is worth having despite the fact that it’s so expensive
because he intends to make a big push and grab the oil fields. The player’s plan might
fail, but the choice is his and it is an informed choice. Case Study 3.3 provides an
example of how to balance compensating factors.
04 3634_CH03
10:28 AM
Page 75
Case Study 3.3 Balancing Compensating Factors
Shapeshifter is a proposed action-adventure game that allows the player to take the
form of various creatures. Each creature has unique abilities. To ensure a degree of
strategic choice, Martin, the designer, builds in a cost in magic points each time the
player changes shape.
Martin is describing one of the creatures at a development team meeting. “The
werewolf is by far the strongest character in the game when the moon is full, and
he’s invulnerable to anything but silver bullets, but he’s only as tough as a normal
man at other times.”
Sandra, one of the level designers, is skeptical. “If we have a full moon level, which
I assume we will, then any player who has the points to transform into a werewolf
will always do it. And on other levels you can see how long till full moon, so you
can save up your points. It’s a no-brainer. Is that good gameplay?”
“It’s not quite that simple,” says Martin. “First of all, didn’t we say we can have
weather effects? If the moon goes behind a cloud, or if another player casts a fog
spell, that weakens the moonlight and the werewolf isn’t so useful.”
Sandra isn’t entirely convinced. “Well, the weather effects are random, so that
makes changing to wolf form just a gamble rather than a gameplay decision. I can
see the point if other players are able to create fog, though.”
“What about the silver bullets?” points out Kay, the software planner. “Either
another player will have one—in which case the werewolf is dead—or not, in
which case the werewolf wins hands down.”
Martin considers this. “Obviously we don’t want it all decided from the start of the
level. So there will have to be a silver bullet somewhere on the full moon level, and
finding it has to be difficult.”
“It still comes down to another gamble,” points out Sandra. “Do I spend a lot of points
changing to a werewolf, or go looking for the silver bullet and hope I find it before the
werewolf player finds me? It’s just luck. I’m not sure these are interesting choices.”
04 3634_CH03
10:28 AM
Page 76
Chapter 3
“You know what?” said Kay. “It’s this simple. The silver bullet is going to be very
effective against the werewolf, which means it needs to be less useful in other
ways. It might be that it does less damage to other characters, or it might be more
expensive. But simply making it rare or hard to find is a bad way to balance it,
because that takes the choice whether to use it out of the players’ hands.”
From a gameplay perspective, compensating factors are worthless if the player only
finds out about them after making the decision that they apply to—for instance, if
the player has to choose a commando squad before seeing the mission they are to
undertake. Maybe the arctic survival expert’s many strengths are counterbalanced
by not having the disguise and fast-talk skills of the CIA man. But all that counts
for nothing if it turns out the mission they’ve been picked for is a hike through
tropical jungle to blow up a bridge.
All decisions that pay off lead to some kind of advantage. Some advantages are long
lasting: You got to the crate first, so you got the bonus. Others are impermanent: I built
my base near the tiberium, sure, but you can still grab it off me.
Impermanence gives the designer another way to confront the player with difficult
choices. Would I rather have pretty good armor for the rest of the game, or be completely invulnerable for thirty seconds? It depends, of course, on the circumstances.
Advantages (and disadvantages) can be impermanent in a number of ways:
They can be destroyed by chance or by enemy action.
They can be stolen or converted.
They can apply to something which you don’t always have (your heavy cavalry,
your jet bike, your wand of flame).
They can have a certain number of uses (six bullets in a magazine).
They can last for a certain time (the spell wears off at midnight).
04 3634_CH03
10:28 AM
Page 77
Impermanence is of course another kind of compensating factor. But it is one that
occurs so often in games, and has such special consequences, that it is worth considering as a separate category of choice.
Shadow Costs
In every game, players are continually being presented with costs and trade-offs. A cost
doesn’t have to mean money or victory points; it can be simply the things you had to
succeed at before you could get to the options you’re facing next. What is the real cost
of a game choice—in terms of time, effort, attention and alternative resources required
to get there? In Warrior Kings, players can hire mercenaries for gold. But the cost of a
mercenary isn’t only the gold you spent to hire him. First you had to build shops, and
townsmen to trade in them, and a tavern to attract the mercenaries to your town in the
first place. So you need to consider everything you had to go through to hire that
mercenary. This is the shadow cost.
Note that shadow costs are related to supporting investments. Tracing a flowchart
for all supporting investments will show you the kinds of strategic choices the player
has to make. Summing all the factors along one particular branch of that flowchart
tells you the shadow cost required to reach each node of it. Case Study 3.4 gives an
example of shadow costs.
Case Study 3.4 Shadow Costs in Age of Empires
The primary resources of Age of Empires are wood and food. Food is pretty much an
inexhaustible resource within the confines of the game, but wood—although easy
to harvest—is finite.
Consider a charioteer. The apparent costs of the charioteer are 60 wood, 40 food,
and 40 seconds to spawn. This unit’s shadow cost, however, changes greatly over
the course of the game. Early on, when the player’s economy is underdeveloped,
both food and wood are expensive. The forty seconds’ spawning time is not so
important, though. Later in the game, wood and food are plentiful and your main
concern is how quickly you can pump out new units. In the end game, if there is
no convenient supply of wood left, each charioteer you spawn is priceless.
04 3634_CH03
10:28 AM
Page 78
Chapter 3
The shadow cost of the charioteer involves other factors too, of course—like the
stables you had to build, and the upgrades you’ve paid to make him a useful unit.
But it is the influence of the fluctuating availability of resources that has most
dramatic impact throughout the game.
A shadow cost is the underlying cost to the player of a decision. You can use the
variability of shadow costs to add subtlety to your game. Vary the environment and
you vary the shadow costs. This is both a challenge for the level designer and an
opportunity to give the gameplay more depth that expert players will appreciate.
Synergies are the interaction between different elements of the player’s strategic
arsenal. They take four forms, as shown in Table 3.1.
Table 3.1 Synergies
Good fit = positive feedback
Economies of Scale: The more units of one type you have, the better each unit is. If your magicians draw
strength from each other, then the value of a new magician is greater if you already
have many others.
Economies of Scope: Of obvious use representing the advantage of combined arms (for example, infantry
units should be supported by tanks). Also covers the use of complementary gadgets
(such as mining lasers with mineral scoops, or a gladiator fighting with net and trident).
Poor fit = negative feedback
Diseconomies of Scale: The first unit is the most useful—after a while it’s not worth getting any more.
Some investments are designed to be increasingly costly—although this
sometimes disguises a drop in the real shadow cost.
Diseconomies of Scope: Mixed troops must move at the pace of the slowest, and must be given different
orders. Sometimes a focused approach is better than hedging your bets and
being able to do nothing well.
Ideally, all four of these synergies should be in play at once because when they interact,
they make decisions of timing more critical. When carefully chosen they also add
verisimilitude and make the game more immersive.
04 3634_CH03
10:28 AM
Page 79
Altering the balance between different synergies critically affects the gameplay. A game
with many economies of scale and scope is a game that gives positive feedback and
encourages the player who is already winning. Chess is a game of positive feedback;
it’s very carefully balanced but even a small mistake is magnified throughout the game.
Games like this tend to tip suddenly as a small advantage becomes crushing. Positive
feedback in a game with chance elements also encourages gambling—having 20 units
is much better than having ten, so if offered a double or quits gamble you’ll take it.
Monopoly displays this kind of gameplay.
On the other hand, if there’s an emphasis on diseconomies then you have negative
feedback. Here, advantages are increasingly hard won, and those who are losing will
rarely fall far behind. These synergies create a “catch-up” factor. Such a game will tend
to last a long time. It’s a good simulation of certain types of warfare, such as a land war
in Asia, or trench combat. How much fun it is will depend on what else you have put
in to keep the player’s interest.
A Final Word About Gameplay
As we said in the last chapter, by way of introduction to this whole topic, a good game is
one you win by doing something your opponent did not expect and making it work.
This is not simply a case of allowing for interesting choices. As the designer, you also
need to make sure those choices will interact. It’s not so satisfying to win a game just by
out-optimizing the other guy on resource production, or whatever. I want his choices to
have made a difference to mine. If the aim is just to complete the level with more points
than another player scored when he tried it, then your choices and your opponent’s
choices aren’t interacting at all. What you have then is a competition rather than a game.
Worst of all is a game where you simply have to know the right thing to do, and choices
don’t enter the picture at all. Then it isn’t even a competition, it’s a foregone conclusion.
We have to stress again that all this only applies to gameplay, and gameplay isn’t everything. Many software entertainment products aren’t games at all. There’s no harm in
that. Grim Fandango is a mystery story requiring you to solve puzzles; it is no more a
game than a book of riddles is a game, but nobody cares because it’s fun.
04 3634_CH03
10:28 AM
Page 80
Chapter 3
So much for gameplay; now for interactivity. Interactivity isn’t gameplay. It’s much
more important.
Most of the time when you’re playing a game, what you’re doing is just that—play.
You’re exploring, running, jumping, or thumb-twitching furiously to unleash a barrage
of bullets—which can be lots of fun, but a game that only has that going for it won’t be
fun for long. Gameplay enriches the experience because it allows you to interact and
take the experience somewhere. It’s the difference between idly kicking a football about
in the back yard and setting up a pitch with goals and having a match.
But gameplay is only one way of interacting. Take a game such as The Getaway. If it
consisted of only a detailed city for you to drive around in, that kind of game would
soon pall. The first thing you want to do is be able to interact with your environment.
Bump another vehicle off the road, smash the doors off a police car, run somebody
down. It’s violent fun—but it’s still fun. Kids play the same way, building bricks and
knocking them down. So you need to have enough opportunities to interact so that
the setting becomes more than a backdrop; it becomes an environment.
A game like that will be better still if you can, for instance, knock over a fire hydrant
and leave it spouting water. Then that pool of water makes the road slippery so that a
passing car skids onto the sidewalk. A pedestrian has to dive out of the way. He hauls
the driver out of the car and punches him. Now what if the driver identifies you, the
player, as the source of the trouble? He points, the pedestrian looks your way, and then
both characters start yelling at you. That way the interaction has created a story with
tension and action with emotions running high.
When you are playing Creatures or Black & White or The Sims, most of what you are
doing hardly involves gameplay at all. Those titles, like many other entertainment
software products, aren’t truly games at all. You don’t mind—you probably don’t even
notice—because you are having fun. But if these products weren’t interactive, you
would certainly notice that. Then you might start wondering if your time wouldn’t
be better spent reading a book, watching TV, or throwing a frisbee around in the park.
A non-interactive computer game wouldn’t win out against other fun activities.
04 3634_CH03
10:28 AM
Page 81
Interactivity is what computers do best. Thus interactivity, much more than gameplay,
is the heart and soul of entertainment software.
Kinds of Interactivity
You can interact with a game or narrative in many different ways. Even so, computer
game designers have pretty much restricted themselves to interaction with the facts.
Their concern is whether I open the first door and get the lady or the second door and
get the tiger.
Maybe this is because designers, like Mr. Gradgrind in Dickens’ Hard Times, believe
that “facts alone are wanted in life”. But in focusing only on the details of action and
plot, they are overlooking other possibilities.
Consider the ways that I, as a player, could potentially interact with a game:
Affecting the game world itself, either by changing the front end settings or (in
simulations and god games) as part of the game itself
Directly controlling the actions of a character or group of characters
Influencing a character’s actions at one remove (for example, by giving him clues
or weapons, like Zeus aiding a favored hero)
Influencing a character at two removes (for example, by leading him somewhere
or pointing something out for him to look at, like a Muse providing inspiration)
Deciding who to follow, rather than what happens—an invisible observer flitting
between narrative strands
Selecting what is interesting to me personally and making the game give more
time to those elements—as any child will, when told a story at bedtime
How many of those kinds of interactivity do you commonly see implemented? The
first, certainly, but only insofar as I can make the game easier or harder, or change the
speed. Why shouldn’t I also be able to say to a computer opponent in StarCraft, as I
can to a human player, “Hey, let’s build up a big army before we start fighting!” or,
“Don’t attack me because I just want to have fun building a city.”?
04 3634_CH03
10:28 AM
Page 82
Chapter 3
The second kind of interactivity in that list—direct control of one or more characters—accounts for pretty much everything else. You find almost no games with the
other kinds listed. And even the direct control option is used in a very restricted way.
I don’t know of any recent games, now that we’re in the era of realtime, that allow you
to switch control between two characters who are not on the same side. Why not?
What inviolable rule says you can only play one side in a game?
This is a failure of imagination. Computers give us the means to create new kinds of
games—and other toys and entertainment genres that aren’t games at all. And yet most
computer games remain, still, mutant versions of other media: films, or novels, or
boardgames. And, since those media have not made much use of interactivity, there has
been no prior example to guide the game designers of today. Case Study 3.5 discusses
one of many possible alternative models for interactivity.
Case Study 3.5 A Different Kind of Interactivity
A couple of years back, Dave was discussing various design proposals with the
managing director of a large development house. One of the ideas Dave was most
keen to pitch was Dalek City, a tie-in to a BBC science fiction series of the 1960s.
“It’s set on Skaro, the Dalek planet,” he began. “The Daleks have been mutated by
nuclear war and can only move about in mechanized travel machines. At the start
of the game they pick up power by induction, so they can’t leave the city.”
“Yep, got it.” said the managing director. “So you have to get upgrades to let them
travel outside the city.”
“Well, it’s certainly possible for the Daleks to get those upgrades. But you don’t play
the Daleks in this game.”
“What, then?” He was getting impatient. Concepts that take more than ten seconds
to explain are hostile territory to managing directors.
“There are all kinds of threats to the Daleks. Various mutants live in the jungle
around the city. Natural disasters like meteor showers can occur. There’s another
race, the Thals, who are their ancient enemies.”
04 3634_CH03
10:28 AM
Page 83
“You play the Thals, then.”
“Not really. That’s the point. You don’t play anybody. You can step in and help the
Thals if you want. Or you could spawn lots of mutant monsters to overrun the
Dalek city.”
The managing director evidently thought that here was something familiar. “That
must cost you resources?”
“No, your resources are unrestricted, up to whatever the game engine can handle.
You could just send in so many monsters, raiders, and natural disasters that the
Daleks would be wiped out right at the start. The point is, say you do that a few
times. Then you try something different: sending in just one monster to begin with.
The Daleks kill it and take it to their labs. They start researching it. Pretty soon
they don’t have any trouble dealing with that kind of monster—and what they’ve
learned will help them in other ways, too.”
“I see; it’s one of these Artificial Life things,” the managing director said. (It was
more of a growl, really.)
“Kind of. The Daleks are prime candidates for A-Life because their psychology is so
simple. They’re paranoid, inquisitive, power-hungry and they hate everything. And
their society is like a type of insect hive.”
The managing director wasn’t even growling now, just giving Dave a worried frown.
Dave decided to press on: “The aim of the game, you see, is whatever you want it
to be. You can just observe the Daleks going about their duties, like your own little
formicarium. Or you can trash their city and watch the little buggers get stomped.
Or you can test them with various threats and see how they learn and develop. It’s
the cruel-to-be-kind method. Eventually you might find you’ve nurtured them to
the point where they can take anything you throw at them. Played that way, the
ultimate aim of the game is to make the Daleks into an opponent you can’t beat.”
04 3634_CH03
10:28 AM
Page 84
Chapter 3
The managing director said nothing for a long while. Dave almost thought he might
be chewing it over. But then he shook his head. “Players don’t like games without a
clearly defined objective.”
Inwardly, Dave sighed. Drawing another sheaf of papers out of his file, he said, “Okay.
Well, in that case I’ve got this really good storyline for a first-person shooter….”
New kinds of interactivity promise a world of possibilities that we have hardly begun
to explore. To fully realize those possibilities, though, we have to be prepared to let go
some of the control that we have come to expect, both as designers and as players.
“Why?” Versus “What?”
Interactivity doesn’t have to be about what happens. It’s natural for programmers who
are also designers to emphasize the logical details. That’s the way they see the world.
But it is often more interesting to think about why something happens, or how it
caused other things to happen.
“If you are a process-intensive designer…then the characters in your universe can have
free will within the confines of your laws of physics. To accomplish this, however, you
must abandon the self-indulgence of direct control and instead rely on indirect control.
That is, instead of specifying the data of the plotline, you must specify the processes of
the dramatic conflict. Instead of defining who does what to whom, you must define
how people can do various things to each other.”
—Chris Crawford, writing in Interactive Entertainment Design, April 1995
Games guru, Chris Crawford, talks about process (the laws of the game world) as
opposed to data (the details of what happens in the game world). It is the blinkered,
data-intensive approach of most designers that has mired entertainment software in
the simplest forms of interactivity. To illustrate this, think back to some of the early
discussions of interactivity using digital TV. Lots of people proposed a system (which
they seemed to regard as a new kind of narrative) in which the viewer could decide
what the characters did. Essentially, they were talking about that unlamented genre
the interactive movie.
04 3634_CH03
10:28 AM
Page 85
Would that work? Imagine an interactive version of the television series ER. You get to
choose. Does Noah Wylie obey the hospital rules, does he give the experimental drugs
to the sick kid, or does he just go outside and play basketball? Is that a rewarding way
to experience interactive drama?
No. It assumes that the point of all drama is simply to find out what happens next.
If it’s good drama there ought to be a lot more to it than that. As we saw in Chapter 1,
“First Concept,” Aristotle listed five elements of drama: plot, style, setting, theme, and
characterization. It’s the last two that aren’t well served by the data-intensive designer.
Going back to the hypothetical example, is the point of the player’s choice to guess
what Noah Wylie’s character would really do? Then the interactivity reduces merely
to an empathy test. Is it to make him do things he wouldn’t normally? Then the interactivity is just a nerdy way to test the storyline to destruction. In neither case is
characterization explored in any detail, and the only theme we are apt to uncover is
that it’s kind of weird when people behave inconsistently.
Now consider a different way we could have played our interactive ER. This time we
don’t get to choose what the characters do or say. Now all we can do is choose which
character to follow around the hospital. Dr. Carter quarrels with Dr. Pratt over a
diagnosis—both storm off. We follow Carter, who gives his version of events to a
third character. Later we switch back to Pratt, who by now has spoken to various other
characters and is pursuing another agenda. It is only later, if ever, that we may stumble
upon the resolution or aftermath of the quarrel that we witnessed at the start.
This example in itself isn’t a sufficient condition for interesting interactive drama, but
it illustrates a couple of important points. The first is that all drama (whether interactive or not) derives momentum from the viewer’s relationship with the characters. A
rule often quoted in writing classes is that a character has to change and develop—
which, in turn, leads to the simplistic conclusion that all stories are about redemption.
In fact, characters don’t have to change to be interesting. But our understanding of the
character has to change. We need to see things that will build on our first impressions
and, ideally, that will surprise us and force us to reassess those impressions. That’s
because the viewer is part of the system—and if that’s true of non-interactive media,
you can bet it’s even more so in the case of games.
04 3634_CH03
10:28 AM
Page 86
Chapter 3
The other point is that interactivity can enhance enjoyment of all the elements of
drama, not only the hard facts of the plot. In the previous example, the protagonists
would behave true to character. Various themes might be revealed, whether the writers
intended them or not. Our motives for interacting would be emotional as well as
logical, and it would remain an immersive experience into the bargain.
The aim of all entertainment is to make the audience care. Interactivity is a new and
powerful technique in the writer’s repertoire for achieving that goal.
05 3634_CH04
10:32 AM
Page 87
Chapter 4
Detailed Design
• The importance of
the detailed design
• Interfacing with software
• Preparing for change
o this point, everything done on the game design has
been the “ivory tower” phase of the project, by which we
mean that the designer should have been the only individual
working on it full time.
At the end of this phase, you should have completed two
documents: the gameplay specification and the designer’s
notes. The gameplay spec is a “vision statement” that
describes the final product and serves as the point that
everybody will aim towards. When the code and the gameplay spec converge, the product will be ready to ship. The
designer’s notes are to be read in parallel with the gameplay
spec. These notes explain your reasoning and allow others
to challenge your assumptions, as well as your conclusions.
The Designer’s Role
“But hang on,” you’re no doubt saying. “Surely a designer
shouldn’t be a one-man band. Shouldn’t he or she listen to
other team members’ suggestions about the game? Isn’t it
better that the game design be a democratic process?”
In regard to the last point, we’re of the opinion that the only
worthwhile thing ever designed by a committee is the
American Constitution. But certainly the designer should
solicit and accept input from others. No one can think of
everything, and a brainstorming session might yield more
ideas in two hours than one person would have in a lifetime.
• Who needs what
05 3634_CH04
10:32 AM
Page 88
Chapter 4
Some developers still fear or distrust the role of the designer. Maybe this is inevitable—
the same way that builders distrust architects. (See Chapter 8, “The Future of Game
Design.”) However, the process of good creative design isn’t a dictatorship with a
designer writing rules that everyone must slavishly follow. Nor is it a case of committee
thinking. It’s a process of filtering and clarification in which the designer takes ideas
from many sources and distills them into the structure that will best serve the game.
So, by all means talk about the game to the developers who will be working on it. Your
marketing team has a role to play as well, so talk to them and find out whether the
game looks likely to fit the commercial requirements that are expected in two years’
time. Most importantly, get your client’s views based on the game treatment and
concept artwork.
What we mean when we say the designer should be working solo up to this point is that
the design should have been no one else’s full-time role. With this distinction in mind, it’s
useful to look at the broad timeline of an existing project, as in Case Study 4.1.
Case Study 4.1 A Development Timeline
Case Study 4.1 shows a development timeline for Corpus, a realtime strategy game.
Note that Corpus (which was released under a different name) was developed in the
late 1990s; modern games tend to require significantly larger teams.
Inspiration Duration: One month.
Process: The initial idea and feasibility discussion, leading to the treatment document.
People: Lead designer with occasional input from architecture and technology groups.
Outcome: The client (publisher) decides whether to greenlight the project.
Conceptualization Duration: Three months.
Process: Writing the detailed design.
People: Lead designer for months two through four.
Outcome: Two documents, the gameplay spec and the designer’s notes.
05 3634_CH04
10:32 AM
Page 89
Detailed Design
Blueprint Duration: Two months.
Process: Planning the tier mini-specs, a series of short documents detailing the stages
of the development. Each document represents a first “best guess” at the solution to
that tier, and a summary of what the tier is intended to achieve.
People: Lead designer and one software planner.
Outcome: Several (for example, six) mini-spec documents.
Technical Architecture Duration: Two months.
Process: Based on the tier mini-specs, the architecture group now prepares a technical design that details the tools and technology components that will be needed.
The lead designer explains the concept in the round, then hands over ownership
to the project leader. The Corpus project leader, a member of the architecture group,
will take the project through to conclusion.
People: Project leader, lead architect, and a software planner.
Outcome: Master technical spec and additional tier mini-specs dealing with technical
as opposed to gameplay issues. These and the game specs now constitute the project
plan. Tier mini-specs are now mapped to milestones (in this case, at quarterly intervals) that form part of the contract with the client.
Tool Building Duration: Four months.
Process: Constructing the components and tool set required for the game. The aim is
to achieve a degree of completion in any aspect of the game architecture that is well
understood and is not likely to need revising due to interdependencies.
Outcome: Tools and also game components that will be functionally complete if not
feature complete. In Corpus, these comprised a 3D graphics engine, level builder,
and a game set.
05 3634_CH04
10:32 AM
Page 90
Chapter 4
Assembly Duration: Twelve months.
Process: A project leader now takes charge of a team (for example, four programmers and four artists) to assemble the game according to a phase-based system on
a six-week turnaround. The project leader finesses the gameplay to conclusion.
There is also a tools programmer who supports the tools and can refer back to the
tools and technology group when necessary. (In theory, the lead designer is consulted at the test-and-review stage at the end of each tier, and doesn’t need to be
involved on a day-to-day basis. He or she is free to get on with the design of the
next game.)
People: Project leader for 12 months; a tools programmer for the first 9 months;
four programmers for 12 months; four artists for the last 9 months.
Outcome: The game and completed tools that are sufficient for a nonprogrammer
to design further levels.
Levels Duration: Four months.
Process: Game levels are built under the direction of the project leader. In practice,
the project leader will be succumbing to battle fatigue by now, so it’s vital that the
lead level designer is a capable and creative individual who can shoulder some of
the burden. In effect, the project leader should be able to share ownership with
the lead level designer during level building.
People: Project leader and three level designers.
Outcome: The finished game with all levels and in-game tutorials in place. The
manual text and artwork are also now finalized.
Review Duration: Three months (at least partially in parallel with Levels stage).
Process: Occurs mostly in parallel with the level building. This may be outsourced,
although, typically (as was the case with Corpus), the publisher has an in-house
Quality Assurance department.
05 3634_CH04
10:32 AM
Page 91
Detailed Design
People: Four testers.
Outcome: Bugs are found and fixed. If fundamental problems are identified in gameplay (although there shouldn’t be any at this late stage), the game can be returned
to the developers for tweaking, but this stage should really culminate in the gold
In developing a game, the total time from concept to gold disk matters less than the
costs-to-output ratio. Looking at the example of how Corpus was planned, you can see
how the costs were kept down. For the first 6 months, only a handful of people were
needed on the project. The bulk of the team got involved for only 12 months.
In an inefficient process, the entire team is put together from day one. The design is
worked on by a game designer, and when complete, is handed over to the development
team. The release preparations usually involve the lead and level designers making
last-minute tweaks to the gameplay assisted by some of the coding and art teams. With
this setup, there is wastage of time and money as employees are idle on the project for
some of the time.
The ideal situation plans to bring people in on the project as they are needed. The
design is worked on by a game designer and a software planner and when ready is
handed over to a development team who have been assembled for the specific project
under the supervision of the designer and the software planner. The release preparations involve the designer and the software planner tweaking the game—with the bulk
of the team having gone on to other things. The number of people working simultaneously on the project is optimized throughout.
However, the culture of the games industry has rarely allowed that kind of model to
work in the past. Historically, the industry evolved from small groups of enthusiasts
who pooled their talents. Nowadays, it’s a brave and well-funded development house
that can afford to have an entire team of maybe a couple dozen people assigned to a
project right from the start. They might get better games that way. The point, however,
is moot, as rarely would a game produced in that manner manage to keep its funding
long enough to see the light of day.
05 3634_CH04
10:32 AM
Page 92
Chapter 4
To contrast with the software factory (see Chapter 11, “The Software Factory”)
approach, we call the old style of development the Twelve Musketeers Model: “all for
one, one for all” (whatever the monthly burn). Is there any precedent for the switchover between these two models? Again, we might consider the film industry. In the
pioneer days, small groups took a camera out into the Hollywood hills (in those days
not a suburban sprawl) and threw a movie together a scene at a time. The script—
when there was one—might be provided by either a joint effort or by whichever technician or actor had the best idea of how to write.
This way of doing things didn’t even last as long as the aging gunslingers who were
often hired to act in those movies. The film industry had to get serious. The route to
making money meant making bigger pictures, which meant larger teams. Pretty soon
the old gung ho system was not sustainable. Efficiency required better planning, and
today we have a much more streamlined process. A writer slogs alone over his word
processor to produce the first drafts of the script. Preproduction work (casting, location scouting, and such) involves only a small team, and only then do the entire crew
of a hundred or more sign on. In postproduction, the project again contracts in scale,
with the director and editor, the foley artists, the music, and special effects.
It is an object lesson in critical-path analysis that the games industry would do well to
Design Documentation
To reiterate, you should now have the gameplay spec and the designer’s notes. The
following sections cover these in more detail.
The Gameplay Spec
The gameplay spec is a highly detailed description of the game that, if studied and
comprehensively understood, allows the reader to visualize the finished product in its
entirety. At the outset, you can consider the gameplay spec to be the very first prototype of the game itself, or you can think of it as a document sent back in time from the
future that describes the game you will have created one year from now. The gameplay
spec is an evolving document, as changes will be incorporated into it throughout
05 3634_CH04
10:32 AM
Page 93
Detailed Design
Now, take that gameplay spec you’ve slaved over. Put it in front of your programmers
and tell them it’s complete. Know what they’ll say? “How can it be complete? You
haven’t got psionics/barbarians/incendiary pigs in there yet.”
The confusion arises because people misunderstand what is meant by the word complete. In games development, it doesn’t mean that a document or code module will not
change. It means that it is in a functioning state: it works, even if not perfectly. If the
author of that document or code fell under a train, you wouldn’t have to throw away
his work. Since it is complete, others can use and modify it if needed.
So, complete does not mean finished. Completion is an ongoing process, and, at each
stage of completion, the game could (in theory) be shipped. It might not be in a state
to sell even a dozen copies, but it could be shipped. A game can only be described as
finished if it has all the features that you could possibly think of and everything works
perfectly, not only in terms of the code itself, but the gameplay, interface, and everything else.
However, in a commercial world that’s an unattainable goal. The French symbolist
poet, Paul Valery, said, “Work is never finished, only abandoned.” If he had been a
game designer, this is what he would have said: “No game is ever finished, only
abandoned.” Accept that. Your target is to get your game through as many stages of
completion as you can before shipping it.
The Designer’s Notes
The designer’s notes are really just a matter of documenting whatever was going on in
the back of your head while you were dreaming up that spec. Include even the doodles
on the backs of envelopes that you did at 2:00 A.M. if you like, just be sure to label
and explain them.
The designer’s notes are useful for two reasons. First, there’s the value of all documentation: if you walk off a cliff while daydreaming about your next game, the project won’t
grind to a halt. Second, there is also a benefit that is analogous to the comments in
programming code. One sentence in the gameplay spec (say, a single decision about the
relative firepower of two tank units, or the reward for collecting squash melons) might
be the result of many hours’ cogitation. You’ll remember the what long after you’ve
forgotten the why. Case Study 4.2 gives an example of the need to document the spec.
05 3634_CH04
10:32 AM
Page 94
Chapter 4
Case Study 4.2 The Need for Documenting the Spec
The game under development is Metropolis, a resource-management sim. John, the
designer, has written a rule that resources are localized in the world. Those quantities of iron and gold displayed at the top of the screen actually exist in the player’s
Civic Center. The Civic Center building itself has limited capacity of 1,000 units of
gold and 1,000 units of iron. Once the Civic Center is full, the player can create
more storage space by building storage silos, each of which holds 250 gold and 250
iron units.
Two months later, Ian, one of the programmers, comes to write the code that deals
with the resource handling. He questions why the storage structures would have
separate capacities for the two resources. “Why not just say the Civic Center can
hold 2,000 resource units of either type? Likewise the Storage Silo holds 500
John never prepared a designer’s notes document, and cannot think of a reason why
Ian’s suggestion should not be implemented. The change is approved.
Another two months later, in testing, Steve, a level designer, detects a problem. He
has reached full storage capacity and needs to build a new storage silo, which will
cost 100 units of iron. But his resource stocks at the Civic Center consist of 1,960
units of gold and 40 of iron, which is not enough iron to pay for the new silo. He
calls the others over to show them.
“We could change it so that workers carrying resources can directly use them for
tasks,” suggests Steve. “That way, as long as the workers you set to build the storage silo are actually carrying the extra 60 iron, they can still do it. It wouldn’t need
to go through the player’s resource stocks.”
“I don’t like that,” counters Ian. “It means changing the way resources are deducted
when you spend them. It’s too extreme a change to incorporate at this stage. Of
course, you could simply get your workers to first do some other task that requires
gold, thus clearing 60 units of space in the Civic Center, and then you’d have room
for the 100 iron.”
05 3634_CH04
10:32 AM
Page 95
Detailed Design
“That breaks Rule Number One,” protests John, quoting one of many Rules
Number One. “Your opponent should be the other players, not the game system
itself. It would mean stopping all workers from delivering gold until the iron
stocks were high enough to…” He slaps his forehead. “Doh!”
“What is it?” say Ian and Steve in unison.
“I already thought all this through, months ago! That was why the original spec called
for separate gold and iron storage capacities. But I’d forgotten the reason for it.”
Using The Design Documents
You must consider two facts:
The gameplay spec needs to be available to your programmers.
The programmers will not read the gameplay spec.
They really won’t. You can tell them it’s vitally important. You can plead with them.
You can even offer royalties. But the programmers will not read that spec.
You may have heard the saying that programmers can’t see the wood for the trees,
while designers can’t see the trees for the wood. It’s true, except that programmers
sometimes can’t even see the trees for the leaves on the trees. This is what makes them
good at programming, the ability to break things down to quantum levels of each tiny
individual step in a process. It’s also why they aren’t going to take a long document
like the gameplay spec and read and absorb it to the point where they’ve built a model
of the whole finished product in their heads.
Indeed, why should they? That’s the job of the design and architecture groups. The
programmers on the team are there to see to the fine detail.
But the programming team does need the design documentation readily available for
several reasons. First, carrying out tasks without knowing the goal is demoralizing and
counterproductive. Second, when some aspect of a task is open to question, it’s more
efficient to be able to refer to the spec than it is to hold a meeting to answer it. Third,
05 3634_CH04
10:32 AM
Page 96
Chapter 4
although it is financially expedient to have only a single person or small group write
the spec, it’s beneficial to involve everybody’s best ideas in evolving it to perfection.
Last, a shared vision of the project contributes to group cohesion and morale.
It’s a dilemma. They need it, but they won’t read it. Obviously, the design must be
made accessible to the programmers in some other way. You could permanently assign
a member of the design group to sit guru-like atop a mountain of 3D Studio MAX
boxes and answer questions whenever they arose. Because it’s unlikely that this would
prove to be cost effective, another solution that is almost as good is to have the design
documentation on an intranet site.
Figure 4.1 shows how the site might be organized. In this case, it is organized as a small
document labeled “Contents Schema” that maps down onto a larger document, “The
Gameplay Spec,” which in turn maps to another large document beside it, “Designer’s
Notes” as well as to numerous small single pages, labeled “Work in Progress”.
Work in
Work in
Work in
Figure 4.1 The Design intranet site.
The Contents Schema is merely a top-level breakdown of the design documentation
into numbered sections, which can be referred to when making changes. An appropriate
level of detail might be as follows:
7.3 Army units (object class)—Army units are characterized by shared behavior
protocols. Non-army units recognize army units and will respond accordingly.
7.3.1 Troops—Troops are human army units that also inherit from the Human
object class.
05 3634_CH04
10:32 AM
Page 97
Detailed Design
Types of troops are as follows:
➤ Foot Soldier
➤ Archer
➤ Scout
➤ Rider
Additionally, comments may be tagged to each entry to provide more detail. So,
attached to “Rider” in the example above might be the comment: “Beats an Archer in a
one-on-one fight starting at bowshot range but is beaten by a Foot Soldier.”
The Contents Schema leads a browser to relevant sections of the full gameplay spec,
which also maps across to the designer’s notes. Of course, hypertext links also cut
across between related sections of the spec, so the viewer is free to explore and learn
about the design via a concept path that he or she decides.
As Figure 4.1 shows, you can also map sections of the spec to work in progress. Thus,
after a while there would be a page showing the latest animations for a Rider character, a
brief description of the character’s function in the game, current values of armor, movement speed, and attack damage. Another page might show a mockup of the user interface.
In this form, the programming and art teams will be encouraged to explore the game
design and participate by suggesting changes and new features. Just be careful that they
don’t get more interested in perfecting the intranet site than in building the game itself!
Fitting Design To Development
So you have your gameplay spec and your designer’s notes nicely presented on an
intranet site. By this stage, we would expect there to have been at least one meeting
in which you presented the idea and sold it to the team. You should be planning on
having such meetings every month. You’ll have had conceptual artwork drawn up, and
you’ll have discussed music and sound effects, which you will have also linked in the
intranet site. Everyone who’ll be working on the game understands the look and feel to
aim for. So now it’s time for the team to get their heads down and start coding, right?
Not yet.
05 3634_CH04
10:32 AM
Page 98
Chapter 4
All you have so far are the broad brushstrokes of the concept. Returning to our analogy
before, you might liken it to the first draft of a movie’s screenplay, which isn’t much
like a shooting script. It lacks camera directions, storyboards, lighting instructions,
and set details. It isn’t a plan yet; it’s simply a goal.
Current thinking is that games are best built according to an iterative process.
Development is sufficiently hazy and risks are sufficiently high that some version of
spiral development is ideal, because it allows regular assessment of all the things you
didn’t anticipate. As well, it lets you closely control the development so that those neat
extra chrome features that could bust your budget are identifiable and can be dropped
if things get tight later on.
Spiral development is consistent with the software-factory model described in Chapter
11. However, it is not the only process that is consistent with that model (the waterfall
concept is another). We’ll explain why we favor it over the alternatives.
Tiers and Testbeds
Much of gameplay is emergent, in that you don’t know quite how all those rules will
work in practice until you try them out. This in itself is reason enough to rule out
models such as the pure waterfall, wherein each part of the process is finished and
coded in stone before the next phase begins. The waterfall model is so called because
each phase flows into the other in a strict order, much like a cascading waterfall.
More useful (and quite common in the games industry) is the evolutionary prototype
or testbed model, which maintains that the very first priority—within weeks of getting
the go-ahead—should be to have a functioning testbed version of the game. The early
graphics can be crude dots on a blank field, but you’ll get the feeling from talking to
exponents of this model that placeholder graphics are almost a point of pride, and the
cruder the better!
The idea of the testbed is that you can more or less design your game as you go along.
While undoubtedly, a creative designer can eventually produce an excellent game by this
model, it does have shortcomings. Development cannot easily be fitted to a schedule.
You know only that “one day” you’ll have a finished game. Also, the process cannot
accommodate critical-path analysis. There is no formal procedure for completing tasks
in parallel, and the whole team must wait while changes are implemented. It is only in
05 3634_CH04
10:32 AM
Page 99
Detailed Design
hindsight that you can see if the order in which tasks were completed was optimal. We’re
not much impressed by the argument that an experienced team gets a feel for which
tasks to tackle first, and it would impress us less still if we were funding the project.
The idea of developing the game through stages of a successive prototype is a solid
solution. These stages are what we have been calling the “tiers” of the build. In each
tier, new components of the game can be added, tweaked, and tested. Existing components from an earlier tier can be refined or expanded. The tier model also has the
virtue of allowing the team to focus on close horizons. Every six weeks a new iteration
is completed, and the game grows in front of your eyes.
A pure evolutionary model falls down, with its eagerness to rush into coding, in the
architecture linking the game components. You can lift out and refine the strategic
AI say, but what can you do if you realize, six months down the line that an early
assumption has saddled you with flawed architecture? Unless your client is very
understanding, it’s too late for a complete rebuild.
That’s why we recommend holding back on the coding just a little while longer.
Produce a testbed to assess your initial ideas by all means, but use off-the-shelf tools to
assemble the testbed and plan to junk it when it’s served its purpose.
To undertake an iterative development, you first need to take the gameplay spec and
turn it into a group of mini-specs. Each mini-spec deals with one tier or phase—a single
iteration of the build—so it is like the testbed approach except that your first testbed is
purely imaginary. The tier documents might say that you can start with just getting a
wireframe environment. Then you’ll have a first-person viewpoint to move around anywhere in the environment. Then you can include collisions with walls and objects, and
then a physics system to govern the dynamics of objects in the game world. Then you’ll
add weapons and opponents. (Here’s a good place to splice in the network code so you
can playtest head-to-head battles, although note that the skeleton of the network code
will probably have been in place from the first iteration.) Then you could sort out the
opponent AI, then the game editor, and finally the levels themselves.
Each tier mini-spec should outline the following items:
The goals—A designer’s wish list of features for that phase
The philosophy—What the phase is intended to test
05 3634_CH04
10:32 AM
Page 100
100 Chapter 4
The expected results—What the phase should have achieved after testing and
The alternatives—Other best guesses if ideas or concepts must change
Suppose you hand your sheaf of mini-specs to your software planner and lead architect. The first thing they’re going to say is that you’ve addressed only gameplay issues,
that there are components needed for the build that aren’t written yet. That’s good in
itself. The tools and components groups can get on with designing and writing those
now that they know how they’re to be used.
Additionally, other tiers will need to be added in-between the ones you’ve identified.
Those are the tiers needed to test technical rather than gameplay issues. This is good,
too, as it helps you explain to your programmers that those initial tier documents
merely show the ideal way that you as a designer would like to approach the build. As
the technical tiers are incorporated, your project moves obviously towards being a
team effort.
Furthermore, the tier documents (like the spec itself) will change during development.
The requirements of each tier will be modified. Some might be split into two. The
order might change. To begin with, it’s all just a first guess. And that’s fine because a
first guess is better than nothing.
If you think of building a game like building a bridge, the tier documents tell you
where the supports will go. It’s an advantage to map out the whole succession of tiers
before the team starts coding because it will then be possible to plan an architecture for
the game, as shown in Case Study 4.3.
Case Study 4.3 Planning the Mini-Specs to Fit
the Architecture
Warbots is a realtime strategy game. The designer, Peter, has outlined an early tier
whose aim is to get the bots moving around on the landscape and fighting each
other. “Basically, I want to be able to click on one of my bots, sic him on an enemy
bot, and he goes over to trash it,” explains Peter. “We’ll deal with the close-combat
rules, damage, and so forth in the next tier.”
05 3634_CH04
10:32 AM
Page 101
Detailed Design 101
“Okay,” says Nick, the lead programmer. “We have the route finding in place
already. We don’t have the fog of war yet—it’s a later tier—but I assume you want
that allowed for. So the bot will ask himself if he’s in attack range of his target. If
he’s not, he asks if he knows where the target is. If so, he accesses the route-finding
module and moves towards his target.”
“That’s fine,” agrees Peter. “Also, if he doesn’t know where his target is, I’d like
a question mark or something over his head—just as a mnemonic for now.
Obviously it won’t go in the finished game.”
One of the level designers, Charles, chips in: “Presumably that’s because bots who
can’t locate their target will have some other behavior we haven’t decided yet?”
“In principle. In fact, the weaker bots maybe will run in circles, or head to the
target’s last-seen position, while the smarter bots will have some kind of search
routine. Isn’t that so?” Peter smiles at the AI programmer, Victoria.
“If there’s time, they can be as smart as you like,” says Victoria. “But let’s rewind to
that bit about the bot asking if his target is visible. Can bots talk to each other?”
“You mean if a different bot on the same side can see the target? Sure. The player
can see everything that’s not in the fog of war. He’ll expect his bots to be able to as
“Okay,” Victoria says. “Let me suggest a couple of ways to do this. We could make
it that the hunting bot accesses his target’s coordinates and compares them with the
field of all coordinates on the landscape that lie within sight range of friendly units.
That’s one way. The alternative is that a player’s side has an AI assistant that is continually updating a list of visible enemy units. Then your bot checks if his target is
on that list.”
“I don’t see the difference,” says Peter.
“It makes no difference in this tier. But the visibility question crops up in several
tiers, and I’m saying we should address it now.”
05 3634_CH04
10:32 AM
Page 102
102 Chapter 4
Nick has been glancing at the mini-spec for the tier dealing with visibility and fogof-war issues. “You’re going to want the fog of war to be obvious in some way? So
the areas your units can’t see are veiled or grayed-out?” Peter nods. “And obviously
enemy units in visibility range will be shown on the player’s screen while other
enemy units won’t,” continues Nick. “So that means we’ll be creating a data field of
all coordinates in sight range and another data field, probably spawned from that,
of all enemy units that can be seen.”
“Apart from cloaked units,” Charles says.
With a wry smile, Nick jots a few notes on the fog-of-war mini-spec. “Cloaked
units remind me you’ll certainly want to put sensor units in also. So we’ll most
likely have two visibility fields—for sensor and non-sensor units.”
“I don’t see where this is leading,” protests Peter.
“It won’t affect your tier,” Victoria assures him. “But the way the design is planned
implies an architecture. I now know there’ll be a group of visibility data fields and
I know how they’ll work, so for now I can just put a routine in the AI that tells the
bot to access those fields. The default, in the absence of any kind of filtering being
written for those fields, will be that the bot knows where his target is at all times.
So the prototype for now will let bots hunt each other all over the map, but, when
we put in the visibility code, they’ll only be able to find opponents they ought to
know about.”
Why Use Documents at All?
One criticism is frequently leveled at this process. Namely, why the emphasis on
preparing and presenting a complete game design document at all? Couldn’t the game
just as easily be designed on a testbed, and presented to the teams in that form?
It seems strange to have to defend the written word, even now we’re into the 21st century. Certainly, written documentation is not sufficient on its own to communicate the
designer’s vision of the game. Hence, the emphasis on concept art, inspirational music,
05 3634_CH04
10:32 AM
Page 103
Detailed Design 103
references, diagrams, and anything else that you can use to get the concept across. But
a written version of the game design is vital, and here’s why.
In board games and face-to-face role-playing games, the game mechanics are transparent. You can see exactly why you get every result. Light infantry forced off the hill? You
shouldn’t have left them without support from your cavalry. Riots in the northern
provinces? Hand out bigger grain subsidies next autumn.
You don’t get that with computer games because the game mechanics are hidden
behind the screen or, more accurately, in the box standing under the desk or television.
This is fine from the player’s perspective (he can fight the Battle of Waterloo in a few
hours instead of three weekends), but it’s useless if you are trying to use a testbed to
explain the workings of the game to your team. And this is why you should use every
means possible to make the design document attractive, welcoming, and user-friendly.
And you must always make sure to keep it at the center of the development process.
05 3634_CH04
10:32 AM
Page 104
06 3634_CH05
10:35 AM
Page 105
Chapter 5
Game Balance
• Player balance and
game balance
• Component balance and
attribute balance
• Symmetry pros and cons
ave you ever admired the elegance of a fine automobile?
It could be a classic like the Rolls Royce Silver Ghost or,
indeed, any well-made machine. There is a beauty in the
perfect marriage of design and function.
So it is with games. To achieve this aesthetic purity, the
design must be balanced. A game without balance is untidy
and ugly, flaws that are reflected in the experience of playing
it. An unbalanced game is less satisfying. Worse, from the
developer’s viewpoint, it means there has been wasted effort.
If the parts of the game are not in balance, there is a risk that
some of them will be used rarely, if at all, in which case the
time spent developing them has gone to waste.
There are three broad types of game balance, all quite
distinct from each other:
Player/player balance is the art of making a multiplayer
game fair so that each player gets no other special
advantage but his skill. There can be luck in the game,
but it must apply evenly to all players.
Player/gameplay balance involves ensuring that the
player’s learning curve is matched by rewards that keep
him playing. The player should not find that his hardest
opponent is the gameplay itself.
• The Save Game problem
• Stone-Paper-Scissors
and variants
• Scalability of design
• Avoiding
06 3634_CH05
10:35 AM
Page 106
106 Chapter 5
Gameplay/gameplay balance means that features within the game must be balanced
against each other. If the Tabu Sword does twice the damage of other weapons, it
must cost more to buy or be otherwise balanced by compensating factors.
We will consider each kind of balance in turn, seeing how they relate to different game
Player/Player Balance
What do you think: Can Sarah Bryant beat Lion every time? And, if that’s so, does it
mean Virtua Fighter isn’t a balanced game?
You can see why the designers of Virtua Fighter didn’t give all of their characters the
exact same moves. That would be kind of dull, wouldn’t it? Half the fun is seeing how
different martial arts styles compete. In real life, it has been said that Gracie jujitsu
beats all other styles in a one-on-one bout. Maybe that’s so. Does it mean that an
authentic beat-’em-up that included a Gracie fighter would be unbalanced? Obviously
it does, if there is only one Gracie character in the game, and if he always beats the
other characters, and if the players are exactly equal in their skill and reflexes. Victory
then will invariably go to the first player to select a character.
Except that’s a lot of “ifs.” Suppose a friend told you he’d been playing a new beat-’emup and knew a character who won every time. Would you just accept it, shrug, and say
it wasn’t worth playing? We don’t think so! Rather, wouldn’t you lead the way to the
arcade and invite him to prove it?
So what if Sarah can whip Lion every time? That’s a challenge. It only becomes a game
imbalance if a complete novice playing Sarah can beat an expert playing Lion. Even
then, the game itself isn’t critically unbalanced as long as it’s possible for both players
to choose from a range of characters. Maybe Sarah isn’t so hot versus Jacky. And how
does she fare when she has to fight her own clone?
In most games, the designer’s aim is to arrange things so that victory is won by skill
and judgment. We don’t mean that there mustn’t be an element of luck. Most strategies
entail a gamble, and judging when a risk is worth taking is part of the fun. However, a
06 3634_CH05
10:35 AM
Page 107
Game Balance 107
balanced design should avoid random elements that favor one player, or whose effects
are so magnified as the game progresses that victory comes mainly as a result of luck.
The most glaring instance of such imbalance is when a player gets a lucky advantage
right from the start—perhaps by starting with his city close to a gold mine, or by entering the game right outside the door of the laser arsenal. How is this kind of undue
advantage best avoided?
Obviously, the simplest way to ensure perfect balance is by exact symmetry, which
means that players have exactly the same weapons, maneuvers, hit points, or whatever.
It also means that the layout of the level would have to be symmetrical so that neither
player benefited from a better position to begin with. As we saw in the Virtua Fighter
example, the problem with symmetry is that it may be the fairest solution, but it’s
rarely the most interesting.
There’s a classic Chinese martial arts movie based on the legends of the Water Margin.
At one point, two of the heroes square off for a duel. They stand watching each other
across the town square, poised in kung fu stance. Hours pass. Days pass. Then a breeze
stirs up the dust, and a speck goes in one hero’s eye. He blinks, frowns, and suddenly
relaxes with a sigh. He bows to the other hero. Such is the mastery of the martial arts
these heroes possess that they know the result of the duel without having to fight.
Their skill was equal. And so a tiny asymmetry—the loss of concentration due to a
speck of dust—was enough to decide the outcome.
Imagine that was a beat-’em-up. You lost because of the wind? Because of a speck of
dust in your eye? You’d take that game back to the store.
No game should ever be decided by factors outside a player’s control. This has a simple
logical consequence: The game has to be symmetrical at the level of significant game
factors. Asymmetries, which are often necessary for reasons of realism or aesthetics,
should be confined to the minor factors that cannot individually sway the outcome of
the game.
06 3634_CH05
10:35 AM
Page 108
108 Chapter 5
Symmetry In Level Design
Symmetric levels work fine in abstract games. The initial setup in chess is symmetrical,
for instance. But if, instead of chess, we were playing a more realistic game that
involved knights, castles, and bishops going to war in a medieval setting, we would
probably find a symmetric level unappealing. It wouldn’t look like a real map. We’d
therefore find it hard to suspend our disbelief.
There’s another, subtler problem, too. A symmetric level is an obvious way to equalize
the odds. In fact, it’s too obvious. As solutions go, it’s almost an insult to the player’s
Better is a level that is functionally symmetric but which is not obviously so at first
glance. So perhaps each player starts with some impassable geographic barrier protecting one flank of his territory, only in one case it’s a mountain range, in another an
inland sea, and in another a lava-filled canyon. The ordinary knights and foot soldiers
with which the players begin the game have no way of crossing such obstacles. Later in
the game, more advanced units might be available, meaning that those barriers are no
longer impassable. That isn’t a problem. If the opponent knows your defense relies on
the inland sea, she has the option of concentrating on building a navy to launch an
attack from the sea. Anticipating that, you might build a defensive navy, or you could
aim to recruit mountaineers to storm your opponent’s fortress first. As long as special
units like mountaineers and ships are balanced in cost, the game is still a fair one. And
there is also the possibility of bluff, which is so vital in any game of mixed strategies.
Seeing a few of your opponent’s ships out at sea may distract you from the conventional siege army that she’s sending in from the other direction.
What about an asymmetric level where players get to choose their start positions?
Insofar as both players have the same choice to begin with, it seems fair. But it’s still
possible that one player could choose a disastrous position and discover that he has no
hope of victory only after playing for some time. That’s not a well-balanced level, and
passing the decision on to the player doesn’t do anything to justify the imbalance. It’s
like the speck of dust in the kung fu duel. Balance requires that there is no starting
position that confers an overwhelming advantage. Giving the player the choice
whether to stand upwind or downwind is no good if you then just say, “Oh, you chose
downwind, that means you lose!”
06 3634_CH05
10:35 AM
Page 109
Game Balance 109
The most general solution is to avoid making the initial setup critically important. If
there are plenty of rocket launchers on the map, having just one of them near to your
start point won’t be such an overwhelming advantage. This begins to touch on the
issue of attribute balance, which we shall be looking at shortly.
Symmetry in Game Design
Symmetry in design means that the choices available to all players are functionally the
same. This is the easy way to make sure the game is fair. The hard way is to give each
player different choices but to try to balance those choices so that every player has the
same chance of success.
Warcraft 2 has both orcish and human players, and the units available to each species
are broadly the same. For example, the orcs have dragons while the humans have
gryphons, but functionally they are not much different, both being tough flying creatures with the ability to bombard units on the ground. Another example are the ogre
mages on the orc side who can use berserker rages and explosive runes, corresponding
to paladins on the human side who can heal injured comrades and exorcise the
If you look closely at the design of Warcraft 2, you see a basically symmetrical picture
with a few differences to lend flavor to the two races. Even these very slight differences
can prove significant. Consider a level consisting of many mountain passes. The orc
player’s explosive runes can be used to booby-trap the passes, giving a huge advantage
in that specific environment.
How much more difficult it is, then, to balance the design when each player has fundamentally different units. Yet this is what Blizzard attempted with StarCraft—and not
only attempted, but largely succeeded. The setting of the game ruled out strong symmetry. In the medieval fantasy context of Warcraft, it was reasonable to posit similar
styles of magic and warfare for the two races; but players would hardly have accepted
that different interstellar species would evolve even remotely similar unit types. The
amount of playtesting Blizzard undertook to get StarCraft balanced must have been
astronomical! It is hard to justify launching yourself on such an uncertain course of
development with no idea when (if at all) you will end up with a balanced game. We
would not recommend it to any developer without very deep pockets. The gamble paid
06 3634_CH05
10:35 AM
Page 110
110 Chapter 5
off for Blizzard, however. Whatever the cost in added development time, StarCraft is a
very finely tuned product—still arguably the benchmark for hardcore RTS games—and
its superior gameplay was rewarded by hitting the top strategy game slot for 1998.
The common sense approach suggests that a just-broken symmetry is far easier to balance than a complete asymmetry. In many cases, just-broken symmetries have more
aesthetic appeal, also. The inimitable Richard Feynman has something interesting to
say about this (as he does about most things):
“There is a gate in Japan, a gate in Neiko, which is sometimes called by the Japanese
the most beautiful gate in all Japan; it was built in a time when there was a great influence from Chinese art. This gate is very elaborate, with lots of gables and beautiful
carving and lots of dragon heads and princes carved into the pillars, and so on. But
when one looks closely he sees that in the elaborate and complex design along one of
the pillars, one of the small design elements is carved upside down; otherwise the thing
is completely symmetrical. If one asks why this is, the story is that it was carved upside
down so that the gods will not be jealous of the perfection of man.”
—The Feynman Lectures on Physics, Volume One
Also remember that it’s only games that ought to be fair. A pure simulation may have
the primary aim of authentically re-creating a historical situation, in which case the
question of balance between players may be only secondary. We mentioned in Chapter
3, “Gameplay,” the example of a Conquest of Mexico sim. Played as the Conquistadors
and their native allies on the one side versus only the Aztecs on the other, this would
be a very unfair game indeed. Played as all the Mesoamerican states on one side versus
only the Conquistadors would be unfair and unhistorical. There are ways to make a
balanced game out of it (by involving many players, for example, or by requiring either
of the two main sides to actively forge and maintain alliances with computer-player
nations) but, in any true simulation of those events, the Aztecs are doomed. You can
make a game fair, but reality rarely is.
06 3634_CH05
10:35 AM
Page 111
Game Balance 111
Player/Gameplay Balance
Recently a game was released—one that many games designers had been looking forward to for almost two years. The early reports made it sound so innovative and
intriguing, but when it finally hit the market, it was almost universally panned. Even
the reviews that praised it for originality gave it a low score. Typical comments went
along these lines: “Needless layers of complexity,” “makes you perform every last chore
yourself,” “tasks are out of all proportion to onscreen rewards,” and “tiresome and
What went wrong is one of the most common design pitfalls: The developers were so
busy dreaming up a nifty idea and festooning it with bells and whistles that they forgot
somebody would actually have to play the darn thing.
Player/gameplay balance means remembering that this business is all about interactivity. Sure, you want a game that’s fair to all players and where all the components and
features are worthwhile, but you also have to think about the player’s relationship with
the game. If you had to retune your TV every time you turned it on, you’d never bother to flop on the couch and channel surf. Likewise, if you have to struggle with a game
in order to extract the slightest reward, it’s not likely to hold your attention for very
long, as shown in Case Study 5.1.
Case Study 5.1 Is This Supposed to Be Fun?
Early CRPGs, for some arcane reason, used to begin with the player having to randomly generate scores in attributes such as strength, dexterity, and intelligence.
Well, those were the early days, and we didn’t know any better. Except that it still
goes on. Look at this quotation from a recent RPG:
“By using the plus and minus keys next to each trait on the menu, you can take points
away from some traits and add them to others to get the balance you want. If you
really don’t like the hand you’ve been dealt, you can click REROLL to get a different
set of values for the various traits.”
—The Baldur’s Gate Official Strategy Guide
06 3634_CH05
10:35 AM
Page 112
112 Chapter 5
“If you really don’t like the hand you’ve been dealt….” Or, to put it another way: “If
you have to be a pain and can’t just go along with what you got first time….” But,
each time you “reroll,” you get a new set of random attributes, which may be higher
overall than the first set. Assuming the attribute values make any difference at all to
the game, you’re going to want them as high as possible, aren’t you? If you’re
patient enough, maybe you should keep “rerolling” until every attribute comes up
with a value of 18. (The range is from 3 to 18, for another arcane reason we don’t
need to go into here.)
And why don’t you keep “rerolling” for hours, until you have the perfect character?
Because you bought this product to play a role-playing adventure, not to watch
strings of numbers changing. This has been set up merely as an endurance test.
When you finally get bored and frustrated enough, you’ll accept your attribute
scores and get on with the game.
As an example of player/gameplay balance, it’s not quite in the Stone Age.
Player/gameplay balance entails balancing challenges against the player’s improvement
curve. This is easy to do in a trivial way in computer role-playing games (CRPGs), for
example, where the player doesn’t have to get better at playing the game because the
character is getting tougher. So challenges can be tailored exactly to fit the character’s
experience level. But this makes for kind of a poor game. Character improvement in a
CRPG should be about widening your options, not just making you and your opponents tougher and tougher.
This aspect of game balance can be reduced to three simple rules:
Reward the player.
Let the machine do the work.
Make a game that you can play with, not against.
06 3634_CH05
10:35 AM
Page 113
Game Balance 113
Reward the Player
The big reward is to win the game, but when you first load up and start playing, that’s
a pretty distant goal. Before any big victories can be won, you have to learn how to
play. You’ll make mistakes, and mistakes are discouraging. So, every time you learn
something new and apply it properly, you need a little reward to offset the discouragement. This seems very obvious—it is very obvious—and yet it’s often overlooked.
Take the case of a beat-’em-up like Virtua Fighter. It takes a while to learn the trickier
combinations, and even longer to apply them properly. The results have to reward the
player’s effort and patience. The reward comes when you see Sarah flip over her opponent’s head and give him an unexpected elbow strike to the kidneys. Both the effect on
your opponent (the gameplay payoff) and the graphics associated with that combo (the
eye candy) have to be worthwhile.
Rewards should widen the gaming experience, not merely fit to points on a learning
curve. “I wanted to learn how to do the flying scissors kick, and I did, and it’s cool,”
isn’t half so good as, “Now that I can do the flying scissors kick, I can see a whole new
use for the reverse punch also.”
Also remember (as we discussed in Chapter 2, “Core Design”) that it’s always better to
reward the player for doing something right than to punish them for doing something
Let the Machine do the Work
This is mainly a question of the interface, which is the player’s channel of communication
with the game and, as such, has no business doing anything except showing the player
what is going on in the game world and helping him to make changes to that world.
Also, remember that the player is sitting with a computer in front of him and that a
computer is a tool specifically designed to deal with a wide range of tedious tasks. If
those tasks aren’t going to be fun, don’t make the player do them. If you look back
over the games of the last 15 years, you’ll find plenty of glaring examples of designers
who have gotten confused over the boundary between a chore and a game feature. For
years, CRPGs used to be boxed with little bits of graph paper so you could draw a map
as you explored each dungeon! It’s like we saw in Chapter 3: If a game option is a nobrainer, there’s no excuse for not having the AI take care of it. Don’t bother the player.
06 3634_CH05
10:35 AM
Page 114
114 Chapter 5
Make a Game You Play With, Not Against
We’ve lost count of the number of games that go to great lengths to set up a vivid
world, atmospheric locations, and a powerful, intriguing story line—and then go and
spoil it all by insisting the player can progress only by trial and error.
You know the kind of thing. You’re approaching the crossbowman guarding the exit.
You think maybe the transparency potion you bought will mean that he can’t see you,
but no. He draws, shoots, and it’s (sigh) back to the last save. So now you know the
potion doesn’t work. Maybe if you try running up to him? No, he’s too quick on the
draw. Back to the last save. Now you try dropping the bottle with the potion. The
guard hears it, comes looking—and shoots you. Now you’re stumped. Just at random,
you try drinking the potion and then dropping the empty bottle. He’s coming this way.
This isn’t going to work, surely? But it does. He walks past, not seeing you. It’s only
now you realize that the transparency potion works only when you’re stationary. He
goes past, still looking around for whoever dropped the bottle, and you sneak behind
his back to the exit.
Is that good design? No, it’s lazy design. The player should succeed by skill and judgment, not by goofing up enough times that the correct solution was the only one left.
Case Study 5.2 provides an example of the save game problem. The reason games
need a Save feature is so we can grab some time to live our real lives between games.
Turning it into part of the game just so the player has to put up with episodes like the
one in Case Study 5.2 seems mere perversity.
Case Study 5.2 The Save Game Problem
A designer was talking about his upcoming dungeon RPG. “I’ve got a great trap!” he
told us gleefully. “The player steps on this platform, it descends into a chamber he
thinks is full of treasure, then a ring of flamethrowers go off and he’s toasted.”
“What if I jump off the platform before it gets to the bottom?”
We meant it as a solution, but he saw it as a loophole. “Yeah, we’ll have to make it a
teleporter instead. You get flamed as soon as you materialize.”
06 3634_CH05
10:35 AM
Page 115
Game Balance 115
“But what is the solution?”
“There isn’t one!” He was astonished at us. “I’m saying it’s just a killer trap. It’ll be
“So, there’s no clue before you teleport that this might not be a good idea? Charred
remains on the teleporter pad or something?”
“Nah, of course not. That’s what the Save feature is for.”
Some people think Save Points are the answer, but that assumes that Saves themselves are
the problem. They’re not. The problem lies in games that are designed around the need to
save—games that are too arbitrary, or where the learning curve makes it hard for you to
progress by dint of skill. (Games that lack player/gameplay balance, in other words.)
Levels constructed around the need to save can destroy the value of a good game design.
A case in point is Myth: The Fallen Lords. Small squads could be deployed in different
formations, creating an interesting hybrid of wargame and RPG that might have been
rewarding to play. One would hope that thoughtful analysis would lead the player to a
good choice of formation, perhaps in some cases keeping a few characters in reserve to
permit flexibility. Instead, too often the levels were pitched to such a degree of difficulty
that you could survive only if you had chosen the single best formation. Flexibility in the
form of a reserve was a luxury you couldn’t afford: the foes would ferociously wipe out
your nonoptimal deployment before you had time to change your orders. The only
answer was to save frequently and, then, with the benefit of hindsight, engage the foe
next time with all your troops placed in the perfect position. Those levels were a shame,
too, because the basic game design of Myth looked so good.
What is the answer to the Save Game problem? It is better player/gameplay balance,
along with level design that has faith in the game design to deliver the goods.
“Experienced garners have come to regard the save-die-reload cycle as a normal component of the total gaming experience. I want to slap all those people in the face and
cry, ‘Wake up!’… Any game that requires reloading as a normal part of the player’s
progress through the system is fundamentally flawed. On the very first playing, even a
06 3634_CH05
10:35 AM
Page 116
116 Chapter 5
below-average player should be able to successfully traverse the game sequence. As
the player grows more skilled, he may become faster or experience other challenges,
but he should never have to start over after dying.”
—Chris Crawford, writing in Interactive Entertainment Design, Vol. 8, No. 3 (February 1995)
Gameplay/Gameplay Balance
The original Warcraft really only featured three essential combat units: the Knight, the
Archer, and the Catapult. Those units comprised most of your army. Okay, there were
also wizards who could summon demons, but you had to disallow them if you wanted
a balanced game. Without the demons, then, Warcraft was a very well-balanced game
that led players to constantly evolve new strategies.
Warcraft 2 had more than half a dozen essential units and several others (like Sappers)
that had very specialized uses but on some levels were invaluable. After looking at the
game for a few hours, we got talking to Steve Jackson, now of Lionhead Studios. It
turned out he was another Warcraft enthusiast. “There are only two ways they could
have got the units so well balanced,” Steve reckoned. “Either they struck lucky first
time, or they must have continually tweaked them for hundreds of hours of playtesting. Either way, it seems incredible.”
We’d be willing to bet it was the latter. Nobody gets that lucky. The game was almost
perfectly balanced. In practical terms, that means that the player never feels there is a
unit he could do without. Nor is there any unit that seems to cost too much, so that
the player would try to avoid having to use it.
One way the designers of Warcraft 2 made sure of that was to build in a strong Stone-PaperScissors (SPS) relationship between the key units. There’ll be more on this later in the
chapter, but for now we need only note that this automatically tends to create a game balance. This is because, with “strong” SPS (such as, where a single unit of one type beats any
number of the next type), you have to play using all the unit types. None are dispensable.
What challenges do we face when balancing aspects of gameplay? Broadly, we must
consider three things:
06 3634_CH05
10:35 AM
Page 117
Game Balance 117
We want there to be a variety of interesting choices rather than a single choice
that always dominates.
This isn’t easy to establish because the optimum choices depend on the choices
other players make.
It’s not easy to see how frequently different choices will be worth making, yet we
need to know that to balance the game.
Sounds like a Catch-22? Not quite, because there are some simple concepts we can use
to get a first guess at the value of different choices, even within a dynamic game.
These concepts were developed specifically to analyze this kind of problem, and since
WWII they have been used extensively to examine problems ranging from nuclear
deterrence to the pricing policy of console manufacturers.
The concepts we’re talking about form part of the analytic arsenal of game theory, and
we’ll be using some of them in the remainder of this chapter.
Component and Attribute Balance
Balancing the elements within a game takes place on two levels. First, there is component balance. This establishes the value of each game choice. For balance to exist, each
choice must not be reducible to a simple value relative to some other choice, or, if it
can be so reduced, other factors must even it out.
Let’s look at an example. In StarCraft, Mutalisks are flying units that can move over
any terrain and have powerful ranged attacks, but cannot fight other flying units.
Wraiths are fast aerial units that can turn invisible but aren’t nearly as tough as
Mutalisks, although (unlike Mutalisks) they can attack other flyers. Observers are light
aircraft that can see invisible units but are no use in a fight. There is no way you could
devise an expression that set the value of each of those units relative to the others
because they do such different things.
Alternatively, consider a pirate game in which the ships you can select from are
Brigantines, Galleons, and Dreadnoughts. The Dreadnoughts are the toughest, next
come the Galleons, and last, the Brigantines. All these units have identical functions in
the game. There is nothing a Brigantine can do that a Galleon or Dreadnought can’t do
better, If the Galleon is twice as tough as the Brigantine, game balance requires at the
06 3634_CH05
10:35 AM
Page 118
118 Chapter 5
first level of approximation that the Brigantine should cost half as much to spawn.
Actually, when balancing costs, you need to be quite careful that you are looking at all
the relevant factors. What we mean by “twice as good” is a unit that is twice as effective
as another in all functions of the units—in this example, twice as powerful in both
offense (damage dealt out per unit time) and defense (hit points).
But we said there were two levels to consider when balancing game choices. The attribute level involves not the relative values of different choices, but the way the choices
interact. Suppose that Brigantines are the fastest ships in the game and Dreadnoughts
are the slowest. From a gameplay standpoint, it’s getting more interesting, but what we
are concerned with now is not gameplay but game balance. Where component balance
dealt with comparison of the three ship types, attribute balance (in this example) is a
question of assessing how important speed will be as opposed to combat ability. If the
game levels are small, are based around fixed goals (such as buried treasure), or operate without a time limit, then combat ability far outweighs speed. If the goals are merchant ships that lack firepower—and if there is no direct imperative for players to
attack each other—then speed becomes far more important.
You can envision this as a number of sets, with each set establishing the relative value
of the Brigantine, Galleon, and Dreadnought with respect to one factor: speed, firepower, upgradability, range, and so on. Attribute balance is then a question of weighting
each factor relative to the others. For example, range may turn out to have no value if
the levels are small. This allows you to construct an average set combining the effects
of all factors. Component balance then requires you to adjust the ships’ interactions
and/or cost to ensure that they all have a useful role in the game.
As a mnemonic, component choices are the ones that are usually embodied by artifacts
in the game: “Hmm, the Vorpal Blade and the Black Javelin are both useful weapons,
but I think I need the Wand of Ill Omen to deal with this troll.” Attribute choices are
more abstract and have to do with how you use the artifacts: “Maybe I’d better not
tackle that troll at all; I can get by using stealth.” Case Study 5.3 provides an example
of component and attribute balancing.
06 3634_CH05
10:35 AM
Page 119
Game Balance 119
Case Study 5.3 Component and Attribute Balance in
Dungeon Keeper
Bullfrog’s Dungeon Keeper is a simulation in which you construct an underground
magical lair while fending off occasional attacks by heroes.
Component balance exists in three forms: between different room types, between
different spells, and between different creatures. You need to decide which room
types to build first and how big you can afford to make them. During a fight, you
need to pick the best spells to use. Creatures arrive at random, and you don’t have
much control over which ones you get, but you do need to decide which are worth
spending treasure on to improve their skills.
Attribute balance occurs in several forms. The most obvious is the relative placement
of dungeon rooms. Creatures must not have to go far to get food or pay, because you
want to minimize the time they aren’t training. Another form involves which creatures
are the best to deploy against different intruders. Faced with fast-moving intruders
such as fairies, you need to hit them quick and hard. Slower enemies like wizards
allow you to skirmish, continually pulling back creatures for rest and recovery. Deep
in your dungeon, too, your aim will often be to capture and convert intruders instead
of killing them.
There is another level of attribute balance in this game that is not widely seen:
the balance between two subgames. Dungeon Keeper can be played either with an isometric third-person viewpoint or by possessing one of your creatures to get a firstperson view. Obviously, the former is better for organizing your dungeon. To encourage you to use the first-person view, the game gives an advantage to your side in any
fight where you possess a creature and personally get into the scrum. Also, there are
some levels where the only way to win is to go to the first-person view.
But was it necessary to achieve that last level of attribute balance? There’s a question.
Suppose you didn’t care for the first-person viewpoint, and you just wanted to play the
dungeon sim? Too bad. If you don’t get through those first-person levels, you won’t be
able to complete the game. So there is no doubt the two subgames were balanced, in
the sense that you couldn’t dispense with either. But whether this particular aspect of
balance enhanced gameplay, interactivity, and player choice is another question.
06 3634_CH05
10:35 AM
Page 120
120 Chapter 5
Not surprisingly, attribute balance is a lot harder to get right than component balance,
which is really just a question of assessing the relative value of different strategies for a
given problem. (Attribute balance has to consider the set of all problems.) Remember
that strategy in this context is not restricted to strategy-type wargames, even though
such games provide a convenient example of these concepts. Strategy simply means the
way the player uses the available game choices—the sequence of moves in a beat-’emup, the mix of units in a wargame, and so forth.
This gives us the key to game balance. If we can form an approximate picture of the
better strategies—that is, the relative frequency that different choices will most
probably be used, even only to the level of a ballpark figure—then we can start to
tweak the costs and compensating factors to reflect that frequency.
Intransitive Game Mechanics Guarantees Balance
Consider a martial arts beat-’em-up game with three main attacks: forward kick, leg
sweep, and stomp. If Al chooses a leg sweep and Beth chooses a forward kick (Figure
5.1), Al dodges Beth’s kick and takes her legs out from under her: a win to Al. If Al
chose a stomp and Beth chose a forward kick, her kick catches him, and now Beth
wins (Figure 5.2). But a stomp beats a leg sweep (Figure 5.3).
Figure 5.1 Leg sweep beats forward kick.
06 3634_CH05
10:35 AM
Page 121
Game Balance 121
Figure 5.2 Forward kick beats stomp.
Figure 5.3 Stomp beats leg sweep.
06 3634_CH05
10:35 AM
Page 122
122 Chapter 5
We said earlier that we’d get back to Stone-Paper-Scissors.
First off, notice that the beat-’em-up example really is exactly equivalent to StonePaper-Scissors. The example we looked at in Chapter 3, based on The Ancient Art of
War, was actually a bit more complex than pure SPS. The reason for that was that you
could play a mix of unit types, and, after winning a battle, you would have some units
left, which would stand you in good stead in the next battle. Whereas in the case of
the beat-’em-up, you can play only one maneuver at a time, and, in each comparison of
maneuvers, there is one winner and one loser. So, in effect, each exchange of attacks is
a new subgame within the overall bout.
We’re playing this hypothetical beat-’em-up. What’s the best strategy? It’s an extremely
simple game, so it ought to be possible to see a good strategy pretty quickly (if there
is one).
Well, you could randomly pick your maneuver each time. This is at least as good as
any other strategy if you have no way of knowing what the other guy is going to do.
However, if you’re playing against a very dumb computer opponent who picks randomly, you can equal its score by playing a leg sweep every time. That way, you’ll each
win, draw, and lose one-third of the time.
Of course, you would soon spot an opponent who always chose the leg sweep. You’d
then start to use mainly stomps, and your opponent would then start using forward
kicks, and so on.
A useful way to represent this game is by means of an interaction matrix like the one
shown in Table 5.1.
Table 5.1 Payoffs in a Stone-Paper-Scissor-Type Game
Leg Sweep
Forward Kick
Forward Kick
Leg Sweep
06 3634_CH05
10:35 AM
Page 123
Game Balance 123
What this shows is your payoff for playing a maneuver (shown in the rows) matched
against the maneuver used by your opponent (shown in the columns). A win is awarded
a positive value, a defeat a negative value, and a draw by zero. This is because the game
as described is zero-sum. In other words, whatever one-player gains, the other player
loses. As we’ll see shortly, even when these payoff matrices are only a very abstract
model of a game, they can be useful in balancing different elements of the design.
The intransitive relationship typified by Stone-Paper-Scissors occurs in even the most
unlikely games. Nail gun beats shambler, shambler beats rocket launcher, rocket
launcher beats zombie, zombie beats nail gun. Or elephants beat chariots beat priests
beat elephants. Why is this such a commonly used game mechanic? To answer that,
remember the discussion in Chapter 3 about avoiding dominant strategies. A good
game design eschews “best” weapons or strategies so as to keep the player’s choices
interesting. A Stone-Paper-Scissors mechanic simply happens to be a very easy way
to make sure no one option beats all others. A great arena for Quake has no perfect
vantage point, so to be top dog, a soldier must stay on the move.
But that isn’t the whole story. Figure 5.4 shows one of the possible evolving patterns of
an SPS-based game. Each point of the triangle represents one of the options, and the
line shows how a given player’s choices will tend to drift over time, with the proximity
to each point indicating the likelihood of picking that option. To start off, suppose our
player’s favored strategy is always to pick the stone. The other player responds by picking paper. The response is to pick scissors. Quite swiftly, the player is forced towards
the middle of the diagram, representing a strategy where all options are equally
favored. There is no way to beat that particular strategy in Stone-Paper-Scissors.
Unfortunately, there’s no way you can win using it, either! Just like in the first law of
thermodynamics, you can only break even.
What if the costs of using each option weren’t the same? In the simple beat-’em-up we
were looking at before, suppose a stomp costs the most energy, and a leg sweep costs
the least. Obviously we’ll now see a different mixed strategy. We might expect to see
the stomp being used less often than the other moves. But wait—if the stomp is used
less often, that would make the forward kick less useful too. It’s not intuitively obvious
exactly what the optimum choices would be. We have to do some math.
06 3634_CH05
10:35 AM
Page 124
124 Chapter 5
Suppose that a stomp costs 3 Ki, a forward kick costs 2 Ki, and a leg sweep costs 1 Ki.
We also need to know how much Ki you knock off your opponent if you win an
exchange of moves. Say that’s 5 Ki. We can now rewrite the payoff matrix with these
costs factored in as shown in Table 5.2.
Table 5.2 Rewritten Payoff Matrix
Leg Sweep
Forward Kick
Forward Kick
Leg Sweep
This is the net payoff, remember. So, for instance, if I choose leg sweep and you choose
stomp, the difference in Ki cost to make those moves is +2 in my favor, but I lose -5
because your stomp beats my leg sweep, so the net payoff to me is -3.
Now, let’s call the net payoff for using each move L, F, and S, and the frequency that
each move is used is 1, f, and s. That means the net payoff for using a leg sweep is
(O × l) + (6 × f) + (-3 × s)
L = 6f - 3s
F = 6s – 6l
S = 3l - 6f
We also know that each of the three payoffs must be zero. Why? Because we are
looking for an optimum strategy. This is a zero-sum game, meaning that whatever
one player gains, the other loses. Both cannot make a net gain. Hence, if there is an
optimum strategy that they both follow, it must give each the same total payoff, which
must therefore be zero:
06 3634_CH05
10:35 AM
Page 125
Game Balance 125
And how do we know that the three components of the payoff, L, F, and S, are equal?
Because the mixed strategy we are looking for is the equilibrium that both players settle down to using after some time, as we saw in Figure 5.4. For it to be in equilibrium,
we couldn’t have the payoff for leg sweep be greater than for stomp, for example,
because then we’d start using leg sweep more. So when we finally get to the optimum
mixed strategy that we both will stick with:
Figure 5.4 Evolution of mixed strategies in Stone-Paper-Scissors.
It’s very easy to solve the equations now and show that the ratio
l:f:s = 2:1:2
which is a very interesting result. It says that leg sweep and stomp are each used 40%
of the time and forward kick is used 20%. Probably that isn’t what you’d have guessed
without doing the math. It neatly illustrates something you’ll see quite often with these
Stone-Paper-Scissors games: If one option is expensive, it’s often the other options that
are most affected.
This is all very well and good, but there is a caveat. Basing your game around StonePaper-Scissors guarantees balance because there cannot be a dominant choice. In such
a game, players must avoid predictability because predictability is a weakness that a
clever opponent will exploit. However, a game in which you must avoid predictable
06 3634_CH05
10:35 AM
Page 126
126 Chapter 5
choices is barely one step ahead of a game with a single best choice. This is why
intransitive relationships like Stone-Paper-Scissors are only the first and most basic
step towards creating rich gameplay. To balance those choices so that they are interesting to the player implies that you will want to design in other factors that influence
how the choices interact—be they weapons, maneuvers, units, or whatever. We saw
one example of that in Chapter 3: At Hastings, King Harold placed his troops on a
ridge so that the normal interaction (archers beat axemen) was changed to archers
equal axemen.
In the case of our beat-’em-up, one route to interesting gameplay is by including combination maneuvers. You can then put together a sequence that you hope will prove to
have a greater payoff than the individual maneuvers taken on their own. Conversely,
skill and judgment give you the chance to anticipate what the other player is leading
up to. It helps even more if certain sequences can build up to several different kinds of
“grand slam.” For instance, a stomp, two forward kicks, and a leg sweep might culminate in a scissors pin; while a stomp, two forward kicks, and another stomp might add
up to a knockout. If I play a stomp followed by two forward kicks, you know I can
summarily end the bout by playing a stomp. You don’t want that, so you think about
countering with a forward kick. But you know I know that’s what you’re thinking, so
maybe I’ll go for the leg sweep and merely pin you instead. And don’t take too long to
mull it over; time is also a gameplay factor! Case Study 5.4 provides an example of
Case Study 5.4 Attribute Balance Using SPS
Peter, the Lead Designer of Warbots, is discussing some gameplay ideas with
Charles, the Project Manager. Charles had requested a circular unit relationship, similar to Warcraft 2, and Peter is sketching out some possibilities.
“Say we have three main combat units,” he suggests. “The Drillerbots are fast.
Frighteningly fast. They can do lots of little attacks in quick succession that take the
slower Hammerbots apart. They make this horrible sound as they do it,
“I know, like my root canal work!” laughs Charles. “Ouch.”
06 3634_CH05
10:35 AM
Page 127
Game Balance 127
“Then you’ve got the Juggerbots. They’re slow too, even slower than the Hammerbots,
but they’re massively armored. The Drillerbot can’t even scratch them—sparks everywhere, but no damage. Meanwhile, the Juggerbot is pulling the Drillerbot apart like a
crab taking a leisurely snack”
“And the Hammerbots? They can get through the Juggerbots’ armor?”
“Crack ‘em like nuts,” grins Peter. “Only it’s not so one-sided as those other matches.
The Hammerbot wins, but it’s a brutal slugfest with scrap metal flying everywhere.”
“Good,” says Charles. “So there’s no best unit. Have you thought about what the
player can do to change the odds?”
“I’m going to talk to Nick about that. He’s still working on the physics system,
which will affect a lot of things besides combat. It could be that the Drillerbots use
laser drills, and they might get less effective as they move away from the player’s
power pylons, or maybe they weaken over time and then have to recharge.”
Charles nods. “As long as it’s an automatic recharge, like the units that use energy
in StarCraft. We don’t want players having to fiddle about sending energy to ’bots
from their resource stocks.”
“Agreed. Also, we could dispense with the mining ’bots. Drillerbots can mine as
well as fight, so they have a versatility advantage.”
Charles thinks about it. “No, keep the mining ’bots, but make them cheaper and
have the Drillerbots better at blasting the ore but not collecting it. It’ll give the player
some interesting choices. And what about Hammerbots versus Juggerbots?”
“Underwater—under liquid methane, I should say—the Hammerbot’s attacks will
probably pack less punch. The Juggerbot fights with these kind of vice-grip pincers,
which won’t be affected so much.”
“It all sounds fine,” says Charles. As a longtime Warcraft fan, he is quite satisfied.
“So much for combat. What about other factors—visibility range, say. Can you get
that to fit a cyclical pattern, too?”
06 3634_CH05
10:35 AM
Page 128
128 Chapter 5
This takes Peter by surprise, “Of course not…” he starts to say. Then, after a
moment’s thought, “Hey, maybe we could, at that! Suppose A has the longest radar
range but the shortest sight range. B has the longest sight range, just a bit under A’s
radar range, but has no radar itself. C has a medium sight range and again no radar,
but also is invisible to radar.”
He writes on a pad, plugging in some numbers. “Yes. A spots B before being spotted
itself. Same with B to C and C to A. Hey, that’s nifty! I never would have thought
you could get an intransitive relationship with visibility ranges.”
“See, I’m too dumb to know what’s not possible,” laughs Charles. “But enough of
this algebra stuff. Which units are we talking about here, A and B and C?”
Peter smiles as another realization dawns on him. “This has a big impact on gameplay.
Say we have the Jugger spot the Driller, and so on, so that each ’bot can see the ’bot
he’s best equipped to take out. That’ll make for a fast-moving aggressive game—the
predator can always see his prey before he is seen himself. But, if we try it the other
way around, so the Hammer sees that Driller coming, the Driller can avoid the Jugger,
and so on… then you have a slower, more considered kind of cat-and-mouse game.”
“Where the cats are made of titanium steel and the mice weigh eight tons apiece,
sure.” Sensing they could be one up on his favorite strategy game, Charles looks a
little bit like a cat with cream himself right now. “Anyway, which way is better?”
“Better?” Peter looks up from his notepad. He couldn’t look more astonished if
Charles had just asked him to calculate the value of the truth-beauty equation. “I
can’t say which is better. It depends on the kind of game you want. We could even
arrange it so that it sometimes works one way and sometimes the other. Making it
vary depending on weather or the day/night cycle is one option. More interesting is
if radar is messed up when there are lots of scrap robot chassises lying about, when
metal buildings are nearby, that kind of thing.”
“I see,” says Charles. “So ’bots might see their targets, pounce, there’s lots of vicious
infighting, but the more the bodies pile up the more difficult it gets to spot your
06 3634_CH05
10:35 AM
Page 129
Game Balance 129
“And the easier to spot the ’bots to avoid. Yes. But even that’s only part of it. We
also have to settle on the attribute balance between visibility range and combat
ability. Because, if the cost of building new ’bots is very high, winning every combat
becomes critically important. If they’re cheap, it doesn’t matter so much. And then
there’s the question of not just whether a ’bot sees another, but the ability of each
‘bot to close the gap or to get away.”
Charles nods. “Okay, you’re right. I was hoping for an easy answer for once, but
there’s only so far you can get with hand-waving and whiteboards, isn’t there?”
“I’ll get Nick to draw up a testbed. You’ll be amazed what a couple days of tweaking
can achieve.”
Combinatorial Explosions
In an intransitive game, how many attributes should you include that are capable of changing the basic SPS relation? It’s a question of game balance. Too few, and the use of those
factors to create a strategy becomes trivial. For instance, if the relationship is archers beat
warriors beat barbarians beat archers—and the only thing I can do in the whole game to
change that relationship is put my warriors on a hilltop, as King Harold did at Hastings—
that’s not exactly an intriguing game. On the other hand, too many modifying factors and
you have a game in which skilled play becomes almost impossible. Here’s a rule of thumb: If
you have N factors that could modify your core game mechanic, and each factor is Boolean
(in any situation it either holds or it doesn’t) then you have 2N possible combinations. The
complexity soon blows up out of control, so you should err on the side of caution.
“In Populous: The Beginning, almost the first decision was whether the game should
have lots of character types or only a half-dozen or so. We noticed straight away that it
would be easier to understand the game experience with a few, very versatile units
rather than many specific ones.”
—Richard Leinfellner, Executive in Charge of Production at Bullfrog
06 3634_CH05
10:35 AM
Page 130
130 Chapter 5
Design Scalability
A related problem arises when balancing many components. As we shall shortly see, it
is possible to construct extensions of the Stone-Paper-Scissors dynamic using five or
seven components—and, indeed, any odd number. If strategies can draw as well as
winning or losing, there are infinite possibilities. Where should you draw the line?
Intransitive component relationships are very inflexible to alterations in the design.
You cannot construct a five-way dynamic and expect it to still work if you remove one
of the components. If you want to try it, skip ahead to Table 5.3 and delete one of the
units: This immediately creates a dominated strategy, rendering one of the other units
redundant. Suppose you’re nine months through development, and work is slipping
behind schedule. The project lead asks you to take out one of the five characters
because there isn’t time to do all the animations for that character. However, this
causes your elegantly balanced design to fall like a house of cards.
On the other hand, it’s relatively easy to add extra components later in development. The
extra component doesn’t have to be symmetrical with all the others. It might be useful in
only a few cases, or it could be a component that is vital to the game but that is only
needed occasionally or in small numbers, like a Lay-the-Dead spell or a scout character.
The lesson is clear. If you’re going to scale your design, plan to scale up, not down.
Other Intransitive Relationships
Before examining the limits of a game-theoretical analysis, we’re going to digress.
Generally speaking, nontransitive relationships like Stone-Paper-Scissors are so useful
in games that it is worth taking a look at some other examples.
How about with more than three options? Is it still possible to construct an SPS-like
relationship? Fortunately, yes. (Otherwise, this analysis would be very limited indeed.)
Figure 5.5 shows another variation using five units.
Here, samurai beat shugenja and ashigaru. Shugenja beat ashigaru and archers.
Ashigaru beat archers and ninja. Archers beat ninja and samurai. And ninja beat samurai and shugenja. No one character type provides a dominant choice, as we can see by
drawing the payoff matrix (Table 5.3).
06 3634_CH05
10:35 AM
Page 131
Game Balance 131
Figure 5.5 Five-way intransitive relationship.
Table 5.3 Payoffs in a Five-Way Intransitive Relationship
Similarly, you can construct a seven-way relationship along the same principles. That,
as they say, can be an exercise for the reader. But Figure 5.6 depicts a less regular
Figure 5.6 A four-way variant on SPS.
06 3634_CH05
10:35 AM
Page 132
132 Chapter 5
This is a variant on the original example taken from The Ancient Art of War. Archers
beat warriors beat barbarians beat archers, as before. Only now there is a new character, the sorcerer, who is beaten by the warrior, beats the barbarian (they’re the superstitious ones, let’s say) and is an even match for the archer. The payoff matrix again
shows that none of the characters is invariably best (Table 5.4).
Table 5.4 Payoffs in the Four-Way SPS Variant
Against a randomly selected opponent, the barbarian doesn’t win as often as any of the
others. Does this mean players won’t use the barbarian as much? Well, no, because, as
soon as I see you aren’t using the barbarian, I will start to play more archers, and your
only counter is to start using the barbarian again. Case Study 5.5 provides an example
of using theory analysis.
Case Study 5.5 Using Game Theory Analysis to Achieve
Bloody Century is an operational-level wargame set in an alternate history where
WWII dragged on into the 1950s, The designer, Gary, explains his concept to the
development leads: “Basically, it’s nice and simple with just five units—tanks,
infantry, motorized infantry, artillery, and tank killers. Tanks beat everything except
killers. Killers beat nothing except tanks. Infantry, motorized infantry, and artillery
compete in a standard Stone-Paper-Scissors form.”
Gary goes to the whiteboard and draws the payoff matrix he has in mind. “Bear in
mind that this is just a first-pass analysis,” he cautions the team. “Unlike in StonePaper-Scissors, units in a strategy game aren’t used up every time they fight. So, if
my tank destroys your infantry, I win a +1 payoff and I keep my tank. How damaged
it may be, of course, is another matter. Here I’m assuming a victory costs about 50%
of a unit’s hit points.”
06 3634_CH05
10:35 AM
Page 133
Game Balance 133
Everyone studies this (Table 5.5) with furrowed brows, and some coffee is drunk. At
last, the lead level designer, Sarah, ventures a comment. “It doesn’t look very well
balanced. Tanks beat almost everything, and killers beat almost nothing!”
Table 5.5 Payoff Matrix for Bloody Century
Motorized infantry
Tank Killer
Motorized Infantry
Tank Killer
“Yes,” concedes Gary, “but because tanks are effective they’ll be used frequently.
Because they’re used frequently, that makes tank killers very useful too.”
“It sounds like a circular argument to me,” decides Oliver, the company president.
‘What I want to know is, what’s the game going to be like to play? Will we see lots
of tanks, or killers, or what?”
Sarah has a further point “And what about the typical mix of units in the game? Are
you saying tanks and killers will be used twice as often, or 10 times as often, or
what? I’d think players will expect the other units to get a look-in.”
“Okay.” Gary wipes out the first matrix and starts to draw a new one. “It turns out
that what I’ve described is one SPS game within another SPS game. It’s clearer if I
write it out like this.” (See Table 5.6.)
Table 5.6 Reduced Payoff Matrix for Bloody Century
Tank Killer
Tank Killer
0 (on average)
“So what does this tell us?” Oliver wants to know.
06 3634_CH05
10:35 AM
Page 134
134 Chapter 5
“The way I’ve assigned the payoffs assumes all units cost the same to build,”
explains Gary. “On that basis, the tanks and the tank killers will be used equally,
with the same frequency as all the other three unit types lumped together.” He
draws a table (see Table 5.7) to illustrate.
Table 5.7 Unit Usage Frequency
Unit Type
Frequency of Use
Motorized Infantry
Tank Killer
“I don’t think much of that!” says Sarah. “You’re going to see three times as many
tanks as infantry units. It just doesn’t feel right.”
“Ah, but we can change that by changing the costs,” insists Gary. “These are the frequencies you’d see if all units cost the same.”
Even Oliver has cottoned on by now. “So what happens if the tanks cost twice
as much?”
“Well, it’s fairly simple, but I won’t go into it all now because it involves a bit of
math, and it’s easier to do on a spreadsheet. Basically, we’d start by drawing a new
payoff matrix matching X dollars’ worth of tank versus X dollars’ worth of infantry,
X dollars’ worth of motorized infantry and so on.”
“For instance?”
“Okay, say a tank costs $3,000, and an infantry unit costs $1,500. So, the matrix
would now pit one tank against two infantry. I said that one tank beats one infantry
but loses 50 percent of its hit points in the process. That means that the tank will
destroy one of the two infantry before it’s destroyed, so the two infantry win with a
net payoff of $1,500.”
“I thought tanks beat infantry,” says Oliver.
06 3634_CH05
10:35 AM
Page 135
Game Balance 135
“Not any more. On a hit-points-for-dollars basis, infantry are now better. You see, I
lost one tank worth $3,000, and you lost one infantry unit worth $1,500, so the net
worth is $1,500 in your favor.”
“Isn’t there a risk that, with the new matrix, with the costs factored in, you could
get dominated strategies?” wonders Vishal, the lead programmer.
“Yes, you’ve got to check for that. There are bounds to the costs. If you go outside
those bounds you get dominance, and then the game can break down. It might
never be worth using artillery, say, and that might mean it’s never worth playing
motorized infantry, and so on. But all these things are easy to see when you draw
the matrix.”
Sarah, at least, is on the road to Damascus by now. “It’s excellent. This means I can
get some idea how changes in shadow costs on different levels will affect how units
are used. That’ll be very handy.”
Unlike Stone-Paper-Scissors or the more complicated five-option variant just discussed,
this has the benefit of asymmetry. The asymmetry makes it more aesthetically appealing, because the player doesn’t merely have to learn a cyclical pattern of win/lose
relationships. The archer and the sorcerer, although they draw with each other, are
interestingly different. Or are they? Let’s do a little math.
We would like to know how often these characters might be used in the game. That’s
the purpose of these payoff matrices, after all: to give some idea of the value of different choices so we can balance them. In this case, let’s call the probability of using these
characters a, w, b, and s, and the net payoff for using a particular character is A, W, B,
and S. We can extract six equations: one from the fact that all of the probabilities will
sum to unity (you have to choose one of the units), four directly from the payoff
matrix, and the last from the net payoff rule.
Thus, we have
1. a + w + b + s = 1
2. A = w - b
06 3634_CH05
10:35 AM
Page 136
136 Chapter 5
3. W = b + s - a
4. B = a - w - s
5. S = b - w
6. A + W + B + S = 0
You might wonder about the last assertion. Why does the net payoff have to be zero?
Well, this is a zero-sum game. We are looking for an optimum mixed strategy. If it were
possible for you to find a mixed strategy with better than zero payoff, your opponent
could use it too—and, logically, you couldn’t both get a positive payoff.
Adding the payoff equations (equations 2, 3, 4, and 5) gives
(w - b) + (b + s - a) + (a - w - s) + (b - w) = 0
Everything cancels out to leave (b-w) =0, which tells us that
The warrior and the barbarian must be used equally often. This means the payoff for
playing the archer or sorcerer is 0, and the net payoff equation (6) is now
We can go further than that. If we’re getting a payoff only from warriors and barbarians, we know that the payoff W can’t be positive and the payoff B negative. If they
were, it couldn’t be an equilibrium. Obviously, you’d play warriors more often and
barbarians less often to increase the net payoff. That would violate our finding that
w = b. The only way to make it an equilibrium, then, is if
And this gives
(a - s) = w = b
06 3634_CH05
10:35 AM
Page 137
Game Balance 137
This is an equation to which there is a range of solutions, bounded by the cases of
(a, w, b, s) = (_, 0, 0, _), and (a, w, b, s) = (_, _, _, 0)
This tells us that, although not a dominant choice, the archer is the most certain to be
used because there is no possible strategy that can afford to dispense with it. More
importantly, though, we can now see that this relationship on its own doesn’t guarantee the dynamic gameplay of Stone-Paper-Scissors that we saw in Figure 5.4. You could
play using only archers and sorcerers, and that would be a stable strategy. On average,
you’d gain nothing by fielding a few barbarians, for example (although neither would
you lose). Certainly, as a designer, you don’t want it to be possible for players not to
bother with two of the character types in your game! Accordingly, to balance things
out, you would have to make sure there are more special situations that favor the other
characters. For example, the sorcerers could be a little bit more powerful at certain
times, or (more probably) have some noncombat functions; the barbarians could get
extra scouting abilities or recover from injury faster; warriors could be the only
character type able to use magic swords; and so on.
How about other four-way relationships? Figure 5.7 shows the classic combined arms
diagram for troops in the pre-gunpowder era. (In fact, this was the core game dynamic
of Dave Morris’s design for Warrior Kings.)
Figure 5.7 Combined arms in ancient and medieval armies.
The principle is one we have touched on earlier. Archers can skirmish against infantry,
whose heavy armor prevents them from getting up close and personal. Heavy cavalry,
though, are fast enough to ride the archers down. Horse bowmen can pull that
06 3634_CH05
10:35 AM
Page 138
138 Chapter 5
skirmishing trick on both infantry and heavy cavalry, being faster than both, but they
cannot prevail against archers, because it is impossible to employ the most powerful
bows from the saddle. Infantry in regular order can stand firm against a cavalry charge
and can present a denser and more effective killing front (which means they win in
a melee).
That’s all fine and dandy but, as we’ve drawn it here, it’s not a valuable interaction system to use in a game. In fact, it would be a disaster. Try constructing the payoff matrix,
and you’ll see that infantry is a dominated choice. It seems, doesn’t it, that there’s
never a case where you’d be fielding horse bowmen and find yourself wishing you had
infantry instead? Did all those medieval generals get it wrong? Should they have
dropped infantry altogether? More to the point, did Dave goof up putting it into
Warrior Kings?
It’s a trick question. Infantry is useful because that combined arms diagram is a simplification. It tells us who beats whom, but it doesn’t say how. Taking another look at the
analysis of what’s actually happening reveals that infantry is the only troop type that
can actually hold territory. The others can win against each other, but they have to
advance or give ground to do so. So, returning to the “best at” rule of thumb we mentioned in Chapter 3, we can say that infantry is balanced against the other troop types
because it is best at holding territory. The importance that the game levels give to
holding territory then determines how often infantry will be used.
To verify that, take a look at Age of Empires. There, resources once collected are nonlocalized. A single villager can sneak in, build a base, and then you can “teleport” your
army right onto the enemy’s doorstep. Supply lines have no meaning. Consolidation of
territory is therefore much less significant than in real-world warfare. And, in consequence, even when the designers halved the population-slot cost of infantry (in the
expansion pack, Rise of Rome), they still couldn’t succeed in making them an attractive
option. This doesn’t particularly harm Age of Empires, which remains a great game; it
just means that the infantry units might just as well not be there.
06 3634_CH05
10:35 AM
Page 139
Game Balance 139
A Game Balance Checklist
Balance in a game takes several forms.
Balance between players ensures the game is fair. Any unfairness is obvious when the
game is played between opponents of equivalent skill, making this aspect of balance
increasingly important as multiplay between human players becomes increasingly the
norm. The only method of making the game absolutely fair is by exact symmetry in
the choices available to all players. However, there may be other reasons for avoiding
exact symmetry (because it is unrealistic or unaesthetic, for example). A just-broken
symmetry can solve this as long as any asymmetries are not likely to magnify in effect
throughout the game. Alternatively, you can provide the players with completely different choices so that each player will use radically different strategies. Absolute balance
is then all but impossible. If you opt for prominent asymmetry, be sure the key differences are peripheral and do not interact to magnify any imbalance that may exist. It is
possible to get away with extreme asymmetry if any unfairness arising from the asymmetry does not become glaringly obvious within the playable lifetime of the game.
The golden rule of player/player balance: A player should never be put in an
unwinnable situation through no fault of their own.
Balance between the player and the game mechanics ensures that the player never
becomes frustrated by an inability to progress. The interface should not present an
unnecessary obstacle to whatever the player is trying to achieve. Remember that small
rewards are needed to guide the player along the game’s learning curve. These rewards
can be cosmetic (a fancy animation, a cut scene) or integral (experience points, further
plot developments, a new weapon). In either case, the best rewards are the ones that
widen the player’s options because this enriches the gaming experience. Having the
player’s avatar go up a level and get more hit points is meaningless if you only make
all the computer-player opponents tougher as well; rather, you need to surprise and
delight the player with new abilities and new kinds of opponent.
The golden rule of player/gameplay balance: The game should be fun to learn as well
as to play, and it should be more fun the more you master it.
06 3634_CH05
10:35 AM
Page 140
140 Chapter 5
Balance between different elements of the game system operates on two interdependent
levels: components (the entities or choices within the game) and attributes (the interaction between components). The purpose of gameplay/gameplay balance is to make
sure that no element of the game is redundant. You can get a useful guesstimate of the
relative value of different game components by drawing up a Q-factor list for each
component. The Q-factor need be no more than an arbitrary rating from one to ten
against each attribute. In a strategy game, for example, you might begin by listing each
unit’s scores with respect to firepower, hit points, mobility, cost, vision range, armor,
etc. The easiest way to ensure that no component is dominated (and therefore dispensable) is to recall the advice of Chapter 3, and make each component the best in at
least one attribute. Relationships following the pattern of Paper-Scissors-Stone ensure
that each component is dynamically rather than statically best. Such patterns oblige
the player to continually alter his tactics and thus help create an exciting game.
It is not necessary to make every component equally useful to the player, nor is it likely that such a goal could be achieved in any but the simplest or most abstract of
games. However, the cost, availability, or ease of use of each component should reflect
its value to the player. This is best refined by continual playtesting throughout the
development cycle.
The golden rule of gameplay/gameplay balance: All options in the game must be worth
using sometimes, and the net cost of using each option must be commensurate with
the payoff you get for using it.
07 3634_CH06
11:12 AM
Page 141
Chapter 6
• Creating immersion
Look and Feel
• Ambience
• Interface issues
n a sense, we now come back to the beginning: Nine out
of ten game concepts start with nothing but the look and
feel. Take a hypothetical designer. Before he knew he was
going to do a horror CRPG, before he decided on the title
Sepulcher, before he thought of the spells with their side
effects and recharge locations, before he devised the combat
system that allows queuing of maneuvers—before all that, he
probably just had an image of a snowbound Gothic ruin and
men in top hats and ankle-length greatcoats approaching the
iron-studded front door with pistols in their hands.
Your game began life as merely a twinkle in your eye, but,
while writing the design, you had to put daydreams aside
and focus on the needs of cold logic. With the design documents prepared and the schedule tacked up on the wall, you
might easily have forgotten that you’re developing a creative
product and not a utility for a bank. Now, you must return
to that original wellspring of creativity in order to consider
the question of look and feel. It’s time to get artistic.
In this chapter, we’ll look at the various ways you can
create a sense of alternate reality so as to promote immersion:
the player’s sense of actually being in the game’s world.
Immersion is primarily achieved through three means:
ambience, interface, and storytelling.
• Raising player
expectations (suspense,
• A toolbox of storytelling
07 3634_CH06
11:12 AM
Page 142
142 Chapter 6
Ambience is everything that contributes to the innate look and feel of the game. In the
lead-up to the release of the Playstation 2, there were some impressive publicity screenshots of Tekken. As well as the two combatants, the screenshots showed a darkly
atmospheric back street and a couple dozen hooting, jeering, fully rendered spectators
gathered in a deserted parking lot to watch these two guys beat each other up.
Having gotten this far through the design section of the book, you don’t expect us to
start drooling over GFLOPS and trilinear filtering. Instead, let’s just take the technology for granted. (As good as it is now, it’s going to get better still.) In the same way that
technology like the SteadiCam has transformed movies, in games it isn’t how it’s done
that matters now; it’s what can be done.
Think about those old-time beat-’em-ups—two fighters in a sparse setting, which was
often no more than a bare stage in the middle of a deserted temple square. Now, you
can have a realistic-looking crowd of spectators in an almost photographically real setting, and that obviously creates more of a story right away. Why are you here to fight
this guy? If you win, who do you have to fight next—that giant in bike leathers who’s
glaring at you from the sidelines? The sneering psycho with the knife? If you lose, how
will the crowd react? Will you have to fight them all? And, if you run, how can you
find your way out of this rat’s maze of back streets?
Okay, Macbeth it isn’t, but suddenly there’s far more depth, richness, and drama than
the beat-’em-up genre ever promised before. All because of a few thousand extra polygons, the game can now include all kinds of ambient features that create a sense of
location and involvement. The designer can hint at a backstory that goes beyond just
beating another guy to score some points. It doesn’t matter that none of those things
mentioned may actually occur in the course of a game. As far as the game logic is concerned, it could still be the same beat-’em-up as before. But now it has atmosphere.
More than that, the fact that you can put in something extra means that the decision to
leave it out can have added significance, too. Imagine a string of bouts in the parking
lot with dozens of street scum looking on. Then the action switches to a dank warehouse. In a makeshift ring marked out by packing crates, you take on your next few
07 3634_CH06
11:12 AM
Page 143
Look and Feel 143
opponents by the light of a single, bare light bulb. A few spectators in fancy suits stand
in the background, their faces cut across by indelible shadow. One of them is smoking;
you glimpse the light of his cigarette. But, unlike the street scum, these guys watch in
silence. The location changes again, and now it’s an empty locker-room with just you
and your opponent and no one to see you fight. We’ve come full circle, but now the
sparse location and lack of spectators tells a story.
That’s what ambience is about: telling a story. Not with intro FMVs and cut-scenes—
that’s cheating, and movies already do that kind of storytelling better anyway. The story
should unfold as part of the game itself, and ambience is all about imbuing each separate element with the backstory so as to create unity in the final game.
We can broadly categorize the contributing factors to ambience as sound, vision, and
The excellent thing about sound in games is that most of the time you hardly notice it,
which means it can work on your subconscious and draw you into the game world
without you even being aware of it. Think of the wistful guitar theme in Diablo, the
stirring call-to-arms of Warcraft, the almost ethnic rhythms of Age of Empires and
Populous: The Beginning. While you’re engrossed in the gameplay, the music adds much
to the sense of atmosphere. And not only the music but other sound effects, too. Case
Study 6.1 identifies several brilliant sound effects.
Case Study 6.1 Sound Effects at Their Best
Bullfrog’s Dungeon Keeper has a particularly effective range of sound effects: the
dank dripping sound evoking the sense of vast subterranean space; the echoing
chip of pickaxes on veins of gold; the burst of crumbling rocks as your imps
carve out new tunnels; the distinctive sounds of the different rooms, the constant
mutterings, chants, and eerie little chuckles that quite brilliantly contribute to the
ambience of a living, magical, sinister place deep below the earth.
07 3634_CH06
11:12 AM
Page 144
144 Chapter 6
Sound doesn’t have to be only atmospheric, of course. The best elements of a game are
those that enhance ambience as well as gameplay. Thus, the priestly chants and war
cries of Populous: The Beginning don’t add only to look and feel, they are also important
cues that tell the player what is happening, And, in Looking Glass Studio’s excellent
Thief, both atmosphere and gameplay are enhanced by the guards’ muttered comments
of “Come out, taffer” (indicating they’re looking for you) and “It was just a rat…”
(indicating that you’re safe…for the time being).
When work began on Warrior Kings, Dave Morris’s first priority was to get the design
documents written so the team could press on with the first tier. For once, look and
feel hadn’t been uppermost in his mind, because he started the project knowing that
Eidos (the publisher) wanted something medieval. Dave had already written an unpublished design, Dominion, that he knew he could cannibalize for the sort of gameplay
features Eidos needed. So, in the parlance of Chapter 1, “First Concept,” to some
extent he skipped the inspiration phase and went straight on to the synthesis.
Three months down the line, the team had completed the first tier to test the resource
management. You could spawn peasants and townsmen and get them foraging, farming, and shopping. You could also place a few key buildings. (They sprouted instantly,
though, as the stages of construction would be slotted in later.) Finally, there was a
placeholder interface and blank slots to show where the mini-map and unit status
boxes would go.
The team worked brilliantly, and the tier was completed a few days early. Everyone was
impressed. A senior executive came round and looked at it. Afterwards, in his office, he
told the team leads, “The king’s palace is rubbish. This game is supposed to be set in
the Middle Ages, but that palace looks like a Disney World attraction. You’re supposed
to see to these things.”
The logical first reaction was why couldn’t he see past the placeholder graphics to the
fact that the team had gotten some working gameplay? Didn’t he understand the point
of the iterative development process?
07 3634_CH06
11:12 AM
Page 145
Look and Feel 145
But the executive had a valid point. Because Dave had been concentrating so much on
designing the gameplay and a new development process, he had kind of overlooked
the style. He was thinking that it could come later, but Eidos had several artists already
assigned to the Warrior Kings project, and, in fact, it was up to the designer of the
game to give them some direction.
One of the members of the Warrior Kings team was Martin McKenna, a very talented
artist from outside the games industry who had a strong track record with fantasy book
covers and role-playing games. He and Dave discussed various ways to give the game a
unique flavor. The Middle Ages is such a common setting that they didn’t want to
rehash something that had been done a hundred times before.
In search of something different, they looked at Eastern European buildings. Martin
started talking about crooked steeples and ivy-covered castles in Grimms’ fairy tales.
Dave was thinking of The Cabinet of Dr. Caligari and photographs of the Kremlin under
a crystal-clear night sky. They remembered the fine, dark look of Jim Henson’s
Storyteller TV series with John Hurt, and the narrow, crabbed buildings of comic book
geniuses like Steve Ditko, Berni Wrightson, and Mike Kaluta.
Martin started producing concept sketches and feeding them to the 3D artists.
Suddenly Warrior Kings started to look like no other game we’d seen. Coupled with
Julia Hunt’s Brueghel-like 3D characters and Angelo Bod’s atmospheric buildings, it
was becoming the Middle Ages seen through a filter of folklore and dream. You could
imagine the superstitious dread of those little scurrying characters as they hurried
home with their bundles of wood, the bustle and squalor of the medieval streets, and
the awesome spectacle of a battalion of knights drawn up for battle on the fields beside
an impregnable castle.
This example shows the great value of conceptual artwork in creating ambience.
Concept artists do not necessarily have to be computer artists. It may be better if
they’re not, as computer art tends to force a finished look, while artistic ideas are often
best developed in a sketchy half-formed way that fosters creativity. The ideal concept
artist is, like the game designer, more concerned with broad strokes and capturing a
general style than in finishing the fine details.
Figure 6.1 shows the kind of artwork we are talking about. Using such pictures in early
documentation helps convey a shared vision to the whole team.
07 3634_CH06
11:13 AM
Page 146
146 Chapter 6
Figure 6.1 One of Russ Nicholson’s concept sketches for Bloody Century.
Attention to Detail
Games achieve cohesion by adhering to a unified look and feel. We decided, for example, that the FMV for Gaze (as we’ll see in Case Study 6.5) should eschew the use of
the color red. The result is a look made of metallic blues and dusty bronze, implying a
world from which the life has been bled. The lead character is then introduced at the
end of the FMV as he slots a carnation into his lapel—a scene made all the more striking because it is the first splash of red used in the game.
Remain true to your style. When there is an element of the art that doesn’t fit with the
rest of the style, the effect is jarring as shown in Case Study 6.2. Concept artwork
assists by showing every member of the art team what you are striving for.
Case Study 6.2 A Discordant Note
When we first saw the promotional FMV for War of the Worlds, we were very
impressed. The weird look of the aliens in their council chamber on Mars banished
all thoughts of Mars Attacks! These insidious tentacled creatures looked like they
had been plotting the invasion of Earth since primordial times. And the Martian war
machines were exactly what H.G. Wells must have intended: massive Edwardian
07 3634_CH06
11:13 AM
Page 147
Look and Feel 147
industrial tech, with valves and pistons and brass finishing. They picked their way
through the ruins of London with the dispassion of insects.
But then one bit of artwork went and spoiled it. In a sequence on the bank of the
Thames, clumping along the embankment came—surely not?—a gun-toting bipedal
robot like something straight out of RoboCop. It couldn’t have been more intrusive
if the lid had popped open and Bugs Bunny had looked out. It was a damned
shame, because they had a beautiful look and feel going, and that robot blew it
away in a second.
“Okay,” you must now be saying to yourself, “sound and vision are obvious enough,
but what has touch got to do with computer games?”
Well, we admit we’re cheating a little bit. By touch, we don’t literally mean the physical
sense of touch, we mean the impression of physicality conveyed by the game’s look and
feel—the handling of the game, so to speak.
This is important. In the early days of animation, cartoonists had a hard time getting
their characters to feel right. The animated characters didn’t seem to move with weight,
momentum, and force. It was early pioneers like Walt Disney who found the techniques of accurately portraying those physical attributes in a medium relying on sound
and vision.
So, it is kind of a cheat, because you have only sound and graphics with which to get
the feel across, but that physical sense is a big part of games. Contrast the comic book
acrobatics of Virtua Fighter with the solidly realistic physics of Bandai’s Ichigeki. Or the
rolling, bouncing vehicles of Big Red Racing with the authentic road handling found in
Metropolis Street Racer.
One of the most interesting talks at the Develop conference in London in 1996, was
given by theater director Graham Bruce. He was discussing how he coached actors for
motion capture so as to help them create a sense of the physical environment of the
game world, using the specific example of Actua Soccer. It was an important reminder
that design documents should always convey the physical feeling of the game environment as well as how it looks and sounds.
07 3634_CH06
11:13 AM
Page 148
148 Chapter 6
The ideal interface is transparent. You would prefer simply to think something in order
to make it happen. Lionhead’s Black & White is one approach to this. The player casts
a defensive spell with a sweeping movement of the mouse, drawing a circle of flame
around him. To throw a fireball, he jabs the cursor towards his target. The designer’s
intention seems to be that the various mouse gestures will become second nature, so
that you can make things happen in the game world almost without thinking.
But when we say the ideal option is the transparent interface, we should add a qualifier.
Transparent doesn’t have to mean invisible. In a racing game, you wouldn’t want an
interface that let you drive your car intuitively, because that’s not how cars are controlled in real life. The transparent interface in that context wouldn’t be one like Black
& White, but one that felt as close as possible to the experience of actually driving a car.
You can help keep the interface transparent by making it enhance the look and feel.
This is the “hide in plain view” approach. StarCraft does this. There is no getting
around the unit status boxes and command icons in such a complex wargame, so the
designers made a virtue of necessity: Each species has its own custom interface that
adds flavor to the playing experience. Case Study 6.3 provides an example of meshing
the interface with look and feel.
Case Study 6.3 Meshing the Interface with Look and Feel
In 1996, Dave Morris wrote a strategy game called 2020: Knife Edge for Domark.
Units could be controlled on two levels: imperatives (direct commands like “go
here,” “attack this building,” and so on) and protocols (general commands like “act
defensively,” “act aggressively,” and so on). It was a game of near-future warfare (set
in 2020, naturally), and so a high-tech interface was appropriate. You would click on
a unit and issue commands from a head-up display just as if you were in the
Pentagon War Room. The project lead, Lee Briggs, described the interface style we
were striving for as “a civilian implementation of a military interface, like you see in
Star Trek: The Next Generation.”
07 3634_CH06
11:13 AM
Page 149
Look and Feel 149
Later, Dave came to write Warrior Kings. In most strategy games up until that time,
units like towers and archers were particularly handy because you didn’t have to tell
them what to do. Ideally, you’d like to set standing orders for all your units, and to
have some flexibility into the bargain. (That is, maybe you don’t always want archers
to skirmish.) Dave’s goal was therefore for the player to be able to set behaviors for
his units so that the game had a strategic and a tactical level, like 2020: Knife Edge.
Players could set large-scale plans using the protocols and fine-tune them in battle
by grabbing individual units and issuing imperatives.
But the high-tech style of 2020: Knife Edge would not have suited Warrior Kings at
all. Clicking on a unit and setting its standing order AI didn’t really mesh with the
medieval setting. Like placing “way points” for patrolling units, it’s a mechanism
that tends to jar you out of the game world. The solution Dave found was to link the
behavioral protocols to different formations so that the player tells the battalion to
draw up in line, column, wedge, or whatever, and this sets the AI for the battalion.
Also this interface became easier to use than the one in 2020: Knife Edge because,
instead of having to click on a unit to check its standing orders, you could see at a
glance from the formation.
Dave adds: “Don’t think I’m blowing my own trumpet. Yes, it was a good idea. But I
should have had that idea a year earlier when I was designing 2020: Knife Edge!”
Sometimes you can’t hide the interface, and you can’t make it enhance ambience either.
It’s quite hard to make inventories and automaps atmospheric when you’re trying to
do an Arthurian role-playing game, for instance. Then at least be sure to minimize the
degree to which those aspects of the interface intrude. The “save game” feature is
usually an intrusive element of any interface, as it destroys the illusion of reality. The
designers of Appeal’s Outcast, however, found a way to incorporate it into the game
world. You save the game by activating a magical crystal that stores an instant in
time. (And, since it produces plenty of light and noise as it does, potentially alerting
your enemies, you have to choose carefully where to use it.) Case Study 6.4 provides
examples of keeping the interface out of the players’ way.
07 3634_CH06
11:13 AM
Page 150
150 Chapter 6
Case Study 6.4 Sometimes Less Is Less
If you can’t make the interface transparent, at least keep it out of the player’s way.
An early testbed of an RTS called Plague had a mini-map that appeared in the bottom left-hand corner of the screen, as in Figure 6.2 (mockups courtesy of Turn3D
Inc). Of course this is highly intrusive and a better solution is that found in Age of
Mythology, where the command icons and selected unit data appear in a continuous
bar filling the lower part of the screen, as in Figure 6.3.
One of the team had an objection. “We don’t need the command icons there all the
time. Once a player has learned the game, he might want to play with just intuitive
right-clicks. For special actions, I’d prefer to just use the keyboard.”
“What advantage do we get by hiding the command icon window?” the designer
asked him.
“The screen’s less cluttered. The only thing you need to look at apart from the main
view is the minimap.”
“In fact, that makes it more cluttered,” the designer explained. “When you put the
interface across the whole lower portion of the screen, it becomes a frame. A frame
doesn’t obtrude on the main view. But as soon as you take away part of the frame,
the player will see that minimap quite differently. Now it’s an obstacle getting in the
way of whatever’s behind it.”
07 3634_CH06
11:13 AM
Page 151
Look and Feel 151
Figure 6.2 Testbed—the minimap is intrusive.
Figure 6.3 Finished Game—the interface frames the view.
07 3634_CH06
11:13 AM
Page 152
152 Chapter 6
Another approach to interface design is to make the game controls intuitive so that
they second-guess what you want. For instance, in Age of Empires, a right-click is interpreted according to what you’re clicking on, If you click on an enemy unit, the command is interpreted as “attack”; if on an empty location, it’s interpreted as “move,” and
so on.
A third option is to second-guess what the player wants or will do according to what
he’s done so far. The game version of Blade Runner infers which of three women the
hero is attracted to by the amount of time he spends with them. This is nice because
it’s very subtle; the player is not even conscious of making a choice.
Allegedly, two factions were involved in the design of Doom. One faction thought that a
strong setting and backstory would enhance the players’ appreciation of the game. The
other took the contrary view: “Story? We don’ need no steenkin’ story!”
Since the release of Half-Life, surely nobody now needs convincing about the value of
story. Sure, in the bad old sense of a story that the game designer had set out to tell
you, storytelling in games was worse than redundant; it was an intruder from another
medium. Games are meant to be interactive. That’s the whole point. They shouldn’t
aim to turn the player into a passive spectator. The designer who wants to do that
should write a novel instead.
But Half-Life didn’t just tell you a story. It helped you create a story. This is more than
simply a worthwhile way to bring storytelling into entertainment software; it is the
philosopher’s stone that turns dross into gold. You see, “look and feel” is about creating
atmosphere so that the player is able to suspend his disbelief, but that is all wasted
effort unless you can make the reader want to suspend disbelief.
If stories did not matter, boxers would never pretend to hate each other before a fight.
Politics would be decided on the issues and not the personalities. Star Wars would
consist of nothing but fast vehicle chases and swordfights. Yet we need to know who
Luke Skywalker is, and why he’s in this spaceship, and who is chasing him through an
asteroid belt, before it really matters to us. All the special effects in the world are just
wallpaper until you have a story.
07 3634_CH06
11:13 AM
Page 153
Look and Feel 153
A Toolbox of Storytelling Techniques
People enjoy stories for lots of reasons. Early man used storytelling to map the landscape around him. A good chat-up line is the beginning of a story. Our whole social
life is woven around the exchange of stories in various forms.
In storytelling, less is more. The most effective films don’t consist of chunks of static
exposition interspersed with bursts of action. Facts are not stated to the audience, but
they are revealed for the audience to notice and interpret. Billy Wilder said that if you
let the audience work things out for themselves, they will love you for it.
It’s true in games too. What the player experiences becomes the story. Elements of plot
should be there to be discovered through action, not shoveled on between levels. Games
are a lean-forward medium, after all, and gamers tend to be bright people. So drop a
hint, plant a seed. Trust that the player will put it all together in a flash of insight.
Erica Wagner, literary editor of The Times, has this to say about the future of entertainment software:
“Sony has already trademarked something they call ‘emotion synthesis.’ Its aim is to
involve you in games the way you become involved in films ... That’s what I want. I love
to read, to listen to stories and go to the movies because I want to be swept out of
myself and into another place. Just now—despite astounding 3D graphics and flawless
violence synthesis—I can’t find that place inside my PC screen.”
—Article in The Times “Interface” supplement, June 9, 1999
Erica Wagner is making a very good point about the direction that entertainment software needs to go, but you don’t need any special technology to tap into the immersive
power of storytelling. Movie makers have been doing it for 80 years. Storytellers knew
all the tricks for creating emotional involvement long before Shakespeare’s day.
Yet not many game designers seem to have the slightest idea about storytelling—so let’s
run through some of the key ideas here. Incidentally, before—when explaining game
theory—we said that it applied generally to all genres of game even though most of the
examples we looked at were from strategy games. Similarly, the clearest examples of
storytelling come from adventure games and CRPGs, but the same techniques can be
used in other genres also.
07 3634_CH06
11:13 AM
Page 154
154 Chapter 6
A little old man runs into the inn where the hero is staying and says to him, “There’s a
vampire up at the castle. You have to kill it.”
Very poor. Try this instead: A fearful old man enters the inn and avoids the hero. When
the hero asks what has frightened him, the old man gives no reply. Then he goes to a
character sitting by the fire and begs him for something, but is refused. “You pawned
it, you only get it back when you pay,” he is told. If the hero pays on the old man’s
behalf, he sees what the item is: a crucifix, “You should buy one yourself if you’re
going to stay in these parts,” remarks the pawnbroker.
Neither version is great literature, but the second is far superior. Why? Because it
involves an obstacle. Instead of being told the setup, the hero/player has to find it out
for himself. This distracts the critical faculty of the mind, which is thereby tricked into
a level of acceptance it would not allow if baldly presented with an obviously artificial
plot device as in the first version.
A story depicts the intrusion of a new world that threatens or transforms the old. In Age of
Empires, a small settlement is attacked by raiders and, in defending itself, grows to become
a major civilization. Or take a movie such as Total Recall—the hero starts by thinking he’s
a construction worker, learns he’s a deep-cover spy, gets drawn into a series of adventures,
and ends by asserting a new identity and bringing about the rebirth of the planet Mars.
Foreshadowing occurs in the early stages of the story when the new world has yet to
intrude on the status quo. It is a way of preparing us for the story by giving a foretaste
of what is to come. In Age of Empires, this would be a small raid by a few enemy clubmen who are easily fought off. In Total Recall, the foreshadowing is quite explicit. The
hero dreams of Mars and then is told by the director of Recall Inc., that he will “be a
spy, go to Mars, get the girl, and save the world.” (Pretty cool. Only the most assured
artists have the confidence to explain all their tricks at the outset and still amaze you!)
Use foreshadowing in an introductory sequence to show what is to come. In Grim
Fandango, we first meet Manny Calavera as he’s about to send his latest client, one of
the recently departed, on a long dangerous walk through the Petrified Forest of the
07 3634_CH06
11:13 AM
Page 155
Look and Feel 155
afterlife. This works as a neat bit of indirect exposition, as the author didn’t have to
say, “Look, this is Manny. He sends souls to their final resting place and the worst
route a soul can take is through the Forest.” And it also works as foreshadowing
because, before the story is out, Manny is going to have to take that walk himself.
The novice author thinks that saving the world is the ultimate goal. Hey, they don’t
come any bigger than that, right?
Well, how about rescuing your little niece who’s gone missing in Central Park at sunset? We reckon that more than measures up. The reason is that saving the world is a
sterile goal. Even the worst author recognizes that and, to solve the problem, they usually add, “The world’s in danger and only you can save it.”
What are they trying to do there? They’re trying to make the threat personal. Because,
when it’s personal, the reader or filmgoer or player will care about it.
Look at Star Wars. Nobody tells Luke Skywalker that he’s going to have to save the
world. In fact, he’s going to have to do that and a whole lot more, but what kicks it off
is a plea for help from a beautiful princess. Princess Leia personifies the incomprehensible goals of struggle against a galactic empire. Luke gets drawn in, and we go with him.
The most obvious personal challenge in computer games is the “What in Sam Hill is
going on here?” effect, used particularly well in Half-Life. Outcast uses it even more
cleverly to distract us from the fact that they are handing us a save-the-world story—
paradoxically made more credible by being told promptly after the intro that we actually have two worlds to save. Nonetheless, when the going gets tough, it is the welfare of
the fragrant Dr. Marion Wolfe that really motivates our hero to get going.
Don’t make the mistake of locating your personal challenge to the player within the
backstory. Players don’t want to watch an opening FMV that tells them their family has
been kidnapped and then throws them into the game to sort it out because, in that
case, the player’s identification with the lead character begins only when the FMV
ends. Dark Earth involved a complex plot of betrayal, mystery, and danger, but it made
sure that these plot elements were uncovered during the game. By the time players had
07 3634_CH06
11:13 AM
Page 156
156 Chapter 6
begun to figure out the intrigue, they’d already been playing for an hour, and then, just
to make sure they took the threat to the world seriously, the game embodied it in a
very personal way by having their characters get infected with a slow-acting poison.
Never mind “You have two hours to save the world,” but how about “You have two
hours to stop yourself from becoming a monster?”
The film director Brian De Palma gave a vivid description of resonance. He likened different elements of the story to charged rods. When they get close enough…CRACK!…
you get a spark that illuminates the narrative. Too far apart and the discharge cannot
happen; too close and you get a closed circuit with no sparks.
One of the best uses of resonance in entertainment software is LucasArts’ Grim
Fandango. The 1930s influence of the setting allows a pseudo-Art Deco motif in the
office block where the hero works, which is the headquarters of the Department of
Death. Traditional Art Deco was influenced by Egyptian art, making it an excellent
metaphor for the afterlife, which is the setting of the game. In Grim Fandango, that Art
Deco style comes with a spin, though, in that the decorative images used are not
Egyptian but Aztec—which pulls us back to the origins of the Day of the Dead festival
in pre-Columbian myth. The 1930s trope also works well with the noirish private-eye
slant and the slightly shabby elegance familiar to anyone who has ever visited Mexico
City. The game begins in the city of El Marrow, which looks like a sleepy town that
time has passed by. Indeed, this impression is accentuated by styling the world of the
living, which the hero briefly visits, on Richard Hamilton’s brashly colorful Pop Art
paintings of the 1950s. (Life is to afterlife as the USA is to Mexico? That’s resonance.)
Case Study 6.5 presents the kind of look-and-feel document a designer might circulate
to the art team at the outset of a project. Resonances are already suggested at this stage.
Rather than point them out, we leave it to you to identify them. (But no prizes because
they are pretty blatant!)
07 3634_CH06
11:13 AM
Page 157
Look and Feel 157
Case Study 6.5 An Example of a Look-and-Feel Document
Gaze is an adventure game set in an alternate reality. The world of Gaze is a single
vast city that is technologically advanced (electric cars, computers, surveillance
satellites) but socially conservative.
Our first view is a gray, unchanging surface that is moving like a featureless landscape below us. Then, catching sight of an observation port, we are able to take in
the shape and size of what we’re looking at. We realize the gray material is the skin
of a dirigible, which moves slowly away like a weightless ocean liner to reveal….
A retro-futuristic city. This is the city of the future as imagined in the 1950s: vast
office blocks, streets like canyons, gleaming skyscrapers of concrete and glass catching the sun. It’s bright, clinical, and overwhelming.
Our viewpoint descends through wisps of cloud around the highest buildings.
Recall the architecture of the Third Reich, Fritz Lang’s Metropolis, the Empire State
Building, The Hudsucker Proxy. The quality of the light is hazy; the daylight turned
brass close to street level by the fine dust of those swept-clean city streets. Sleek cars
like huge cryo-capsules whoosh down tarmac avenues on silent tires. Looking along
the block, the avenue goes on and on until lost in the distance, unchanging like a
reflection in a pair of blue mirrors,
The crowds on the streets are uniformly dressed: the men in dark suits, the women
in gray or white dresses. This is not a world like ours with a dozen different fashions
and colors. And that means that the occasional splash of color on a hoarding or in a
window display is all the more striking.
And it’s quiet. The cars are electric and make very little noise. The people hurry to
work without a word. In our opening shot from high above the street, the first
sound you hear is just the lawnmower hum of the dirigible’s rotors.
What feature of all this is startling? We see it as the camera spirals down, taking a
leisurely view of the streets and the people and then turning towards the center of
the city as it reaches ground level. We’re now looking into the burnished bronze
07 3634_CH06
11:13 AM
Page 158
158 Chapter 6
glare of the sun. What we didn’t see before was a massive statue that towers above
the buildings, matching the highest skyscraper. At first it might evoke a resonance
with the Statue of Liberty, but then we see the spiked crown, the balance, and the
blindfold. This is not Liberty. It’s Justice.
Main playing screen
Gaze is an adventure game, and the main screen is a third-person view like in Enter
the Matrix or Max Payne.
Something we must decide: Does the view ever cut, or is it a continuous tracking
shot throughout? Grim Fandango and Dark Earth use the cut and all shots are static,
allowing pre-rendered backdrops. This favors adventure games with strong storylines, because you can use the cut to create suspense: a sudden high angle with the
hero far below, a shot from behind as a door opens, and so on.
In games such as Tomb Raider and Enter the Matrix, the story matters less. Action is
more important, and so a smooth tracking shot is sustained throughout. Where
every action counts, the player doesn’t want to keep switching views.
The graphics engine will determine if the number of characters on screen would be an
issue. It would be nice to be able to at least hint at the heaving mass of humanity filling the streets during the rush hour, so as to make more of the utterly deserted streets
during the rest of the day. Obviously, the first-person viewpoint always has the advantage that it’s one less character on screen. In any case, Gaze is a game of suspense and
nail-biting tension rather than in-your-face bloodbath action, so, in fight sequences,
we’d expect only a few opponents to be on the screen at any one time.
Our thinking on this has been that we’ll probably go for a continuous third-person
tracking shot most of the time, as per Max Payne, with very occasional cuts or prescripted camera movements at key dramatic moments.
Overview screen
Our original impression had been that between encounter areas, we’d switch to a 2D
map of the city on which you’d click to go to a new location.
07 3634_CH06
11:13 AM
Page 159
Look and Feel 159
The problem with that is that it’s not immersive. It would be far better to have a
seamless way of moving between the two views. The ideal would be to pull back
from the hero in the close-up main view, and keep moving up and away until you
had a high-angle shot of the city with the hero now a tiny figure down on the
streets. Not quite as ideal, but almost as good, would be to start pulling out and
then cut to the high-angle shot.
Obviously, we’ll want to keep the screen as uncluttered as possible. We’d prefer to
avoid having a status bar. Instead we’ll show injuries on the hero himself (a torn
shirt, scratches, bruises, and so on) and by the way he’s moving (bad injuries cause
a limp, he hunches down nursing his arm, and so on).
Movement control and combat is like in Tomb Raider (which has the virtue of being
familiar to more players than any other game interface).
Selecting items from your inventory takes you to an extreme close-up of the hero
pulling items from inside his jacket, while the full range of items in the inventory is
shown across the bottom of the screen. (This view will be more immersive than
switching to just a clinical scroll-through item list.) You pick items using either
arrow keys to get him to pull out one item after another or with function keys tagged
to the full inventory of items shown at the bottom of the screen. You can reorganize
items in the inventory so you’ll have at hand those items you’ll need in a hurry.
We need to decide how to handle items that are dropped. We could say you can
deposit items only at a chest, say. Otherwise, it’s possible to get a very cluttered
screen with far too many objects on it. Another way is to have a generic “dropped
object” graphic and you discover what the object(s) is/are only when you pick up
that graphic. Let’s discuss this one.
For character style, think of those chunkily drawn private eyes in big suits that you
get in comic strips from the 1950s. Bob Kane, creator of Batman, seems to have been
the main influence on that style.
07 3634_CH06
11:13 AM
Page 160
160 Chapter 6
What we’re envisaging is that most people in La Vista (the city) seem heavy set and
move a little stiffly. Their chiseled features evoke a robotic impression. Don’t make
too much of all this, though; be subtle. Just because they’re conformists doesn’t
mean they have to stumble around like sleepwalkers!
But our hero, Bracken, is a free spirit, not a cog in the machine, and the graceful way
he moves tells us that. The same is true of Gaze, the mysterious woman for whom
the hero is searching and who seems to hold the key to changing this stagnated
world. When we get onto motion capture, the actors should be thinking of Bracken
as a hawk: proud, swift, capable of both fierce concentration and ferocious activity.
Gaze is a tropical fish: fragile, languid, and ethereally beautiful. The incidental characters share a kind of ‘50s uniformity so, to mark a contrast with that, imagine the
protagonists cast as silent-movie era stars: Rudolph Valentino for Bracken, Louise
Brooks for Gaze.
Resonance can be used with varying degrees of subtlety. The least overt involves
repeated images that defy instant critical analysis and so work on a very deep level.
Examples of this kind of resonance—which we may call motifs—are hats in Miller’s
Crossing, water and the lack of it in Chinatown, and the use of discordant sounds and
background noise in Touch of Evil.
(Why all the examples from cinema and nothing from games? We’re coming to that….)
More obvious are symbols that visually express the subtext. Hitchcock’s thrillers often
symbolize inevitability or danger by shadows that seem to form a web or the bars of a
cage. Orson Welles’s Touch of Evil is set on the border between the USA and Mexico,
and the story concerns a moral border: the gray area between the two lead characters,
played by Welles and Charlton Heston. We categorize these as more overt forms of
resonance because they operate at the mind’s liminal threshold.
Least subtle of all is resonance to another story, movie, or game. Charitable critics refer
to this as intertextuality or hommage (presumably on the theory that saying it in
French makes it more respectable). This kind of reference occurs very, very often in
games, particularly in the introductory FMVs, whose authors seem to want to copy not
07 3634_CH06
11:13 AM
Page 161
Look and Feel 161
only the style of movie directors like Sonnenfeld, Raimi, or Tarantino, but the specific
scenes used in the movies, too. The other name for this kind of “resonance” is plagiarism, and it is a sure symptom of unoriginality. Advertisers plagiarize all the time, but
that’s no excuse for anyone else. We’re game developers, artists in an exciting new
medium; we ought to show the heights of creativity. One sign that entertainment software is maturing as an art form will be when we start to see designers employing more
subtle and original forms of resonance.
Storytellers know that people are perverse. When we’re being told a story, there’s a little bit of us that is kicking against it, continually whispering, “It isn’t true!”
Imagine the start of a movie. Bruce Willis is sitting in a dingy strip-club treating yesterday’s hangover with a large scotch. Two guys in regulation black suits and shades walk
in. “There’s a terrorist who’s sworn to kill the President. We need you back.”
What does Bruce do? Jump to his feet and go straight to work? No. Instead he snarls,
“I’m retired,” knocks back his whisky, and turns to look at the girls. If he’d agreed at
once it would only allow our subconscious imp of the perverse to grumble how unrealistic that was, how he was just doing it for the sake of the story. By showing resistance, however, he makes us start to think, “Come on, man! Agree! There’s no movie if
you don’t.” The storyteller suckers us into rooting for his story even before the main
character does. There’s no better way to enable Coleridge’s “willing suspension of disbelief.”
You can use the same trick even better in games than in movies or books because, in
an adventure game or an RPG, the player’s identification with the hero is much
stronger. In Grim Fandango, Manny learns about an evil conspiracy and is asked to join
the struggle. But he refuses: “I’m not looking to join any military organization. I just
want my job back so I can work off my time and get out of this dump.” The player
knows he’s going to get drawn into the struggle anyway, but Manny’s resistance to the
idea makes it all the more credible.
So, despite what the Borg may say, resistance is not futile. It makes a far stronger story.
07 3634_CH06
11:13 AM
Page 162
162 Chapter 6
Plot Points
Every storyteller knows the importance of confounding expectations. Being told you’re
going with Gandalf on a quest to Mount Doom would be boring if that’s exactly what
happens, no more and no less. But, in The Lord of the Rings, that isn’t all that happens.
Gandalf gets killed by the Balrog when the journey has hardly begun. Nobody was
expecting that—not the characters, nor the reader. This is what is called a plot point.
Plot points pivot the story around in a new and surprising direction, as shown in Case
Study 6.6. This quickens interest in the story and expands its scope.
Case Study 6.6 An Unexpected Development
Dungeon Master (not to be confused with Bullfrog’s Dungeon Keeper) was a seminal
CRPG of the 1980s. The setup was that a powerful sorcerer, the White Lord, sent the
heroes on a quest to retrieve his staff from the underworld and return it to him so
that he could use it to destroy the Dark Lord.
As the quest went on, there were little clues that the White Lord might not be the
happy-clappy sort of fellow that he made himself out to be. And, lo and behold, if
you did take that staff back to him, he suddenly decided that you’d been tainted
with evil and started trying to kill you.
The plot had suddenly been completely changed. Previously the player had had a
nice, clear goal: get the staff and everything will be okay. Only that turned out to be
100 percent wrong. Now you had to figure out a new goal all by yourself.
Fortunately, by this time in the game you would have found that you understood
every spell you could get from combining runes—except for one. And, if you’d figured out the principles behind the magic system, you might start to have an inkling
what that mystery spell was for. So the designer of Dungeon Master not only slipped
in a neat bit of storytelling but succeeded in meshing it perfectly with the gameplay.
Adventure games benefit most from plot points, but you can also use them very
effectively in designing specific levels for racing, platform, action, and strategy games.
Aristotle identified three kinds of plot point: the reversal, the discovery, and the
calamity. There is no reason why a plot point shouldn’t combine these functions.
07 3634_CH06
11:13 AM
Page 163
Look and Feel 163
Aristotle cites the example of the divine messenger who comes to relieve Oedipus’s fear
about his mother but ends up revealing she is his lover—a discovery and a reversal. A
calamity as defined by Aristotle is a destructive or painful act.
In story terms, a plot point might be trying to save the kid but ending up being
responsible for her death (a reversal), finding the Swiss account number in the murder
victim’s cell phone (a discovery), or a bomb going off and killing a major character
(a calamity).
To take an example based on a strategy game, the player might discover that the direct
assault he was planning is impossible because of the cliffs in front of his foe’s stronghold. That’s a reversal. The second comes when he realizes a smaller commando-type
force can be sent in through the mines under the stronghold. That’s a discovery.
Movie executive Fouad Said described these as whammos and believed they should
occur every ten minutes or so. Each plot point, or whammo, turns the story around in
a different direction. When you get a really major whammo…
“Luke, I am your father. “
…it changes everything. These mark out the spaces between acts. In the case of an adventure game like Max Payne, figure that you get a whammo at the start of every level (sometimes a double whammo, as when Max enters the deserted hotel and it immediately bursts
into flame). The big whammos divide groups of levels into acts, and each act can be summarized as a distinct section of the story. In the classic three-act structure beloved of
Hollywood adventure movies, the three acts are set-up, conflict, and resolution.
There is no reason for a game to follow the three-act structure. In fact, in terms of story
length, a game is more like an entire season of a TV show in one box. Consequently you
should have as many plot points as you need to keep the story interesting. If you can
work the plot point so that it occurs during a level rather than requiring a cutscene,
so much the better. It will have more impact on the player that way.
Stories generally work better if early plot points deepen the mystery or throw new
challenges at the player. Later in the story, the function of plot points should be to
clear up mysteries (not necessarily completely), so that the action is freed up to shift to
a quicker pace. When you are sketching out your storyline, imagine an overarching
07 3634_CH06
11:13 AM
Page 164
164 Chapter 6
structure where the really significant plot points become the main supports of the
structure. Then fill in lesser plot points to become the sub-supports, and so on. Hence
in a forty-hour game, the player might (on average) be expected to encounter really big
plot points at ten, twenty, and thirty hours in; fairly major points at two, four, and six
hours, and so on; and minor points at half-hour intervals.
Lastly, remember that structural analysis of storytelling is not a science, and even if it
were, it would provide you with no help at all in constructing a good story. An awareness of plot points and the various theories of where to locate them (from Aristotle to
Joseph Campbell to Syd Field) can at best only give you a steer. The only test that
matters is to “suck it and see.”
We’re going to return to one of our favorite bugbears, the “save game” feature, to show
how it encourages lazy storytelling. Imagine an adventure game with an unbeatable
opponent. The hero won’t have a hope if he fights him one on one. He’ll have to run
away, lead the bad guy back to a trap he passed earlier, and sucker him into it.
How would the designer show that this particular foe was unbeatable? Nine times out
of ten, he wouldn’t even try. He’d just say to himself, “The player will probably try to
fight the bad guy a couple of times. When he realizes that he’s just getting killed straight
away each time, he’ll soon go back to the last save and try to think of something else.”
That’s poor because there are many very effective techniques for building suspense so
that the player begins to expect trouble ahead. Foreshadow by having the corridor littered with old skeletons, the baddie’s earlier victims. Better, make it a hallway filled
with fossilized heroes. Then every step will have the player wondering where the
Gorgon is lurking. Use occasional momentary releases in the build-up of tension so
that you can crank the suspense up a notch. So, in this case, you might have a sound
startle the hero and then a rat scurries out from a pile of rocks. The hero sighs, but the
player now sees a shadow with snakes for hair rising up behind him…
You can also demonstrate the danger to the hero. Show another character, perhaps a
companion of the hero, being ripped apart or turned to stone. And you can tantalize
with false hope. Perhaps the hero shoots, causing a cave-in that seems to bury the
Gorgon, but when the dust clears, she is unscathed. And create a race against time:
07 3634_CH06
11:13 AM
Page 165
Look and Feel 165
Cracks are spreading fast across the roof of the cavern, and the hero has to lure the
Gorgon back to the trap and return before the way ahead is blocked for good.
The reason why suspense should be used more in games is because any death of the
lead character destroys the illusion. Suspense allows you to create fear and expectation
in the player without recourse to the “save game” feature.
One picture is worth a thousand words. Consequently, because games are primarily a
visual medium, there’s no reason to use dialogue to convey something that the images
can do better. For example, a cutscene shows the hero talking to a wild-eyed guy who
is setting the timer on a makeshift contraption. The dialogue doesn’t need to say:
“Once you activate the timer, this bomb will go off in seven minutes.” We already can
see that it’s a bomb, and we know what timers are for. Instead the hero gingerly takes
the bomb and says, “How long?” and the bomb maker replies, “Plenty of time. Smoke
a cigarette, fill out your tax return, and make your weekly phone call to Mom. Just
don’t start War and Peace.”
By the way, that illustrates another objective of good dialogue, which is to make it
serve more than one purpose. The bomb maker is telling the hero that he doesn’t have
much time once the bomb is primed. Every example he gives is a reminder of mortality—cigarettes, taxes, Mom. Then he caps it with ironic emphasis because starting a
war is exactly what the hero is going to do. Hence the dialogue is overdetermined, and
on a subliminal level, it’s bombarding the listener with a whole bunch of associations
and resonances that make sure you have his full attention.
Don’t have characters telling each other what they already know unless you can reveal
something interesting at the same time. The trailer for Deus Ex worked brilliantly
because it smuggled all its exposition in a dialogue that expressed the relationship of
the two would-be tyrants. Exposition was superbly handled in The Getaway, too,
because dialogue not only revealed detail, but it colorfully revealed how that detail was
seen by the characters of London’s criminal subculture.
Here’s one of the supreme examples of effective dialogue:
“Do you expect me to talk?”
“No, Mr. Bond, I expect you to die.”
07 3634_CH06
11:13 AM
Page 166
166 Chapter 6
That’s great because it takes us (and Bond) by surprise and explodes an emotional
bombshell that powers the rest of the scene. Whereas look at this:
“You can’t just let them die! It’s inhuman!”
“Whoever said that I was human?”
That dialogue is boring because it fails to express character, and most of us are more
interested in character than we are in facts. This applies even when the facts in question concern the deaths of thousands or millions of people. Facts like that are just statistics. It takes a story to make sense of them on the human level.
We already saw in Chapter 1 that theme is the inherent question posed by a story. Is
freedom more important than contentment? What makes us who we are? How powerful is love? For the record, the films we’re thinking of now are The Matrix, L’Enfant
Sauvage, and The Last Temptation of Christ, although you could no doubt think of
dozens of other movies with those same themes.
First off, your game shouldn’t answer the question. Even a book or movie should only
pose the question and then explore the specifics of one case. A story is not a manifesto.
In an interactive medium like computer games, forcing an answer to the theme on
your player is doubly bad. The story is created when the game is played. You cannot
know the exact experience the player will get, nor what he or she will conclude about
your chosen theme.
Reflect the theme in incidental details to add texture to the story. Shakespeare wasn’t
afraid to do this overtly in Hamlet, staging a play within the play to highlight the hero’s
moral dilemma. On a minor but very effective level, in Age of Empires it is poignant to
see a lion killing a villager while your armies battle to the death on a grand scale. More
significantly, Blade Runner uses a recurring motif of eyes (the windows to the self) to
subliminally remind us of its theme: “What is the soul of Man?”
The resolution of a story should be:
07 3634_CH06
11:13 AM
Page 167
Look and Feel 167
Hard won—No reward is satisfying if it’s attained too easily. (This is not often a
failing of computer games!)
Not obvious—You don’t want the ending the player has seen coming for the last
10 hours. (All the same, in hindsight it must make perfect sense.)
Satisfying—Usually this means morally satisfying (the hero wins), but the best
authors can make it work as long as it’s aesthetically satisfying (such as tragedy).
Consistent—The ending must be in keeping with what has gone before in style,
theme, and plot development.
Achieve closure—The resolution must do just that: resolve the problem of the
story. (The king is dead; long live the king.)
Case Study 6.7 provides an example of an unsatisfying conclusion.
Case Study 6.7 An Unsatisfying Conclusion
There were many ways that Might & Magic 2 broke new ground, but elegant resolution wasn’t one of them. Through most of the game, the player had taken a party of
adventurers to different cities and ruined castles; explored elemental realms of fire,
water, and so on; and fought a sufficiency of orcs, vampires, and other monsters in
various locations, usually underground.
The game took the form of several mini-quests building to an impressive climax.
The player knew the date that Armageddon was prophesied, so it was soon possible
to feel the icy breath of fate. After many exciting adventures, the player approached
the final labyrinth. There was a long struggle against a succession of powerful monsters that guarded the way to the inner mystery.
And what happened then? Our sorcerers and knights were confronted with a control
panel and were told that the computer that steered the planet was malfunctioning.
To avert collision with an asteroid, you had to solve a coded message. A clock started. You had 15 minutes. And the coded message? It turned out to read: “Fourscore
and seven years ago, our fathers....”
That was it. You solved it, the asteroid missed, the computer thanked you, and you
could go home. Imagine. Why didn’t Tolkien think of that one?
07 3634_CH06
11:13 AM
Page 168
168 Chapter 6
In the film The Warriors (incidentally based on Xenophon’s account of the Greeks in
Persia), the heroes get stranded on the wrong side of town during a gang war. The
story involves their journey back home to Coney Island. They arrive home at daybreak,
and it seems the nightmare is over. But it isn’t easy to wake up. In a mirror of the foreshadowing technique, a threat from the chaos of the story still lingers even when they
think they are safe. Use this in your games. Everything has been building to the fight
with the end-of-level boss, so there are no surprises to be had there. Put in another
challenge after that—something more personal like a test of trust, or a last simple
hand-to-hand fight when all the hero’s weapons are used up. The sense of catharsis for
the player will be far greater.
Sometimes the hero can’t make it back. He can create a new order out of the chaos, but
he has to sacrifice himself to do so. This is what happens to John Wayne’s character
Ethan Edwards in The Searchers: There is no place for him any more in the homestead;
he has to walk away back into myth. Computer games don’t often make use of selfsacrifice, but it is the strongest kind of resolution. Remember that the cost of saving
the world cannot come cheaply.
You probably know the Chinese curse: “May you live in interesting times.” Well, all stories
are set in interesting times. We don’t get to hear about the time Sir Gawain went shopping
for a loaf of bread. Instead, we hear about when he had to marry a hideous crone or when
he had a year and a day before the Green Knight would hack his head off.
The interest value of a story lies in the fact that something has happened to cause an
upheaval in the status quo. The period of the story is a time of change, even outright
chaos, when the normal rules are suspended. Anything might happen. In a lighthearted story, the aim may be only to restore things to the way they were. But, even if
the story ends with the return of the old order, there must necessarily be the change
that T. S. Eliot described: “The end of all our exploring will be to arrive where we
started and know the place for the first time.”
The inner change in the hero is in fact the point of the story. It is what all stories are
about. A protagonist, dormant and waiting (if often unwilling) to change, is catalyzed
by events that challenge him to grow or (in the case of tragedy) to fail in the attempt.
The quest in The Lord of the Rings is not just a long physical trek to Mount Doom
07 3634_CH06
11:13 AM
Page 169
Look and Feel 169
to destroy the One Ring; it is the moral journey of Frodo from innocence to selfknowledge. He starts the story as good because he doesn’t know what evil is—an easy
choice. By the end, he knows that evil is seductive, but he resists it all the same. He
has grown up, and so the story achieves a satisfying closure.
The Sum of the Parts
You can use many techniques for creating atmosphere, engaging the player’s emotions,
and unfolding the action of the game in an effective and exciting way. Game designers
too often have sought inspiration from a narrow range of sources, leading to diminishing returns as the same products are endlessly reiterated in slightly different forms.
Examining other narrative media, particularly cinema, gives us suggestions for how
to enhance our games.
Ultimately, however, entertainment software is a unique medium like no other.
Computers give us the means now to create toys, mysteries, puzzles, and stories, either
individually or all wrapped up in one package. The authors of the past can point the
way, but it is up to us to create a new entertainment medium worthy of our audience.
07 3634_CH06
11:13 AM
Page 170
08 3634_CH07
11:15 AM
Page 171
Chapter 7
• Talking to the experts
Wrapping Up
• Concept and execution
• Coping with change
• Making teams work
y way of summarizing this section, we asked a number of
the game industry’s luminaries to give their own take on
design and development. Before going on to their comments,
however, we should provide a caveat.
There is a true story told by an Englishman who studied
karate. He went to train in Japan. His teachers were at pains
to correct any romanticized Western notions he may have
had concerning the Oriental martial arts. “There is no magic
about the martial arts,” they told him. “Results are obtained
by skill and strength; these will only come through constant
Sometime later, the Englishman was staying at a rural
Japanese inn with several of his teachers and a very elderly
Chinese priest from the Shaolin Temple. The inn comprised
a huge wooden hall with a central pillar supporting the roof.
During the conversation, to demonstrate a point about the
focusing of ki (energy), the Shaolin priest got up and swept
his hand against the stout wooden pillar. He seemed to use
no effort at all, and yet the roof of the inn boomed, the
reverberations lasting several seconds.
After the old priest had gone to bed, the Englishman had a
question for his teachers: “I thought you said there wasn’t
anything magical about the martial arts. How about what we
just saw?”
• Technology pros
and cons
• What makes a
great game?
• The future
08 3634_CH07
11:15 AM
Page 172
172 Chapter 7
The Japanese teachers smiled. “There’s nothing magical about the way ordinary folk do
the martial arts, that’s right,” said the senior sensei. “But that old man is from Shaolin!”
The moral of the story is to remind you that the people we’re going to hear from in
this chapter are some of the major developers working in entertainment software. It’s
instructive to hear the way they went about creating their great successes, but don’t
assume you can just leap in and do it the same way and get a success overnight.
Metaphorically speaking, these guys are from Shaolin….
The Professionals
The “software factory” model advocated in this book does not accommodate extensive
changes in design during development of the product. We assume that you will be
starting with a detailed design and attempting to limit changes to a minimum. This is
simply because it is the most cost-efficient development process. If you can stick to
only 50% of a plan, you’re still a long way ahead of the guy who makes it all up as he
goes along. Moreover, a detailed design is the only way to make sure a large team
remains focused on the same goal.
To allow for the unexpected, we suggest using a developer testbed as part of the design
process. However, we do not advocate turning the testbed into the game itself via an
evolutionary process. Instead, we advise using the testbed to create a game design and
technical architecture for the project and then proceeding by means of iterations (what
is called spiral development) to the finished game. The aim of those iterations in the
software factory model is not to evolve the design, but to focus the build toward the
design. There are other methods, but each has its cost in time, money, and risk.
So it is interesting and instructive to see how some of the classic games of the past
were produced. We spoke to some top industry luminaries, including
Will Wright, creator of The Sims
Ian Bell, co-creator of Elite
Mike Diskett, co-founder of Mucky Foot
08 3634_CH07
11:15 AM
Page 173
Wrapping Up 173
Peter Molyneux, co-creator of Populous, Dungeon Keeper and many others;
co-founder of Lionhead
Glenn Corpes, co-creator of Populous; co-founder of Lost Toys
Brian Reynolds of Firaxis, designer of Alpha Centauri
Chris Crawford, creator of Balance of Power and many other classics of the ‘80s
Dave Perry of Shiny Entertainment, famous especially as the creator of Earthworm
Jim and MDK
Bill Roper, design guru at Blizzard
The Game Concept
In Part I, “Game Design,” we’ve worked from the premise that games begin with a raw
concept. We said in Chapter 3, “Gameplay,” that you should aim to begin with a features-based description of the game. The specific rules may change, but those features
are the game you want to end up with. Consequently, it’s interesting to find out how
some games now acknowledged as classics were expected by their creators to turn out.
To start with, can you tell us the game you are most pleased with? And what made you
think it would be a commercial success?
Peter Molyneux: That’s a tough one. I’m not completely happy with any game. If I had
to pick one it would be Populous, the first game I ever did.
At the time I thought Populous would be viewed as quirky and weird. I didn’t imagine
it would go on to sell 3.5 million copies. The first glimmer I had that this game was
going to be special was when I met a journalist who had played it and was so enthusiastic that I thought he must be mad.
Glenn Corpes: Populous. At first we had no idea it would be so successful, but when
we were able to play the multiplayer game it was very obvious that other people would
like it, provided that Peter could code a convincing computer opponent.
Mike Diskett: Being a commercial success isn’t really something that can be planned
for or designed for. Primarily I design the game that I want to play and want to work
on, and hope that it catches the imagination of the public.
08 3634_CH07
11:15 AM
Page 174
174 Chapter 7
Bill Roper: I am in somewhat of a fortunate position in that I am incredibly proud of
every game Blizzard has published. If I had to choose one that particularly exceeded
the expectations of the development team and ourselves, it would be Starcraft: Brood
War. Although it is an expansion set in name, it has been compared to complete
sequels in regards to the amount of new content and gameplay. Through the course of
the project, the goal of the team became, “We need to outdo Starcraft in every way,”
and I feel that this was, for the great part, accomplished.
Chris Crawford: Siboot (“Trust & Betrayal”). As for commercial success—it wasn’t. But
it was such a GLORIOUS failure!
Dave Perry: Earthworm Jim was probably our most challenging and most fun game to
develop. I had just started Shiny and so was sleeping on the floor to get the game done.
When I was not programming, I was learning to pay payroll or dealing with insurance
and building maintenance. It was nuts! But very rewarding.
We took it to the E3 show in Atlanta, and the response was amazing…just what we
needed to understand that the hard work was worthwhile. It was also funny, and I
think that Jim taught us that games really are just a mix of entertainment—and humor
is one of the strongest kinds.
Will Wright: I’d have to say The Sims. Otherwise I guess I’d pick SimCity 2000. I think
we got the tuning just right on that one, and it paid off in long shelf life.
Ian Bell: Elite. It was an obvious genre that no one else had done.
Planning for Change
We said in Chapter 5, “Game Balance,” that it makes more sense planning to scale up
rather than scale down. However, it seems likely from the fact that development schedules are frequently underestimated that it must be more common to have to drop features for which there isn’t time to code.
With this in mind, we asked next about ideas that had been hoped for at the start but
that didn’t make it into the finished game.
08 3634_CH07
11:15 AM
Page 175
Wrapping Up 175
What was the initial concept like compared to how the game turned out in the end?
Ian Bell: Less well-rounded.
Mike Diskett: Very different. I’m very much a believer in having a very brief initial
design concept, a single paragraph synopsis. Then let the game evolve during the
development. It’s quite inefficient and involves throwing lots of work away, going down
the wrong path and back tracking, but I think it provides the best gameplay.
Will Wright: One of the features that we started with on SimCity 2000 was a dynamic
hydrological model. This allowed the player to create streams that would run down a
watershed along a realistic path and accurately respond to later changes in terrain
(pooling, flooding, and so on). There is something intrinsically interesting about playing with water and the way it tends to flow. We ended up with the first half of this feature (placing streams and watching them run down hills) in our terrain editor, but the
ongoing dynamics proved too costly (in CPU) to include in the final design.
As far as pure simulations go, I’d have to say I’ve always been the most proud of the
model we built for SimEarth. As crude as it was I still have yet to see any other
attempts (on any computer system) to integrate all the aspects of global dynamics that
we were modeling: axial tilt, core cooling, magma confections, continental drift, atmospheric formation, global climate, biome distribution, evolution and diversification of
life, and the effects of global civilization.
Glenn Corpes: With Populous, there was no initial concept. There was look and feel
but no game. The “fractally” generated levels were there from day one, the basic grass
slopes and water were there from day two, the unique ability to model the landscape
with the mouse was there from day three, but the gameplay part of the game was
thrashed out later, partly inspired by this landscape demo and partly by the genius of
Peter Molyneux and the rest of the team talking about it in the pub.
Peter Molyneux: Populous started out as a real-time war game and ended up being
called a “God Game” by the Press. The best games are the ones that evolve around a
coordinated development strategy.
08 3634_CH07
11:15 AM
Page 176
176 Chapter 7
Bill Roper: The first concept for Brood War was to create a “traditional” expansion
set—add a new unit or two, some new multiplayer maps, and a new campaign. Also
we wanted to do it in three to five months after Starcraft shipped. As we got about two
months into the project, it was obvious that we were not going to be happy with just
creating what people had come to expect from an expansion set, and we set our goals
much higher. This meant not only going back to our developer and restructuring our
deal to fall more in line with our new desires, but it also meant adding additional internal resources that we had not originally intended on putting onto the project. We
ended up taking a little over eight months to finish Brood War, but in the end we
released a game that met our goals and we believe benefited and excited the people
who play our games.
Chris Crawford: Radically different. The original concept had the player and a little
creature called a siboot marooned together after a crash. Their only means of communicating was through an “Interspecies Iconic Language” (IIL) that used symbols rather
like an extended version of airport sign symbols. The IIL was about the only part of
the design that made it through to the final product; I called it “eeyal.”
Dave Perry: Earthworm Jim was made as we went along. There was no original design
document, just a bunch of passionate guys that knew what they were doing. That
meant that we could develop the game in the direction of things that were working. I
loved it that way because it meant we could add anything at any time. Nowadays,
teams shudder when they think that might happen. They call it “feature creep.”
Was there anything that you’d hoped to include but that didn’t make it into the game?
Ian Bell: Dramatic reward sequences were omitted owing to lack of memory.
Mike Diskett: Being able to go inside every building, a feature we removed because the
player would be distracted by interiors that had no relevance (but would feel the need
to explore them just in case). Also short cuts through sewers used up too much memory on the PlayStation and Jellymatter, a physics system for objects that deform, squash,
or bounce. This would only work with floating point numbers, which are too slow on
08 3634_CH07
11:15 AM
Page 177
Wrapping Up 177
Glenn Corpes: A lot of the ideas from Populous 2 were ideas that never made it into
Populous—stuff like the extension of the mana system and more landscape modifying
effects such as the whirlpools and the tornadoes. I also wanted to see a higher definition landscape with more varied curves, but this didn’t happen until Populous 3. I also
would like to have seen a more flexible and intelligent automatic town building around
any stretch of flat land. This would have resulted in nicer looking towns but may well
have done nothing for gameplay.
Dave Perry: We wanted to have more and more and more sub-games, but we just ran
out of space in the cartridge. That meant that the “cow sponge bath” level never saw
the light of day.
How did the game change as it progressed?
Ian Bell: It became more general, with the player being able to take more roles.
Mike Diskett: The main character being a cop was just a bit of background info initially, but has evolved into being a major part of the gameplay.
Glenn Corpes: In the early stages, Populous was very hectic with huge waves of people
pouring into many fights. This caused a lot of problems (not least speed and memory)
as it was very difficult to actually finish a two-player game without a couple of hours of
intensive clicking. It had to be changed into something where the player felt able to
gain control of a level.
Peter Molyneux: Dungeon Keeper’s original concept was, “You play the bad guy.” This
was, I thought, one of the best ideas I’d had. In hindsight the way the idea was implemented caused it to be not such a compelling game as it should have been. A lot of this
was the fault of the interface. For many of my games, the area I am least happy with is
the interface. That’s why on Black & White, I had an objective to be innovative with the
interface design.
Bill Roper: Although our original specification (for Brood War) called for two new
units for each species, what those units actually were changed drastically. As opposed
to taking our first idea and just making units that seemed cool, we stopped and looked
at the data we had coming in from our players in regards to balance issues. Although
we strove to make Starcraft an incredibly well-balanced game when it came out, we
08 3634_CH07
11:15 AM
Page 178
178 Chapter 7
were still eager and open to perfecting it if at all possible. We saw Brood War as a
chance to do so by adding units that would balance out not only the early-, mid-, and
end-game, but also that would redesign our air combat paradigm.
Also the scope of what we wanted to accomplish with the single-player campaign shifted towards telling a much more involved story with in-game cut scenes and hundreds
of scripted events. This meant that we would need to do some serious work on the
map editor as well, which was not in our original idea. The team was dedicated to
coming out with an expansion that would not just enhance the original, but would
supplant it in every way. Every decision we made that changed the game—whether in
graphics, mechanics, balance, technology, sound, music, or story—was made with that
goal in mind.
Dave Perry: Every day Earthworm Jim changed in erratic directions but always towards
a common goal. By using a character that had no limits—he could go to any planet,
pick up anything, and so on—it means you have complete freedom. The problem that
many designers face today is that they are making “reality.” Reality is pretty dull.
Chris Crawford: Oy, did it change! This game was a radical departure from anything
that had come before. I found myself reworking the design every time I ran into a
brick wall—which was often. I went from two-person journey to multi-person
interaction because I wanted a more varied set of interactions available to the player.
The basic conflict underwent many changes: First it was man versus nature; then it
was man versus man versus nature; and I finally settled on man versus man.
For that I needed a clean basis of conflict. My theoretical work at the time led me to
believe that circular non-transitive combat relationships offered the most interesting
possibilities, so I developed a conflict system based on such relationships. I’m quite
proud of that system, and I’m sad to note how little the rest of the industry has
picked up on the enormous potential of such relationships. True, they require some
artificiality—I can’t think of many circular non-transitive relationships in the real
world, but artificiality has never held back game designers in the past. I suspect that
most designers just don’t understand the concept.
There were lots more changes, but I’ll stop on this point.
08 3634_CH07
11:15 AM
Page 179
Wrapping Up 179
Brian Reynolds: The important thing is always to get something running that you can
actually play, even if it doesn’t play the whole game. That way you can actually try the
game out and see what the strong points and weak points are. Then you can revise
your prototype to enhance some of the good and eliminate some of the bad, then play,
revise, play, revise, etc. If you don’t play your own game, you won’t be able to make it
fun for anyone else.
What specifically influenced or caused each change?
Ian Bell: Pieces falling into place.
Mike Diskett: Some changes are through necessity—the work involved is just too
much for the benefit to the game. Other times the changes are a result of a very gradual evolution, small steps leading to a large unexpected overall change.
Glenn Corpes: The first effects in Populous were earthquake, volcano, and flood. Early
games consisted of one player building flat land as quickly as possible while saving for
flood while the other bombarded him with earthquakes and volcanoes. Other effects
were inspired by the need to speed up the end game. For example, the knight was
designed to allow the leading player to invest people and mana in a character that
could mop up large areas of enemies with little help from the player.
Brian Reynolds: As I’m playing my game, I jot down the things that irritate me and the
things I like or want to enhance. Then as soon as I’m done playing I go to work on
making the changes.
Dave Perry: We used a system where each person on the team would have to submit
ideas as drawings, even if they had no drawing ability. It made the meetings funny as
we looked at the programmers’ “poohey’” drawings, but also it put us in the right creative frame of mind. We had many, many, many more ideas than we could ever use in
the game.
What would you change now with the benefit of hindsight?
Ian Bell: Some coding stratagems were inefficient, but gameplay-wise it was about
08 3634_CH07
11:15 AM
Page 180
180 Chapter 7
Glenn Corpes: We had to limit the number of characters and buildings in the game
because we just couldn’t move or draw them fast enough. A few years later, Populous 2
was able to draw more sprites and move more characters on the same primitive
machines as Populous, all thanks to us getting better at programming. I’d love to have
seen an unlimited, more “alive” version of Populous 1 rather than the fussier, more confusing and over-featured Populous 2. So with hindsight I’d say that we slightly missed
the point of what made Populous 1 so good when we did the sequel. (Didn’t miss it by
quite as far as those doing the third did though.)
Chris Crawford: I’d give the characters more complex interactions. The current design
doesn’t give them much maneuvering room: They can just talk, deal, and betray. I’d
build in another modality of spoken interaction.
Will Wright: I had a game I was working on before SimCity (and after my first game).
It was called Probot. It was on the Commodore 64 and was about 90% finished when I
put it away and started in earnest on SimCity. I always wondered how it might have
done in the market.
Dave Perry: I would have arranged for a bigger cartridge!
The Technology
Innovative concepts frequently demand innovative technology. However, the commercial reality of development, particularly for small companies, usually prohibits extensive R&D. In a perfect world, you would not commence development on a leadingedge game without at least proof-of-concept for the technology required. In fact the
ideal would be to start out with fully tried-and-tested technology.
Elsewhere in this book, we have given our development doctrine on this point in this
way: A game should not be green-lighted unless the technology required to develop it
is already in existence. (That can include fall-back technology, however.) R&D should
be just that; it should not draw from the development budget.
08 3634_CH07
11:15 AM
Page 181
Wrapping Up 181
How did the technology surprise you? Looking at the positive aspects first of all, how did
it help you in ways you hadn’t expected?
Ian Bell: What most surprised me was how the versions which had to make do with a
character-based display rather than a true bitmap display were in many ways superior
in terms of plotting speed, there being less RAM to clear.
Glenn Corpes: Populous was the game it was because of the landscape modification. (I
would say that it was my idea.) It turned out that this was a very therapeutic, fun way
to spend your time while waiting for more people to breed and the mana to build up.
This was all a lucky byproduct of the limited technology.
Chris Crawford: I was stunned at how fast my inverse parser ran. That’s another niftykeen technology that went unheeded. I had feared that it would gum up but even the
most complex parsings ran faster than the display.
Dave Perry: We pushed the Sega Genesis to its limits. Every trick over years of development were all in that one title.
Brian Reynolds: Even though we’ve come to expect a dramatic increase in processor
speed and equally dramatic increases in memory and hard drive space every couple
years, it still surprises me how much more powerful the machines get during the
development of a single game. For me, more memory and faster processors always
means better AI.
But in all the years I’ve been doing this I think the biggest outright surprise has been
the way in which the coming of Windows has revolutionized the industry in a positive
way. It made game development much easier from a technical point of view, and at the
same time allowed us to reach a lot more potential players. Anyone who remembers
the “config.sys hell” of DOS days has an idea what I’m talking about.
What about the negative aspects of technology? How did it inhibit you? What difficulties
did it present?
Glenn Corpes: I felt very limited by the single slope steepness and having only eight
possible altitude levels of the landscape. At the time they seemed like the biggest problems in the system, but with hindsight these limitations were probably largely
responsible for the success of the game.
08 3634_CH07
11:15 AM
Page 182
182 Chapter 7
Peter Molyneux: Technology is both the boon of development and the bane. It moves
so fast that when you first design a game, you think the design is going to be capable
of X only to find within a year you could have done X times 2. The temptation is then
to redesign around the new technology, and this is why a lot of projects slip.
Ian Bell: The primary difficulties were lack of code space and RAM space.
Brian Reynolds: Increasing technology levels also increases customer demands. As we
progressed from floppy disks to CD-ROMs, and soon most likely from CD-ROMs to
DVD, the amount of “disk space” available for a game increases, but so do the
demands to fill up that space with quality gameplay. It has become more and more
expensive to make a top quality game.
Dave Perry: The game could have used more voice samples. But they take up tons of
cartridge memory, so we were limited to Groovys and Whoa Nellies.
One of the messages of this book is that game development has changed utterly in
moving beyond its origins as a back-bedroom industry. Development teams in those
days might comprise of two people—or fewer. Rather as with other entertainment genres (such as cinema and music) this is rarely now the case.
We have argued herein that formal process is an efficient method for medium to large
teams (10 people or more). The software factory model is the template for one such
formal process. Its key strengths are the ability to plan, predict, and track—vital when
the cost of developing a blockbuster game is getting so high.
How far did you stick with your original schedule?
Peter Molyneux: Anybody who says they can predict the actual release date of an original concept is either a genius or a fool. I am betting they are a fool. Even if you have a
script that details every single point of the game and every hour of someone’s time, it’s
still impossible to predict. The best thing you can do is have objectives to make the
game playable as soon as possible, thereby taking out the “It looks beautiful, but what’s
the game?” syndrome that often takes the most time.
08 3634_CH07
11:15 AM
Page 183
Wrapping Up 183
Glenn Corpes: We didn’t have a schedule. The game took seven months from start to
finish. Bullfrog may not have survived if it had taken over a year.
Chris Crawford: I held pretty close to the schedule. I can’t remember the exact dates,
but I know I wrapped it just in time for Christmas. I now suspect that the game would
have been better if I had slipped the schedule by three months, but I’ve always taken
pride in coming close to my deadlines.
Brian Reynolds: For whatever reason, I’ve tended to have better luck with schedules
than most. My first five products were all delivered within two weeks of the planned
date. But Alpha Centauri ran significantly (several months) late.
Dave Perry: We used to make games on time, but now as we keep trying to push technology, we guess time very badly indeed. I have learned a lot over the last couple of
years, and I have a real respect now for the right balance of technology and innovation
in a new game.
Ian Bell: There was no schedule.
How effective was your progress tracking? Did you know exactly where you were in
your project to the nearest percentile? If not, what would you do differently to get more
accurate tracking next time?
Glenn Corpes: If you’d have asked me about this at the time, I’d have asked you to
repeat it in English.
Chris Crawford: All that went on informally inside my head. This was a wild blue sky
project, requiring not just one but several technologies that had never been invented
before and a fundamental structure that had never been tried before. Nevertheless, I
readjusted the internal schedule with the passage of time to ensure that I’d meet the
final deadline.
Peter Molyneux: The way we track our progress is to have weekly meetings where
everyone states what they’ve done in the last week and their objectives for the next
week, following a development strategy. This has to be centered around at least one
person in the team who has a very clear picture of the final game.
08 3634_CH07
11:15 AM
Page 184
184 Chapter 7
Dave Perry: We tend to get more accurate towards the end of the game. Our games
cruise along, getting enhanced all the way, trying new ideas… Then, when the light
appears at the end of the tunnel, it pops out in no time.
Ian Bell: Progress is made, not tracked.
Would you describe the game development as like the voyage of a ship from port to port,
or perhaps an organic process like the growth of a tree? Or can you discuss some other
analogy that you feel illustrates it better?
Chris Crawford: Well, not speaking for myself, I can note that there are some designs
for which a third analogy works better: The Humongous Heap. The designer just keeps
shoveling features onto the pile until it’s so big that it bulldozes its way through publishers, distributors, and retailers who figure that a pile that big has to have a pony
inside it somewhere.
For me, the process is more like a Napoleonic battle: First comes the grand speech to
the troops, full of phrases like “For glory!” and “Vive La France!” Then comes the brutal slugfest in which I hurl my talents at the problem with utter disregard for anything
other than victory. If I’m lucky, I breach the enemy defenses and plant my flag on the
hilltop. If not, it’s a long and painful retreat. I’ve experienced both.
Glenn Corpes: For many years I was sure it was the tree method. New genres seemed
to be ripe for the picking after a year or so of messing about. These days it’s closer to
the voyage, but not too close. Lost Toys’ philosophy is to carefully pick where we are
going (nowhere too crowded with competition, hopefully) and head rapidly in the general direction with quick, interactive prototypes.
Mike Diskett: Development is like a mountaineer climbing up a cloud-topped mountain without sight of the summit, occasionally slipping down, changing tack round the
mountain, but when the summit is within sight he knows exactly where he is heading.
Bill Roper: We at Blizzard have always viewed game development as very organic, but
much more like some form of controlled chaos. As developers, we have to be willing to
constantly be open to changing the game for the better during the design and development process. We have made changes to our games days before we shipped to ensure
better gameplay, better balance and—most importantly—that the game is fun. In this
08 3634_CH07
11:15 AM
Page 185
Wrapping Up 185
constant evolution of the game, ideas come from all areas and not just some single
designer who sits upon high sending down changes on stone tablets for the masses to
integrate. We do maintain some structure over the process, but we have always been a
very “seat-of-the-pants” company when it comes to the creation of our games.
Dave Perry: It’s like a nightmare. You go to sleep expecting it to be a nice straightforward sleep, but then the nightmare begins. Major stress. When it’s all over, you wake
up and need a lot of caffeine to get your body going again. The difference is that game
development lasts for 18 months!
No, it’s kind of organic really. A game grows until it looks like someone would want it,
and then you package it up and let them pluck it off the shelf.
Brian Reynolds: Much more the latter—new ideas accreting on top of old ideas, and
you can’t tell for sure from the early and middle stages what the end result will be.
Ian Bell: It’s like the evolution of an ecosystem.
What do you think of the application of formal process? For example, component-based
design, planning for reuse, change control, formal code reviews, and so forth?
Ian Bell: Probably a good idea.
Chris Crawford: Absolutely necessary for projects with more than one worker or a
budget in excess of, say, $100,000. We must realize, though, that this stuff does to creativity what a foot of water does to a cross-country hiker.
Glenn Corpes: Treating the programming portion of the game development process
very seriously is a good thing. Standards, modularity and an interest in technology like
source control should be expected of any good programmer. I don’t see any problem
with mixing disciplined programming practice with a less formal game design process.
Mike Diskett: I hate the software engineering creeping into game design. Formal
processes can only work if you know from the start exactly what the game is, what the
final thing will be. But if you want to explore different avenues and keep the freedom
to branch the game down a different route mid-development, then these formal methods just get in the way.
08 3634_CH07
11:15 AM
Page 186
186 Chapter 7
Dave Perry: It will be important in the future, but programmers in small teams have
had so much sloppy freedom, it will be a challenge to lay down “corporate” procedures
that they will graciously accept.
Brian Reynolds: Ugh.
The Team
The first question on this topic concerned the entire team working on the project—not
just the artists and coders, but management and marketing too.
How did you ensure the team meshed?
Ian Bell: By having a very small team.
Chris Crawford: That was the easiest part of the project: There was no team! I did it
all by myself. It’s always best to concentrate on your strengths and avoid your weaknesses. I know some good designers who couldn’t manage their way out of a paper
bag—I’m one of them—but I’m hard put to identify any good designer who’s also a
good manager.
Brian Reynolds: We had the good fortune to found a company with a handpicked team
of veterans who already knew they liked working together. As we’ve grown we’ve done
our best to perpetuate that environment.
Glenn Corpes: That wasn’t my job at the time. The team meshed because it was small
and everyone cared. In my opinion this is the key to game development, which is why
we formed Lost Toys rather than stay at Bullfrog, which has 160+ employees.
Peter Molyneux: If there is one secret to a successful game it is finding a team that gets
on together. If everyone you’re working with believes in the project, and if they know
you and trust you when you say it will be a major success and they will get rewards,
then development of a game can be the most wonderful experience.
Dave Perry: The team got on great. Just the music choices were a problem—we had
music wars. Then, when the insults flew, I once had to make two guys hug in my
office. It’s all in a day’s work, but God how I hate Johnny Cash!
08 3634_CH07
11:15 AM
Page 187
Wrapping Up 187
If a team doesn’t need to change, it’s possible to operate on a gung-ho team ethic. Small teams
frequently operate best this way. However, if new team members need to be integrated, formal
process is one way of ensuring the integration goes smoothly.
Did you need to add any new roles as the development progressed?
Glenn Corpes: This was years back, so 95% of the programming and art was by myself
and Peter Molyneux. There were other programmers, artists, and testers at different
Dave Perry: We had so much animation in the game that we needed to hire in people
who did digital ink and paint. That means that our animators would just draw the pencil lines, then an intern would clean the lines up, then another would color in the
frame. This had to be done to thousands of images—way more work than we ever
Chris Crawford: No.
Ian Bell: No.
Costs and Timelines
Computer games have notoriously been plagued by expanding deadlines and escalating
costs. It is not uncommon to hear of products shipping up to a year or more late, and
there are many instances of games that turned out to be great successes but that had so
tried the patience of the publisher that they had been on the verge of cancellation prior
to release.
“The game is ready when it’s ready,” is a luxury that few developers can any longer
afford. Try sitting in front of your investors and telling them that you want $5 million
but that can’t show them a business plan because game development isn’t a predictable
industry. You will get short shrift.
The methodology proposed in this book is to undertake development in iterations,
or “tiers.” Within each tier, you first plan what you want the tier to achieve. Then
you build it, then test, and finally review how much of the plan was achieved and
08 3634_CH07
11:15 AM
Page 188
188 Chapter 7
what changes need to be accommodated in the next tier. Each tier of the build adds or
refines features within a modular structure that allows elements of the code to be independently checked. The final aim of this process is to achieve “completion”—which we
define not as the game being incapable of improvement, but an ongoing process, at any
stage of which it is possible to draw a line under further development and still have a
shippable product.
What do you consider to be the typical development time frame for a AAA title?
Glenn Corpes: Two or three years, which is one or two too many.
Mike Diskett: Two years for an original title.
Brian Reynolds: These days it seems to be about one-and-a-half to two years for us,
and I think that’s on par with the rest of the industry.
Dave Perry: It is now 18 months—expect that to be 24 soon.
Peter Molyneux: A minimum of two years, more probably three years from concept to
box on shelf.
Bill Roper: As computers, video cards, sound cards and speaker systems get more powerful and less expensive, the expectations increase as to what a top quality game is
going to deliver. Add into this the increase in bandwidth and better and better connectivity to the Internet (at least in the U.S.) and gamers expect a fantastic single player
experience as well as an exciting and smooth multi-player setting in which to game for
months, not weeks. This increase in expected performance and content means that
team sizes have greatly increased and producing a AAA title is becoming as big a risk
as coming out with a top film.
What this all boils down to is that the accepted design cycle for a game of these proportions is about two years. This is very challenging for developers as it is becoming harder
and harder to predict where technology will be that far out, and I think that we are
seeing some of the fallout from the rapid growth of that technology, as games that have
been in development for one to two years are either cancelled or delayed even longer so
that the development team can try and get it caught up with the new standard.
08 3634_CH07
11:15 AM
Page 189
Wrapping Up 189
Faced with the opportunity to question developers of this caliber, we felt it was important to try and find out where that vital spark of creativity comes from. Though the
same caveat applies as before (that is, you may not obtain the same success just by
reproducing the methods), it’s interesting to delve into these special indefinable qualities of gameplay and look and feel—what in movie terms is called the lightning in the
What features do you particularly dislike in games and would not accept in one of your
Dave Perry: I hate games that take a manual read just to play around with. I want to sit
down and play, then when I get stuck, reach for the manual. I guess I am too impatient.
Glenn Corpes: Too much emphasis on storyline or full-motion video (FMV). Games
are supposed to be interactive, so storyline and FMV can only detract from this even if
they are top quality. I’d love to live in a world where everyone felt this way, and we
could just ditch both.
Brian Reynolds: Interfaces that don’t allow full keyboard control of the game. Mouse
interfaces are essential for learning games, but being forced to use the mouse repeatedly after I’ve learned a game just makes my hand hurt.
Will Wright: I really don’t like overuse of FMV. On the other hand I try not to be too
dogmatic about anything. If I have strong preconceived biases about certain features
then I’ll be less likely to approach the design task with an open mind.
Chris Crawford: The deliberate pandering to the least noble elements of the human spirit.
Ian Bell: Gratuitous violence. Poor control design. Juvenile plot lines.
What (if any) were the biggest groundbreakers in the industry?
Glenn Corpes: Tetris was an incredible game. Every programmer in the world was
kicking themselves for not having the idea and coding it themselves in an afternoon.
08 3634_CH07
11:15 AM
Page 190
190 Chapter 7
Also Dungeon Master: the first role player that wasn’t for dice-juggling perverts. And
Ultima Underworld, for doing it again in a “realistic” world. And all id games for staying ahead of the “technology curve” and having the nerve to keep the gameplay at its
most basic.
Will Wright: FS-1 Flight Simulator: This was the first computer game I ever bought. I
was totally amazed that there was a little toy world here that I could fly around in.
This was the first PC flight sim (by Bruce Artwick), it had wireframe graphics and was
quite primitive by today’s standards. But still, it was a self-contained universe with its
own rules of physics that I could play in.
Pinball Construction Set (by Bill Budge) was the first game I played where the object of
the game was to construct something. I really enjoyed the interface (drag and drop)
and the open-ended nature of this. PCS was quite influential in the original design of
M.U.L.E. was the game that proved that economics (and multiplayer gaming) could be
fun. Also it was a brilliant example of how you could design a game that included both
competition and cooperation (in a great balance).
It’s really too bad that most people can’t experience these games any more.
Peter Molyneux: A product in the recent past, which I consider is groundbreaking in
the industry is Quake/Doom. Its simplicity, compulsiveness, and sheer adrenaline rush
is a perfect reflection of what games should be.
Dave Perry: Donkey Kong Country, Command and Conquer, and Street Fighter.
Ian Bell: Colossal Cave, Elite, and Tetris.
What do you think of sequels?
Ian Bell: Easy.
Glenn Corpes: A good thing if there is a good reason to do one, a bad thing in nine
cases out of ten though, including a few I’ve been involved with.
08 3634_CH07
11:15 AM
Page 191
Wrapping Up 191
Mike Diskett: Sequels are great from a development point of view—you know exactly
what the game is and what the work involved is. The public knows exactly what they
are getting, but to go beyond version two of a game without taking a totally new direction means dragging the game out further than it should sensibly be stretched.
Dave Perry: They are great if the first game was awesome. I hate, hate, hate lamer
sequels to lame games.
Chris Crawford: I did one and was none too pleased with it. Sequels should be separated by at least five years. Otherwise, they’re just marketing exploitation of the customer base.
Bill Roper: Sequels can be good and bad. Without sequels, we would never have seen
Aliens—but then again, we never would have seen Highlander 2, either. Basically, a
sequel is going to give you what you put into it. If you just want to ride on the laurels
of a brand that you have developed, you can do that. At Blizzard, we fall in love with
the universes we build, and sequels give us a chance to more fully explore and develop
those worlds. We also strive to do more than just give players more of the same in a
sequel in that we try to build upon the successful parts of the predecessor while finding ways to improve what we or our fans didn’t like.
I think a great example of this was the leaps that occurred between Warcraft: Orcs and
Humans, Warcraft II: Tides of Darkness, and Starcraft (the spiritual sequel). If you look
at each of these games, you can see where basic gaming tenets were preserved while
the interface, gameplay, graphics, connectivity, and so much more evolved. If developers can use the original as a foundation and then build a greater game upon that experience, then they are creating what I envision a sequel to be.
Would you compromise your next design to get a favorable publishing deal up-front?
Chris Crawford: Certainly not. I’ve spent the last eight years as a starving artist; why
should I want to break such a clean streak?
Mike Diskett: If you need money up front from a publisher to create a game then the
design is always a compromise between marketing’s perception of what will sell and
the developer’s individual dream.
08 3634_CH07
11:15 AM
Page 192
192 Chapter 7
Glenn Corpes: It’s not a case of compromise. If you think you have a really cool game
idea but publishers “don’t get it,” there is a strong possibility that you don’t have a
good game design and are just being overly idealistic. Making sure that it’s a game that
a publisher will want to publish (which equates to a game people want to play) is part
of the design process.
Dave Perry: Shiny is lucky—we get to try different ideas. Our games are made for people
like ourselves; you won’t ever see us making Barbie games.
Peter Molyneux: Quite often marketers and publishers have specific objectives for the
products that they sign. I think developers, including Lionhead, should be reasonably
flexible so long as publishers can back up their desires with deal points.
Brian Reynolds: If I know what product I want to do, I’m sometimes willing to agree
in advance on a title and genre, but beyond that we’ve always reserved creative control
of the game and gameplay for ourselves.
Ian Bell: It would depend on personal financial circumstances.
The Future
At the launch party for a non-games company, one of the investors asked us, “Why
don’t you guys in the games industry do more games like Myst? I really liked that one.
But all the others—Doom, Quake, Tomb Raider—just leave me cold.
Of course, that’s only one view. It would be easy to ignore if the person in question hadn’t
been a venture capitalist with tens of millions of dollars at his disposal and if there
weren’t strong commercial evidence that most games still belong to a niche market.
The question boils down to this: “How much interactivity do people want?” It’s unlikely
we’ll see the videogame version of the Victorian family round the piano—put away the
dinner plates, plug in the X-Box, and get gaming. Not with games as they are today.
If people wanted that kind of thing they’d roleplay, they’d join drama groups, they’d play
board games. Every weekend there would be a murder mystery dinner party.
Of course there will always be a solid core of gamers. Many developers disdain the
mass “market.” Let them watch “TV.” But the interactivity made possible by computers
and consoles suggests that there could be products with appeal to both the gamehead
08 3634_CH07
11:15 AM
Page 193
Wrapping Up 193
market and the mass market. These are games that could be played in different ways
according to what the user wants. In short, interactivity should not be about forcing
the player to make choices; it should be about giving the player control of his choices.
We will examine the issue of where games should go from here in Chapter 8, “The
Future of Game Design.” First, let’s see what the games gurus think.
Can games in their current form attract nongamers, or must games adapt to them?
Peter Molyneux: The future of the games industry has not even begun to be tapped.
There have been products like Tomb Raider and sports games that have shown the
potential of designing games for the mass market as well as the die-hard gamer. This
doesn’t mean “dumbing down” games, but making them much more accessible. When
this truly starts to happen, then games will be huge.
Glenn Corpes: I think games are slowly seeping into mainstream culture. It can’t be
stopped; it probably can’t be accelerated. Aiming games at the mainstream is a nonstarter unless you want to work on a sports title.
Mike Diskett: Games are too ugly looking, over-complicated, and require too much
work from the player to be truly mass market. Mass market games either have to be
incredibly simple so as to be immediately comprehended, or complicated with depth
and subtleties beyond the capabilities of current hardware.
Will Wright: I’m thinking that they’re going to meet in the middle somewhere.
Imagine some guy walks into a Walmart and buys Deer Hunter one day because his
friend told him to check it out. He brings it home, tries it out, and enjoys it. It was
simple and easy. Now he’ll probably feel more confident buying a somewhat more
advanced game the next time. Now if he goes back to the store and buys Falcon 4.0
he’ll probably be getting more than he bargained for. But of course there will always be
a market for things like Falcon for the hard-core gamers.
So I would imagine that as the mass market consumers become more computerliterate, our industry will evolve the other way. Instead of mostly hard-core games out
there (with a few Deer Hunters sprinkled in), we’ll end up with a bell-curve that peaks
in the intermediate level games and tapers down at both the beginner and hard-core
ends of the market.
08 3634_CH07
11:15 AM
Page 194
194 Chapter 7
Ian Bell: Games will have to adapt to attract non-gamers.
Brian Reynolds: I think the question is a false dilemma, so I guess my answers are “not
usually” and “not really.” There is a very significant core market of gamers, and one
can make a perfectly decent living writing games for them. It’s not essential to bang
your head repeatedly against the mass-market wall. At Firaxis, we tend to eschew
market research and instead concentrate on writing games we want to play. We assume
that if we write something we like to play there are enough people like us out there
who will want to buy them.
Bill Roper: The answer is a little bit of both. We at Blizzard have always made games
that will be appealing to the core game player without alienating anyone who sits
down and wants to see what this whole “computer gaming thing” is all about. We cannot afford to be strict purists who only make games for the hardcore gamer. There is an
audience for computer based entertainment that is growing at an alarming rate, and
they need content. At the same time, however, we do not need to compromise our
games so that they are unsophisticated and unchallenging.
Making games that are simple to learn but difficult to master is a lofty task that we
have set for ourselves, but if you can get it right, you start seeing not only the life-long
gamer enjoying the experience, but also someone who has never played a game before.
What it all really boils down to is three simple letters—F U N. If the game is accessible
and fun to play, you will find that your audience will continue to expand with the
number of people who have computers who want to have something entertaining to do
with their technological marvel.
Dave Perry: Games need to adapt to non-gamers. It’s like the way games on the PC
used to ask for IRQ numbers and DMA settings. You just have to make it really simple,
and then you give people a chance to feel good.
Chris Crawford: Games will never break out of their rut. Lord knows I tried to help
the industry break out of its rut, but I failed completely, and the industry is so solidly
set in its ways that I wouldn’t even make the attempt now.
A decade from now, there will be interactive entertainment based on storytelling, but I
think it will be independent of games. Think in terms of bookstores versus comic book
08 3634_CH07
11:15 AM
Page 195
Wrapping Up 195
What is your one favorite thing that appeals to you from any one game of your choice?
Glenn Corpes: Rocket jumps in Quake.
Mike Diskett: The fear instilled in the player in Dungeon Master when deep in the dungeon your last torch runs out, and you’re left alone in the dark with just the sounds of
monsters moving around you for company.
Ian Bell: The ability to play with élan. By this I mean a game that may be played with
unnecessary flamboyance—with consummate style as opposed to mere efficiency.
Chuckie Egg was such a game.
Dave Perry: Multiball in Pinball. With about five balls going, you walk away sweating.
08 3634_CH07
11:15 AM
Page 196
09 3634_CH08
11:15 AM
Page 197
Chapter 8
• Benefits of design
The Future of Game Design
• Checklist of good
design elements
• Where games are now
n this chapter, we will review basic gameplay essentials
and then take a speculative look at how game concepts and
design may evolve in the future. First, though, we’ll debunk
some of the myths surrounding formal design and its effect
on the creative process.
The Necessity of Design
Here’s a quotation that you might think has a familiar ring:
“In these companies there is a designer who directs by word
alone and who seldom or never dirties his hands writing
code. Designers say to the other team members, ‘Program
this feature in such-and-such a way,’ and yet they do no real
work themselves.”
Does that sound like something you’ve heard before? Some
dyed-in-the-wool developers still think a formal design isn’t
necessary, and in a moment we’ll review the arguments they
usually give to support that view. But first—a confession.
The quotation above didn’t read quite like that in the
original. Instead of “designers,” it referred to “architects,”
and for “writing code,” substitute “cutting stone.”
The statement is from the thirteenth century, and it bemoans
the passing of the “good old days” (the twelfth century)
when stonemasons could just fling up a new cathedral
without having to bother with architects’ plans. But back
then, as now, there was good reason to abandon the old
• Where games are
going next
09 3634_CH08
11:15 AM
Page 198
198 Chapter 8
trial-and-error process. Building a cathedral was a decades-long project that involved
a team of several hundred men. Also, cathedrals that were built according to intuitive,
unplanned, “see how it goes” methods were proving unreliable; that is, sometimes they
fell over. Unlike software, when cathedrals fell over they killed people.
The quote shows that people have always lamented the need for change, because
change normally goes in the direction of making things more formalized, logical, and
safe. The pioneers don’t care for the sound of that, be they cathedral builders, movie
makers, soldiers, or astronauts. Or even game developers.
Don’t Be Afraid to Plan
When we first began working in the computer games industry back in 1995, there still
was very reactionary opposition to detailed design and the formal development process.
Generally, our experience was that artists were willing to embrace the need for change
but many programmers tended to be quite fearful of it. A couple of objections the programmers would raise were, “We don’t want designs handed down from on high,” and,
“You can design business application, sure—but every game is a brand new project.”
Fear, perhaps, is too strong a word. These were intelligent, talented, experienced
people, and naturally they were wary of changing the way they did things. As Case
Study 8.1 reveals, they resisted the concept of detailed design mainly because game
development had historically been a hit-and-miss affair. The idea of writing the design
in stone and sticking to it throughout an 18-month project seemed absurd. And there’s
a good reason for that: it is absurd. As we discussed in Chapter 4, “Detailed Design,”
the gameplay spec is an evolving document. It is not a Holy Writ; instead it is a combination of a vision statement, a blueprint, and a repository of changes.
Case Study 8.1 Design Saves Time
Back in 1996, David was designing sim-type product (which we’ll call Catastrophe)
in which players had to build and manage a kingdom. Early in the design process,
David began to consider having a bunch of initially neutral city-states that would
declare allegiance according to various factors, including proximity to a player’s
units (especially army units) and how well each player was managing his economy.
09 3634_CH08
11:15 AM
Page 199
The Future of Game Design 199
At that stage, the project had just one programmer assigned to it. David asked him
to knock up a quick testbed. “Just dots on a screen that change color with allegiance will do,” he said. “I want to get a rough idea of how far the allegiance effect
will spread, because that feeds back into the player’s economy. If we’re going to get a
runaway effect, I’d rather start building damping factors into the design now—or
drop the idea altogether if it’s not workable.”
A few days later, the programmer had a display with dots, but they didn’t do anything.
“I haven’t had much time because I’ve been busy with DirectX,” he told David. “But the
game engine will be ready in a few months anyway, so can’t we test your idea then?”
“I was hoping to be past the purely experimental stage by then. I need to get a
guesstimate now to put in the design spec.”
“Oh,” said the programmer. “Well, we hardly ever stick to the spec anyway, so why
Let’s return to those objections to the design process: that it kills the fun of creative
coding and that it is logically impossible to plan a new thing anyway.
When Peter Jackson was filming the Lord of the Rings movies, you can bet he had a
script, storyboard, and shooting schedule. There were more than a thousand people
involved in making those movies. Everybody on the crew had to know what they were
supposed to be filming every hour of every day. Do you suppose that killed the fun?
Do you suppose having a detailed plan stifled people’s creativity? Not a bit. There was
room for improvisation. Having a plan just meant that there would be a safety net if
improvisation failed. There will have been many aspects of the movies that were added
by various people during shooting, and other ways they will have changed in postproduction. The script for a movie isn’t a straitjacket. It is a visionary framework that
actually facilitates creative input.
Remember we saw in Chapter 4 that the existence of a design doesn’t mean that the
designer is necessarily the sole author of the project. Formal design certainly doesn’t
mean that developers will become serfs toiling without any control of their destiny.
Ideas might come from brainstorming sessions in which all can contribute. The design
can and will change during development, as the spec is an evolving document.
09 3634_CH08
11:15 AM
Page 200
200 Chapter 8
And what about the difficulty of writing a design when your game is something completely new? Some would argue this is a logical impossibility, but that is evidently not
so. If it were, even a description of the new concept would be impossible, because just
a five-page treatment is the first step in the design process.
Again, the fallacy derives from a misunderstanding of what the design is supposed to
achieve. We have to stress once more that a design (and especially a design for iterative
development) is never intended to be 100% correct from the start. Instead, initial
design provides you with a first “best guess” that, since it typically takes only 12 manmonths out of a total development of 200 or more, pays for itself if it’s only 12% right.
(Our own estimate is that a good design, even on a highly innovative project, is actually closer to an 80% match with the finished product.)
So, to summarize, why document the design? Because the approach of developers to
date has been like medieval alchemy. “Creativity can’t be planned for!” they gasp. But
in fact it can, and as developers we should strive to be scientific and not let ourselves
be held back by superstition. From the outset of your design, you can apply principles
based on game theory and storytelling techniques. These principles can save thousands
of hours of development time—and that means hundreds of thousands of dollars.
Why Design Is Fine
The advantages of a good design are that it is good for morale, good for the budget,
and good for the game.
Teams thrive on a shared vision and work with more enthusiasm when given clearly
defined goals that reward their efforts.
“When men have a mission, they arrange matters to accomplish it. Without one,
they don’t.”
—Gifts of the Night (DC Comics, 1999) by Paul Chadwick and John Bolton
That is what a detailed design is: a shared vision, accessible to the entire team that
everyone can analyze and comment on. The tier system of design that this book
describes allows the whole team to focus on creating the game in stages. A monthly or
09 3634_CH08
11:15 AM
Page 201
The Future of Game Design 201
bimonthly turnaround means that each tier represents an achievable goal wherein the
contributions of all the team members are visible.
This is very different from the kind of process that programmers legitimately fear,
where they are simply issued with lists of tasks each week. Incredible as it seems, such
a system was in fact applied to all internal developments at at least one major publisher until the late 1990s. This is the exact opposite of detailed design as we have defined
it, as it makes the design opaque and allows none of the coherence and Gedankenexperimentation that a tier-based design and development will encourage.
Robert H. Dunn, in Software Defect Removal (1984), estimates that a design error left
undetected until testing will take on average 10 times longer to fix than one detected
at design time. With today’s much bigger projects, the statistic is probably even worse.
Imagine all those problems showing up a month before your shipping date. Small
wonder that some old-time development managers have gotten the idea that games
can be completed only by going to the mattresses. Many of those problems could be
avoided—and the developers would get better working hours—if things were handled
better at the design stage.
Any detailed design allows scheduling, resource inventorization, control, and tracking.
Without these, it is impossible to accurately budget a project, and, without an accurate
budget, it is unlikely the project will ever be funded—unless you have a rich and very
indulgent uncle.
The Game
An iterative or tier-based design is even more vital for any product that has to incorporate new or untried features. If you absolutely, positively have to ship by a certain date,
the tier system will let you fit the design to that date. If development is slipping, you
may have to trim some features (or even drop them entirely), but the modularity
inherent in the design allows you to do so while still keeping control. Case Study 8.2
stresses the importance of keeping all designs up to date.
09 3634_CH08
11:15 AM
Page 202
202 Chapter 8
Case Study 8.2 Keep the Design up to Date
A few years ago, Rachel was brought in as design consultant on a CRPG that we’ll
call Golem. This project had been in development for over a year with no real sign
of progress, and the feeling was that it was turning into a house of cards, with no
cohesion or robustness in the design. The development director’s hope was that, if
they could tighten up the design, the software issues might also come under better
After chatting to the team and looking at what they had so far, Rachel took the
design spec home to read it. When she’d finished, she had a three-page document of
queries but wasn’t even sure she’d been given the latest version of the spec. “This
describes a third-person game,” she said to the lead programmer, “but what I’ve seen
is first-person.”
“We needed to cut the number of polygons on screen,” he explained. “First-person
is one less character.”
“Fine, but that will change the interface design also.”
He agreed. “Bill, the designer, said we’ll tweak the interface when we’ve got the
other bits working.”
Next, Rachel spoke to Bill. She liked much of his spec and told him so. “There are
some great features. I especially liked that the player will be able to prime spells
ready for casting. That’s a neat tradeoff of versatility against speed.”
“Thanks,” he said. “Mind you, I can’t stand CRPGs. What I really wanted to do was
a first-person shooter.”
“I thought it was tending that way. Also, one of the artists told me there’s a high
likelihood of dropping the thief and warrior characters now, so you can only play as
a sorcerer?”
“Yes, that’s right. The first-person view saves character artwork, but there were too
many other issues thrown up, not least being the complication of designing levels
that would be challenging to all three character classes.”
09 3634_CH08
11:15 AM
Page 203
The Future of Game Design 203
“It’s the old combinatorial explosion problem,” Rachel said. “Three character types
means nine times the work.”
“At least. The new design is a lot simpler.”
“Good.” They were getting somewhere, she thought. “I was sure there must be a
new design. Can you run me off a copy?”
“It’s up here!” Bill laughed, tapping his head. “We’re so far behind schedule, you
didn’t think I’d have time to write up every design change, did you?”
The most frightening thing is that Bill wasn’t kidding. They never did have any further design documentation from that date, apart from technical specs drawn up by
the programmers. Rachel recommended a full retrofit of the game design to date,
with further changes to be placed under change control. The company decided
against these measures on the basis that they would cost too much time at such a
late stage. In fact, it was the failure to implement them that proved costly. About a
year and $750,000 later, Golem was canned.
If you don’t have a design, you may find yourself frantically tweaking and changing
gameplay the day before shipping. Doing so can never be a good idea. If you didn’t get
it right earlier, how could an eleventh-hour change do the trick? The advantage of a
clear design is that, not only does it describe and plan for the gameplay features you
expect, it also gives you a rational framework within which to make any required
Essentials of Game Design
Before moving on to consider where entertainment software is headed, let’s review
some of the essentials of game design. You can rely on a few useful questions as a
litmus test.
09 3634_CH08
11:15 AM
Page 204
204 Chapter 8
Is it Original?
Few games are completely original. As we discussed in Chapter 1, “First Concept,”
originality is a relative term in any case. But if your design is worth developing, it has
to have some features that no other game of its type has tried before.
Of course, these features are the ones that will give you the most trouble in development. Even if the feature is one that’s well understood in another genre, importing it
into your game will have new effects—which is precisely why you should identify and
consider those features right at the start of the design process. As we have said
throughout this section, development will entail changes to the design, but knowing
what your key features are—and anticipating the development difficulties they might
create—is far preferable to blundering about with no plan to guide you.
Original features don’t have to be gameplay features. You might simply have decided to
do a wargame with funny cartoon characters. And why not? After all, look and feel
alone can make a game different. As long as there is something to distinguish your
game, you will know where to focus your main efforts during development, and you
can be sure the finished game will stand apart from others on the market.
Right after the initial treatment stage, compile a list of unique selling points (USPs)
that define how your game concept is unlike any other. These are the things that make
it special and will justify a development team spending a year or two creating it. If the
concept has no original features, junk it. The world just doesn’t need another female
archaeologist with two big guns. Do it differently, or not at all.
Is it Coherent?
Good game designs are built around a core vision that works like a seed crystal. As the
game is built, if changes need to be made, the core vision keeps them targeted on the
final goal. It ensures that the game features serve a common thematic purpose. For
example, if you intend to make a game in which the player makes long-range strategic
plans so as not to have to fiddle with detailed micromanagement, you might think
twice about a multilayered interface that, although original, militates against the core
vision of ease-of-use.
09 3634_CH08
11:15 AM
Page 205
The Future of Game Design 205
A core vision pulls the look and feel of the game together too. Not only do these
“chrome” elements assist each other, but also they should enhance the experience of
playing the game. When all the elements of the game are working to the same end, you
have a product whose internal resonance guarantees aesthetic appeal. On the other
hand, a combination of incoherent game features and artistic styles is simply a mess.
Is it Interactive?
Ezra Pound had a saying that he was fond of quoting to poets: “Whatever can be said
well in prose can be said better in prose.” He was trying to stop poetry from trying to
do something it wasn’t suited for, to get poets to instead concentrate on the unique
strengths of their own medium. In the same way, the entertainment software of the
future will hopefully move away from other media like cinema, defining itself by what
it can do best—in particular, by offering maximal interactivity.
We discussed specific kinds of interactivity in Chapter 3, “Gameplay.” A broader but
still useful distinction is between high interactivity and low interactivity.
Most computer games today use high interactivity, as the player has a strongly proactive role. The story should unfold directly from what the player sees and does, because
the player’s expectations are that his role is proactive, which means he will be impatient if forced to sit back and be told a story. The best designs avoid lengthy dialogues,
chunks of clumsy exposition, and extended full motion videos (FMVs) that remove
control from the player. It is possible to tell a story implicitly and with great economy;
you do not need lots of dialogue. Think of entering a drifting spaceship. One of the
cryo pods is damaged, and a long-dead body is slumped across a table nearby, a bottle
of pills in its hand. You don’t need to spell it out. The story is all there.
Low interactivity, on the other hand, is reactive. At its simplest level, it is represented
by the audience at a play, hissing and calling out boos and hoorays—as long as the
actors take any notice, that is! You use low interactivity when programming a stack of
music CDs to play the tracks you want, or when channel-surfing on the TV.
Low interactivity is a way for entertainment software to reach a larger market, since
more people choose reactive leisure (such as watching TV or a ball game) over proactive leisure (acting or playing a ball game). In the more story-oriented genres that will
09 3634_CH08
11:15 AM
Page 206
206 Chapter 8
evolve out of today’s adventure games, we should see flexible narratives that the viewer
can dip into with as much interactivity as he or she wants.
Is it Interesting?
In Chapter 3, we examined Sid Meier’s description of a game as “a series of interesting
choices.” We also saw that interesting choices are difficult choices.
An elementary grasp of logic tells us that just because all interesting choices are difficult doesn’t mean that all difficult choices are interesting. Being presented with six
treasure chests, five of which are booby-trapped and where there is no clue to guide
you, is a difficult choice but distinctly poor gameplay. (And poor storytelling, too.)
So don’t include features whose only effect will be to annoy the player. The principle
of a good gameplay design feature is that it presents the player with an upside and a
downside, either of which may vary according to other factors. The choice is thus nontrivial to make, but, if player gets it right, he is rewarded by success and a further layer
of choice.
Remember also that the foregoing applies only to gameplay, as we defined it in Chapter
3. Not everything that happens in a game has to be gameplay. However, it should
always be interesting. The stories that grab our attention are the ones where people
change, surprising things happen, and we get to see places in a new way. Interactivity
is interesting in a story if it presents us with real emotional or moral choices. Any
sequence of events—whether interactive or not—will bore the player if it’s obvious.
The mind demands stimulation.
Is it Fun?
You can include only so much in a game, so the trick is to make sure that whatever is
included will enhance the player’s enjoyment.
As an example, consider artificial life, which has hovered around the fringes of the
industry as a buzzword for the last few years. We are now starting to see a clutch of
that make much of their A-life credentials. We’re interested in the technology ourselves, but a pitfall exists for the overeager designer, because sometimes A-life (like
real life) just isn’t any fun.
09 3634_CH08
11:15 AM
Page 207
The Future of Game Design 207
An online RPG world in which all the computer characters are sustaining a real
economy is worthless if the player-characters form a subculture that’s unconnected to
that economy. A strategy game in which you have to keep rounding up your soldiers
because they’ve lost their nerve and deserted isn’t clever—it’s just witless. In an adventure game, having to remember to keep punishing your hero so that he does what you
tell him is…well, actually that could be fun—a true “god game” in the Old Testament
style—because it enhances your relationship with the game’s central character. But
it’s for that reason that it’s worth including, not because it’s a nifty example of new
The lesson is that the technology, like everything else, must serve the needs of the
game and thus the needs of the player. Never let yourself get carried away by a cool
idea, technology, or amazing effects. Always ask first: Is this fun?
The Future of Design
Suppose that, in order to be a game designer, you needed qualifications in business
management, artificial intelligence, creative writing, computer science, and game theory. If that were the case, there wouldn’t be many game designers, and we wouldn’t get
the best concepts possible. We’d just get the best that could be dreamed up by people
with degrees than a thermometer.
Look at it another way. Suppose the only novelists in history had been people who
trained as typesetters as well. Some of them might have written great novels—works of
genius even. But those works would have been fewer, because the talent pool would
have been smaller.
In the 1980s, the programmers of a game were most likely the designers as well. (They
tended to do the artwork, too, which explains some of those early Spectrum titles.)
The classic games of the 1980s were created this way, but no one should lament the
fact that formalized roles and more accessible technology mean that it is now possible
to be a designer without also having to be a programmer. It just makes the talent pool
wider. The creative programmers will continue to design as well, just as some camera
operators also write screenplays.
09 3634_CH08
11:15 AM
Page 208
208 Chapter 8
In the future, we should see more individuals such as screenwriters, artists, novelists,
and board-game designers enter computer game design, which will revolutionize what
we expect from entertainment software. Some of these people will have no idea of the
limitations of the technology. It’s not the job of the designer to worry about that, at
least in the initial design phase. Not knowing what can’t be done, they will demand it;
and their development teams, rising to the challenge, will deliver games astoundingly
different from anything we’ve seen before.
“Most game designers are programmers and very familiar with technology. Programmers
are held back in an artistic, creative way. They know the limits of technology.”
—Roberta Williams, Creator of King’s Quest, quoted in Game Design 101
Making Designs More Generic
If there will be one overriding evolution in game design (as distinct from game content), we would say it will be a move towards starting out with a generic design. This
is consistent with (and indeed a logical consequence of) the software factory model.
Such designs will allow designers to start arguing from the general to the specific,
which will make the design concepts more portable and robust.
Thus, instead of beginning with a design that specified game resources as food, wood,
money, and piety, the designer would now begin by defining the attributes of resources.
He might say that resources can be localized (physically resident somewhere in the
game world) or nonlocalized (held intrinsically by the player and therefore not subject
to capture by the enemy). They can also be collectable (waiting to be picked up) or
creatable (requiring some action to spawn them). They can be perishable (decaying
over time) or indestructible, and so on.
By this model, piety is nonlocalized, creatable, and indestructible, whereas food might
be localized, collectable, and perishable. The advantage comes when the designer
moves on to his next game—another wargame, but now set in the modern world and
involving guerrillas and military juntas. He decides that one of the resources will be
morale, which is spent whenever you want to repair damaged buildings. Broadcasting
stations generate morale; friendly losses or enemy propaganda reduce it. It also
declines over time if you have soldiers your own city streets. Rather than having to
design this resource entirely from scratch, the architecture group already has a template
09 3634_CH08
11:15 AM
Page 209
The Future of Game Design 209
for including it: nonlocalized, creatable, and perishable. In practice, the resource template is rather more complicated than that, and with more attributes, because it aims to
allow completely customizable resources for any game. Now, this can mean many
attribute slots, but they are actually grouped in classes (attributes involving existence,
persistence, and so on). In many games, those attribute slots would never be switched
on or would simply default to some higher level of classification. The advantage is that
you could, in theory, take units out of the terrorist game and drop them into the first
game and they would still be ready to function—which obviously would be handy if
you are aiming for any level of reuse.
Nonsymbolic Design
If you throw a ball and take many high-speed photographs of its flight, you’ll see that
the trajectory the ball took is a parabola. But the ball didn’t follow that path because
gravity told to move in a parabola. A parabola is just a symbolic concept in the analytical domain of mathematics, and the universe doesn’t know anything about mathematics or analysis or symbols. These are human concepts. In reality, there are just a bunch
of physical processes, each of which deals only with the processes and circumstances
just before and just after it. So, the ball is at one position, and gravity tells the ball’s
velocity to change, and the ball’s velocity tells its position to change.
This is the opposite approach to that taken in most software applications. There, processing power is at a premium, so the sooner you can go to symbolic constructs, the
better. The tradeoff is that software can crash when your symbolic “shortcut” misses
something that the one-step-at-a-time approach would have taken in its stride.
Researchers in artificial life have identified an analogous problem:
“The classical AI approach has been criticized because the symbols and symbol structures on which planning and decision making are based are not grounded in the real
world. The problem is that unequivocally decoding sensory data into a symbol and turning a command without error into its intended action may be unsolvable.”
—Luc Steels, “The Artificial Life Roots of Artificial Intelligence” in Artificial Life (MIT Press, 1997)
09 3634_CH08
11:15 AM
Page 210
210 Chapter 8
Here is an example: Suppose you are putting a monster into your new Frankenstein
adventure game, and the idea is that it will jump out of its vat when the player enters
the laboratory. Instead of putting in a lot of complicated AI to do with detecting
humans and having the goal of wanting to kill them, you just choose the shortcut of
placing a trigger tile inside laboratory door. When the player steps on the trigger, the
monster will appear and attack.
Okay so far, but what if the player manages to get onto the tower roof, jumps down,
and, by some fluke, manages to land safely on the balcony of the laboratory? Now he
can explore the lab, get all the power-ups, and read the journal about the monster (an
entry that is supposed to be poignant if he’s just fought and killed it but is meaningless
otherwise). Only when the player goes to leave via the door does the monster climb
out of its vat and growl, “You shall not steal my master’s secrets!”
When cutscenes were pre-rendered, one of the purposes of tying events to a trigger
point like that is that you could be sure of where the player’s character would be standing (and the state of the laboratory, in this example) when the cutscene began.
Nowadays, it is more likely that the cutscene will be generated in the game engine. The
cutscene thus becomes an example of “machinema” which can be different every time,
depending on the game state at the time it is triggered.
The use of the trigger point illustrates symbolic design. The designer assumes there is
only one way for players to enter, and that’s via the door. The alternative nonsymbolic
approach would recognize that the true trigger event is the monster’s awareness of
intruders in the laboratory. Whatever way the player enters the lab—even if by
teleportation—the game still responds appropriately.
Discussing Deux Ex at the GDCE conference in London in 2002, designer Harvey
Smith of Ion Storm cited how nonsymbolic design is changing games. In testing a
maze level (the walls of which were set high enough that the player couldn’t jump
them) the developers discovered an ingenious way to escape the maze. A player could
fix a limpet mine to the wall and use this as a stepping stone to jump out of the maze.
Harvey Smith pointed out that old-style designers might have regarded this as a bug,
but in fact it was an extra opportunity that enriched the gameplay.
“We need to reward the goal,” he concluded, “and not the method the player uses to
achieve the goal.”
09 3634_CH08
11:15 AM
Page 211
The Future of Game Design 211
In the past, the nonsymbolic, step-by-step approach was not practical. The processing
capability wasn’t available to deal with that and graphics too. Hence design used a
symbolic approach, and the idea of one correct solution to every problem became
ingrained. But now much of the graphics work is done by the video card, and computers are doubling in power every 18 months or so. At last, it is starting to be possible to
create “uncrashable” games by avoiding the need to design using symbolic shortcuts.
Case Study 8.3 compares nonsymbolic and symbolic designs.
Case Study 8.3 Comparing Nonsymbolic and Symbolic
In the original Warcraft, peasants collected gold by entering a gold mine and bringing sacks back to your town hall. At the start of the game it was always worth
spawning peasants because the more peasants you had, the greater your revenue
stream. However, there came a point when the peasants started to get in each
others’ way. Adding more peasants would then lead to traffic jams as the peasants
encountered each other on the streets of the town and would have to back up to
let others get past. The situation was alleviated if you planned your town with
wide streets. Additionally, it was not a good idea to place your town hall too close
to the gold mine—giving a little more space also helped avoid traffic congestion.
Now, an economist could derive an equation to describe the flow of gold to the
town hall. The factors would be the number of peasants, the placement density of
the town buildings and the distance from the town hall to the mine. We can imagine that it would be a pretty complex equation. The point is that the designers of
Warcraft never needed any such equation. They simply programmed in the basic
rules and behaviors and the economic simulation emerged directly from those.
Contrast this with a game like Caesar 2, which used underlying equations to create
a simulation of an ancient Roman city. This approach is less satisfying because the
player is not directly viewing the reasons for success and failure. Instead, when
playing a game like Caesar 2 (or any simulation of its type) you are trying to build
an abstract match to the game’s underlying equations in your head. The simulated
economy and the gameplay are less visible, lessening the sense of immersion.
09 3634_CH08
11:15 AM
Page 212
212 Chapter 8
The Future of Games
Computer games have been around for approximately 25 years. In the form we know it
today, the medium is really only some 15 years old. Nevertheless, throughout this section, we’ve drawn on references dating back to King Solomon and Aristotle and
through Shakespeare to Orson Welles. We’ve related computer games not only to the
technological leap of the last decade, but to 4,000 years of human culture.
We’ve done so because we don’t see computer games as froth for kids, as mere “brain
candy.” If we felt that way, we wouldn’t be working as game designers. We believe that
games—entertainment software—are potentially the most exciting development in the
creative arts since man first drew a bison on the wall of a cave and started to tell a story.
But notice we said “potentially.” It’s that potential the industry has hardly begun to tap.
Everything to date has just been the groundwork. It’s in the future that those foundations will grow into an entertainment medium that will rank alongside literature,
music, and cinema. And the future starts now.
Over the last decade, computers have revolutionized the way we look at home entertainment. However, the revolution so far has been only experimental. It has yet to give
birth to a new medium in the way that the fusion of photography and drama created
cinema, for instance.
No one has ever really looked at what the market wants, or even who it really is. And
even when we find out, we often shy away from the lesson. CTI’s Deer Hunter [was] a
slideshow of a game, [but] the U.S. mass market lapped it up, fueling clones from Big
Game Hunter to Turkey Hunt until the barrel scraping began.
“Deer Hunter revealed that people were starving for game styles that the industry hadn’t
even considered making. It also showed that people would buy hunting products. Only
one of these lessons was learned.”
—Owain Bennallack, writing in MCV, 2 July 1999
The designers of the next decade (including many of the readers of this book) will take
entertainment software from the level of a hobby and make it into a new art form. This
is why we have repeatedly said that it doesn’t matter if your product is a game. If you
09 3634_CH08
11:15 AM
Page 213
The Future of Game Design 213
like games, all is well and good. But, just so long as you create something that people
can interact with and enjoy—and that couldn’t be done better in any medium other
than software—you are doing your job as a designer.
Games on a computer are more attractive, exciting, convenient, and immediate than
games played on a board. Multimedia products provide more interesting presentation
and better look-up capability than a reference book. Computer toys let you play with
unlimited resources. All these products are worthwhile, but none of them exploits the
opportunities inherent in software to create a completely new medium. It is as if we
were still at the point where movie theaters showed nothing but short films of onrushing trains and racing meets. What we have to do now is get to the point where a new
technology metamorphoses into a new medium.
And what is the really exciting thing about entertainment software? It isn’t going to be
only one new medium—it’s going to be a whole bunch of them.
The Next Decade
As the entertainment software industry matures, there will be more formalization,
creating a clearer distinction between genres. Why should products as diverse as Ico,
Rez, Animal Crossing, Pikmin, Max Payne and Ghost Master all be classified as “games?”
Hard-core gamers may think there’s considerable overlap in the markets for those
games, but the wider buying public would find them all quite different.
“Another hugely derided game [is] Riven, the sequel to Myst. ‘It’s not a game!’ we cry.
‘Then stop giving us games and give us more of this instead,’ many punters respond.”
—Owain Bennallack; MCV, 2 July 1999
We will see an increasing trend towards more specialized entertainment software magazines and towards defining genres for display in stores. (Just as in bookshops, where
the mass market doesn’t want to search through the equivalent of thrillers, sci-fi, and
bodice-rippers to find what they’re looking for.) You will know the genre you’re interested in, and there will be magazines and web distribution sites devoted to just that genre.
And what will these genres be? We might expect them to be evolutions from the existing game genres, plus a few more that nobody has thought of yet. The demands of a
wider market will mean that, in comparison to today’s games, they will be:
09 3634_CH08
11:15 AM
Page 214
214 Chapter 8
More accessible—People will want products they can play right away.
More flexible—The user will have more choice in how to use the product.
More realistic—Vastly improved AI, physics, and graphics will transform the products of the future.
More fun—No more struggling against the game system; the mass market will not
have the patience for badly designed products.
Completely different—Only the diehards will continue with games as we know
them today.
The Strengths of Software
Let’s look at what entertainment software can do really well. If we were compiling a list
of USPs for the medium, we could say that it has the edge over other media in terms of:
Depth—The background can be far more fleshed out. The inhabitants of the game
world can have their own independent existence.
Freedom—The true payoff of interactivity is that the user can make the product
deliver what he wants.
Persistence—You can get engrossed for hundreds of hours, experiencing the ultimate in escapist entertainment.
Multiplay—Entertainment software empowers groups of people with the ability to
create a mutual narrative.
These are the areas in which we may expect to see real advances, as entertainment software (“playware?”) defines itself as different from other media. So, we expect to see an
increasing trend towards multiplay and for game worlds to feature better AI and
physics that will enhance the verisimilitude of the setting. A much greater range of
interactivity is also required, allowing the player to choose how the game is used.
(Better AI will assist here, too.)
It’s possible even today to wring an additional level of interactivity out of games. That’s
what we’re doing whenever we use cheat codes. Although the cheat codes give some
extra degrees of freedom, the downside is that they tend to be used only by the inner
09 3634_CH08
11:15 AM
Page 215
The Future of Game Design 215
cadre of hard-core gamers (the very people who need them least) and are in any case
unsupported. The use of cheats isn’t catered for by the design. In most cases, using
cheat codes doesn’t enhance the game; it breaks it.
Instead, games should try to empower the user with real choice. Our ideal for a strategy
game, for example, would be one in which you could choose your own role within the
world: commander of an army, ruler of a civilization, or Populous-style god. This way,
the same product could be both a wargame and a sim-civilization game. We’d also like it
if you could decide the kind of opponents you’d face, both by selection of the computer
player AI and by altering constants in the game world, (Increasing the likelihood of
famine occurring could thus switch a computer player from peaceful farmer to desperate raider.) This is a whole other level of interactivity, letting the player choose how to
play, instead of constraining them with the preconceptions of the designer. With the
release of Dungeon Keeper 2, which has several options to let you customize the game
you want, we are finally beginning to see this kind of interactivity being made available.
In the case of a storytelling or adventure game of 10 years hence, you might decide one
evening to watch a software movie starring Victor Virtual. Later, you could decide to
switch characters, or demand longer fight scenes. Or you might interfere more directly,
by feeding clues to the hero. So you can experience The Odyssey, say, from the hero’s
own viewpoint, or as Poseidon (the god he had made an enemy of) or as Athena (the
goddess who occasionally helped him).
It’s probable that the specifics of these new media won’t turn out quite as we’ve
described them, but the principle is clear. Interactivity will be about degrees of choice,
as well as the choices themselves.
The Crossroads of Creativity
To get some idea of the different forms into which computer games will evolve, it’s
helpful to look at the ways that they entertain.
Games as Stories
There is an increasing trend in the games industry to try to find rules for structuring
creativity. This trend is being driven by management who cite the prevalence, in other
creative industries, of theories such as Syd Field’s “paradigm” of movie plots.
09 3634_CH08
11:15 AM
Page 216
216 Chapter 8
It is understandable that videogames executives don’t want to accept that creativity is a
black art. Content is the single biggest determining factor in the success of any entertainment product—way ahead of production efficiency or marketing. If only it were an
exact science! The executives’ desire to believe this is so great that they will selectively
look for evidence of successful formulae. Every so often, for example, a pop band will
synthesize a hit using a formula. But is that evidence of a usable paradigm? We are not
talking about Lennon & McCartney or Björk or Eminem here, after all; we’re talking
about the one-hit wonders you can’t remember a few months later.
A tool that accurately analyzes hit entertainment is not necessarily a tool that will help
you create it. In the search for a science of entertainment, some development companies have hired psychologists to help them inject emotion into their games. In many
cases, these are the same companies that are treating development like an art when it
should be a science, and now they are treating creativity as a science when it should
be an art!
If you want emotion in your game, hire screenwriters, playwrights, novelists, poets,
directors, or actors. These are the people who know how to tell stories.
In Chapter 6, “Look and Feel,” we looked at some perennial storytelling techniques
that can be usefully employed to intensify the gaming experience. However, we have
been saying that entertainment software must become a new art form altogether.
Simply borrowing the techniques of other media such as cinema will not do the trick.
Famously, Walt Disney made it his goal to produce a cartoon that would make people
cry, and he achieved this goal most notably with the death of Bambi’s mother. The
ending of the adventure game Outcast (Appeal) is, if not quite tear-jerking, at least
poignant and emotionally mature. However, it achieves this effect entirely through cinematic techniques. The last 10 minutes of the game include a couple of sequences in
which the player is proactive, but those sequences have no effect on the final outcome.
The story in Outcast, although moving, is told by noninteractive cut scenes.
For entertainment software to make its mark, we need to adopt a new approach to
interactive storytelling. This will come via improved AI and full physics systems that
do not tell a story but rather allow the player to participate in the creation of a story.
09 3634_CH08
11:15 AM
Page 217
The Future of Game Design 217
Any game that can be reduced to a walk-through is simply mechanistic. We want to
create games that can’t be solved that way. The game just has to be “lived,” and the
only way to “solve” the problems it creates is to ask yourself at the end (“death”)
whether the outcome and experience was satisfactory. You can have the concepts of
“right” and “wrong” without necessarily implying correct and incorrect solutions.
Novels, and many other art forms, grew out of the desire to teach morality. They show
the reader the consequences of people’s actions. Until recently, virtue was usually
rewarded and vice mostly punished, except in satires. Frodo suffers, but his determination to do good wins the day—not so different from Bunyan’s Pilgrim. The author
shows us his characters’ behavior, explains their motives, and tells us the consequences.
What we can do through games is to allow the reader to learn the consequences of different courses of action via experimentation and participation. What if Frodo were to
offer the Ring to Saruman to save the Shire from the Nazgul? This is arguably a worthy
decision with noble motives, but what would happen? Why shouldn’t he put the interests of his friends, his family, and his neighbors above those of people he knows nothing about? Given the choice between saving your brother or your best friend, which
would you do? Which should you do? Would it be different if it were a choice between
your mother or your best friend?
There’s no correct solution, just different ones—and you can argue forever about
which is right. That’s how you create something of lasting value and not just a puzzle
or challenge you have to beat.
—Matt Kelland, Creative Consultant at Short Fuze
We have said that entertainment software must move away from old models like films,
novels, and plays to find a new way to tell stories. In fact, face-to-face RPGs could
provide a useful template.
When preparing an RPG session, the referee begins by devising a setting—a town by a
lake, for example, under quarantine because of a plague. He populates this with the
main nonplayer characters (NPCs), whose goals he defines. Thus, the notes might
read: “Lord Shonu: crafty, cautious, extremely wealthy; wants the Key of Time but isn’t
prepared to die for it,” and so on. When the referee has specified all the NPCs, he can
infer the story that would occur if not for the players. The players’ characters (PCs)
09 3634_CH08
11:15 AM
Page 218
218 Chapter 8
perturb the situation with their own actions, aiding some NPCs and opposing others.
Because the referee knows the NPCs’ motivations, goals, and capabilities, he can then
decide how they will respond, and so on.
The point is that the role-playing session itself becomes a process of story creation.
None of the participants decides the plot in advance. Instead, it emerges from the
actions that everybody takes, which means that it would be possible to run the same
scenario with two player groups and get completely different stories. (For instance, in
one game, the players might kill the bad guys and rescue the princess and the king
rewards them. In the other, they act so bad that the evil guys come to work for them,
and they jointly collect the ransom.)
An RPG with a human referee is, of course, capable of a lot more variety than any
computer game. However, you can get stories from the simplest of systems. The rules
of soccer are almost trivial (well, except for the offside rule), and soccer consists of
nothing but those rules and the laws of physics. And yet you get unlimited stories
once that simple game is combined with human social dynamics.
This was the principle behind Conquerors, the virtual gladiatorial sport concept outlined in Chapter 1. The games companies didn’t really understand it as they thought
it was another online game. In fact, probably only 10% of the audience would actively
play the game itself. Even the BBC, whose producers were highly enthusiastic about
the concept, didn’t really get it. This is because the point of Conquerors is to create a
community in which stories are told. A story is a cascade of events and consequences,
and it happens that a gladiatorial bout is one of the simplest forms that a story can
take. If, on the other hand, the concept just becomes an audience watching two people
play a computer game, then it loses the character element that should draw you into
the world of the story.
The point is that stories don’t have to be told. Stories happen. You could tell a story of
how you went shopping for Christmas presents. It isn’t much of a story if somebody
doesn’t know you. But you might well tell it to your friends—and they’ll tell you the
stories of their week. That’s what stories were all about originally. Nowadays we have
formalized art forms and so an author can craft an elegant story that works for a wider
audience. That’s linear storytelling. Alternatively, a game designer can create a rich
environment that allows the players to explore, interact, and create their own stories—
which is what we can and should now be doing with games.
09 3634_CH08
11:15 AM
Page 219
The Future of Game Design 219
Games as Visual Arts
Entertainment software, although rich with unique potential of its own, shares
elements in common with other arts. In particular, we might look at two terms used
in cinema: mise en scene, which is the organization of images in space, and montage,
which is their respective organization in time. Case Study 8.4 provides an example
of mise en scene.
Case Study 8.4 An Example of Mise En Scene
About halfway through Appeal’s Outcast, there is an FMV where the hero, Cutter
Slade, responds to his lady friend’s parting words, “We’ll have words about this
later,” by saying to himself, “I’m all tingly with anticipation.”
Suddenly, he feels he’s being watched. He turns to look towards the impregnable
fortress in the heart of the city. In one continuous shot, the camera pans up over
the wall of the fortress and continues up and up until it finds the villain of the
story standing on his balcony, gazing at the city where he suspects Cutter Slade is
Although FMVs are not really what entertainment software is about, it’s impossible
to deny that this is a remarkable case of mise en scene. To bring us close to both
hero and villain in one uncut sequence is something that even cinema would rarely
attempt. And, if it happened in a movie, it still would not impress us as being as
“real” as in the context of an adventure game where we have at least an illusory
sense that these characters are bustling about their lives at all times.
To illustrate the difference, consider a very famous movie moment: the shower scene
in Psycho. The murder scene takes approximately a minute, and there are at least 60
different shots used. This is montage. By organizing the shots, Hitchcock creates
dislocation and panic. Now suppose he had simply used a single shot for the entire
scene—as he did for the murder in Torn Curtain. Instead of identifying with the victim,
we now become onlookers at the scene of a crime; instead of empathic terror, we feel
objective horror.
Theater uses mise en scene but not montage. Novels use montage as a matter of
course, simply because they are forced to describe a scene one point at a time.
09 3634_CH08
11:15 AM
Page 220
220 Chapter 8
However, they cannot employ mise en scene without evoking montage, because the
order of describing objects and people in a room inevitably becomes significant. You
would need a picture to detach that from montage. As an extreme simplification of a
complex subject, we can say that montage creates narrative and intense emotion, and
mise en scene creates setting, ambience, and reflective emotion.
What about computer games? Except in FMVs (which are little movies anyway), the
computer game evidently uses only mise en scene. This is because montage requires
the viewer to be a spectator and not the controller of the action. Montage would work
only in a weakly interactive game, say a murder mystery game in which your only
choice was which character to follow, In a strongly interactive game, moments of montage can be used very briefly for effect, such as in the marvelous sequence early in the
original Alone in the Dark when a monster runs across the lawn and crashes through a
window. However, these must be used sparingly, or they cease to work and become
only an annoyance. You do not want to keep destroying the feeling of immersion and
forcing your player to sit back and wait while he watches snippets of FMV.
The huge advantage that entertainment software has over other visual arts lies in what
is going on outside the frame. In an adventure game, you could give a message to a
character, and he walks off. He later returns with a reply from his lord. This can happen in a film, of course, but there it is a scripted event. In the case of the adventure
game, it’s possible for the designer to construct a complete environment so that the
character might be waylaid, lose the message, or give it to the wrong person. (Okay,
this wouldn’t have been so easy in the past given it would have eaten up a lot of processing power, but we’re talking about the next 10 years now.) You get the same effect
in a simple way when you see a priest in Age of Empires walk out of sight behind a wall
and you race units to either end, knowing you have him trapped. It works because,
even though the priest has moved “off-screen,” you know he’s still there.
One of the unique strengths of entertainment software is evoking a world that persists
even outside the player’s immediate vicinity. This is a cut above the fictional worlds of
cinema, which are merely credible. Game worlds will become more than credible; they
will, in a sense, become real.
Games as Sports
The ability to accommodate multiple participants makes entertainment software perfectly suited for sports. And these don’t necessarily have to be real-world sports. Sports
09 3634_CH08
11:15 AM
Page 221
The Future of Game Design 221
that would be impractical because of lethal danger levels (combat golf, lion wrestling)
or size and cost (hundred-a-side football) are easy to stage by the medium of software.
Software sports stars may become as well known as real-life baseball and football players, and not everyone need be an active participant. The old game model would insist
on a virtual sport in which the user took part, but why shouldn’t you be a spectator
instead? Perhaps in 10 years’ time we might come home from work, get out the beer
and pretzels, and connect to view the final of the World Quakeball Tournament.
Games as Toys
A toy is something that we play with in order to have fun, which is why entertainment
software and playware are better terms for what we do than computer games. All games
are toys, but not all toys are games. Our aim as designers should be to create something that is interactive and fun, whether or not it is a game.
We are emphatically not predicting the end of gameplay in games. Gameplay is a
perfectly valid way of having fun. Gameplay will continue to be an absolute priority
in the strategy and management genres, for example.
But is gameplay needed in other genres? At the moment, gameplay is present in a
peripheral sense in products like Vice City when it comes to choosing your weapons
and vehicle for a specific mission—do you need resilience, speed, or road handling, for
example? But gameplay here is only part of a much greater whole. It helps to enhance
the product, but it is not central to it. We might suspect that overt gameplay elements
are emphasized only by historical accident in what we now call games. It is because
very early games like The Hobbit found it easier to emulate the form of a game than
a story.
The real innovation came with SimCity. That was the point in the evolution of entertainment software when people could say, “This is very definitely a toy. It appeals to
adults, and it’s popular.” Whether it was a game or not (by formal definition) didn’t
enter the equation. (Simulations tend to contain gameplay simply because real life
does, and those “interesting choices” are one of the things that make a simulation
rewarding. But simulations are not defined by their gameplay content.)
09 3634_CH08
11:15 AM
Page 222
222 Chapter 8
Everybody accepts that software toys (like Pokemon, for instance) are fine for children:
“Encourage exploration. Make your environment free and comfortable enough that
children will be willing to try things out until hitting on what you want them to do.
While they’re figuring it out, they’ll still be having fun.”
—Tzvi Freeman, “Power to the Kids!” in Game Developer, September 1997
If you’ve ever played a game like Dungeon Keeper or Age of Empires with a small child
watching, you’ll know that children really appreciate the fun aspects of the game. “Make
the lion chase that man,” they’ll say, or “Pick up the goblin and make him squirm!”
But it isn’t just children who respond that way. Adults can enjoy entertainment software in the same way. Someone we know who doesn’t play games enjoys Age of
Empires because he likes ordering his little men around to build a city. What annoys
him is when the enemy players storm in and destroy it. For him, that spoils the fun.
Unlike children, we adults like to formalize gameplay just to make what we’re doing
seem respectable. We can’t simply enjoy a perfectly good bit of fun like chucking a
beachball around—we have to go and give it rules. We’re not quite sure whether it’s
okay just to play. But play isn’t the same thing as tomfoolery. Play can educate the
mind, stimulate the imagination, sharpen the wits, gladden the heart, and enrich the
soul. In the future, we’ll see much more awareness of entertainment software as a new
kind of toy, and the products will be all the more diverse and better for it.
Games as Entertainment
The games industry is still, in 2003, the most insular of all the mainstream entertainment media. It doesn’t want to be. It looks jealously at those other media and wonders
what it’s doing wrong. Why isn’t it as cool? The problem is that where the games
industry has come from has saddled it with the wrong mindset and skills to easily
reinvent itself so as to put the emphasis on entertainment.
Here’s an example. Theater and film are different media. You have to write for them in
a very different way. And yet one of the leading Hollywood screenwriters and directors,
David Mamet, is also an accomplished playwright.
09 3634_CH08
11:15 AM
Page 223
The Future of Game Design 223
Compare that to the games industry, where many console developers don’t even want
to work with PC developers, never mind talent from other media. One senior games
executive—a man who founded and ran a very successful development company—had
a meeting with a team from Xbox2 to discuss the future of games. Afterwards he said,
“Why did they keep talking about television? I hate television. It’s boring.”
Television: a medium that reaches an audience easily ten times bigger than movies.
And yet a game executive will dismiss it and its relevance to his industry in three sentences—and be happy to be heard doing so. Talk about having your head in a box!
The industry has to wake up and see that the way ahead isn’t merely licensing the
year’s hit movies. What the movie industry realized about 20 years ago was that the big
bucks don’t come from just having a successful movie. Monsters Inc. cost $110 million
and grossed around $250 million. Given the massive risk in a hit-based industry like
movies (or games), that wouldn’t be much of a profit (if indeed it was a profit at all,
after marketing costs are factored in). Except that the profits from Monsters Inc. were
really closer to $1 billion—that’s because of all the toys, lunch boxes, pillowcases,
books, videogames, sponsorship deals, and so on.
The typical videogame remains niche. The characters and stories would be hard to
pitch as a movie or TV show. One former Hollywood studio boss said to the authors
recently: “We’ll buy game licenses if they’re selling well, but they need a lot of work.
Those guys think eye color and bra size are enough for a character biography! The
development work just hasn’t been done.”
To be fair, he was talking about licenses based on games from several years ago. But,
even though the situation is improving (we are now starting to see some games that
have absolutely first-class, original story ideas, such as Breed), the industry still needs
to bring in a lot more creative talent. Moving your company to Los Angeles or
Charlotte Street, London, is not enough. The culture needs changing too.
The games industry should be developing intellectual properties to a quality that can
potentially work across all media. Games publishers are worried about taking such a
step because they see their market as just the people who buy videogames. But the
market for The Matrix Reloaded or X-2 goes way beyond the people who will actually
go to a cinema to see those movies. Games developers need to see what they do as creating entertainment brands—“root class” IPs—and it just happens that the game is the
09 3634_CH08
11:15 AM
Page 224
224 Chapter 8
first instance of the brand. This step would allow the developers to leverage much
greater revenue from their successful IPs. Until they take it, they’ll always be at the
mercy of the publishers.
The Way Forward
A novelist said, “Games will never be art.”
Well, we say that art isn’t a rarefied concept for the elite few. All forms of entertainment give rise to arts. They may not be formal arts, but they are arts nonetheless.
Computer games may not be there yet, but it’s an inevitable trend.
So far, the technology of entertainment software has overshadowed the art. Now, armed
with ever-improving graphics, physics, and AI, we can look forward to a decade of
exploring the medium’s artistic potential. It will move away from the tropes of cinema
and literature to define itself in ways that we are only starting to anticipate.
The way forward will come from design because designers are the artists of the new
medium, just as medieval architects were the artists of stoneworking. As designers, we
must insist that what has been achieved in computer games is not good enough. We
must devise new concepts to make us worthy of our role at this, the most exciting
moment in the history of human entertainment.
We’d like to end this section with two quotations. The first is from George Bernard
Shaw: “The reasonable man adapts himself to the world; the unreasonable one
persists in trying to adapt the world to himself. Therefore all progress depends on
the unreasonable man.”
The other quotation is from Sir Christopher Cockerell, inventor of the hovercraft:
“If it wasn’t for the silly chaps, we’d still be in the Stone Age.”
Among the readers of this book are, without doubt, those “unreasonable” men and
women who will pioneer the transition of entertainment software from a niche hobby
to a whole series of brave new media.
10 3634_Part 2
11:16 AM
Page 225
Part II
Team Building and Management
Chapter 9
Current Methods of Team Management
Chapter 10
Roles and Divisions
Chapter 11
The Software Factory
Chapter 12
Milestones and Deadlines
Chapter 13
Procedures and “Process”
Chapter 14
Chapter 15
The Future of the Industry
10 3634_Part 2
11:16 AM
Page 226
11 3634_CH09
11:17 AM
Page 227
Chapter 9
Current Methods of Team
• Brief history of game
development methods
• How games are developed today
• Problems with current
• Types of team members
n this chapter, we’ll discuss the methods that are currently
used to manage the development of games. Although the
overall model of game development has changed substantially over the last few years, it still retains elements of the
“spare bedroom” mentality in which it is rooted.
This chapter describes the brief history of the techniques
that have been used in game development. It draws some
parallels with more mature methods that are used outside
the games industry (such as those, for example, that are used
to develop database software). Although this may sound
irrelevant, think of a game as a realtime database with an
exciting interface.
Admittedly this may be an oversimplification, but the fundamental differences between mainstream and game software
development are very small. This rationale will be explained
later in this chapter, and we’ll look at some examples of the
similarities between the two.
• Exceptions to the rule
11 3634_CH09
11:17 AM
Page 228
228 Chapter 9
The Current Development Model
Except in some of the more enlightened game development firms, the general recipe of
software development is as follows:
1. Find about a dozen coders. Try to make sure that there is a good spread of specialization, such as an AI specialist, a graphics programming expert, a sound
expert, a good programmer, and whoever else is available on short notice.
2. Appoint a quirky genius as the lead programmer, usually because he is the sort of
person who can code Quake III in his sleep, and knows everything about anything. (See Chapter 11, “The Software Factory.”)
3. Put them together in a small room with a staff of artists at their disposal.
4. Allow to simmer for an 18-24 month period, stirring periodically and adding soft
drinks and pizza to taste.
5. Be prepared to allow a little extra cooking time, but also be prepared that the
result still may be half baked, or that the developers will have been frazzled to a
Another oversimplification to be sure, but this is basically the method still used in
many parts of the computer games industry. Of course, there are a few more interventions and checks from management, and some sort of plan is occasionally written (at
least at the outset) but, whatever the intentions, in the end it frequently boils down to
the same old “code like hell” methodology.
The Origins of the Industry
The evolution of the computer games industry can be compared to that of cottage
industries. These industries sprang from individual craftsmen but didn’t flourish until
the industrial revolution, when they were mechanized and started producing greater
output at reduced costs, taking advantage of the economies of scale.
Take the pottery industry as an example. It started with individual potters who produced small quantities of pots, plates, and the like, entirely by hand. The more successful of these soon learned that, in order to expand and increase their business, they
11 3634_CH09
11:17 AM
Page 229
Current Methods of Team Management 229
had to take advantage of the scalability of certain processes. This was not done simply
by using more teams of individuals, but was instead achieved by automating and
organizing the less individualistic stages of the pottery-making process, and using a
central pool of designers.
A place still remains in the market for the individual manufacturer with a specialized
product, but the associated costs per unit are much higher than they would be for a
manufacturer who takes advantage of economies of scale allowed by the factory methods. The output of a lone manufacturer cannot simply be a scaled-down version of the
output of an efficiently running factory.
That Was Then…
The computer games industry developed out of garages and spare bedrooms all over
the world, but mainly in the United States and Great Britain. These programmers
were enthusiastic amateurs, writing code for machines such as the ZX Spectrum,
Commodore 64, and Amstrad 464, all of which are ridiculously underpowered compared to today’s computers. Consequently, the scopes of their game projects were much
smaller than those of today.
The limitations were such that one programmer could quite feasibly design and write a
game within a few weeks without having to reuse any code from previous projects.
Because of the tight memory constraints of early computers, it was often better to
rewrite custom code every time to squeeze every last bit of performance and space out
of the system. At most, the “team” would consist of one programmer, an artist (if the
programmer was artistically challenged), and occasionally a freelance computer musician. Teamwork, such as it was, meant simply a group of people who worked on separate parts of the same program and hardly ever needed to touch the work of another
team member. It was possible then to hold an entire game design and architecture in
your head without once having to set pen to paper.
Hardware was so limited that the emphasis had to be on gameplay rather than on presentation. Similarly, the programming emphasis was on small and efficient code. There
was no operating system to interact with, and, due to the universal nature of the hardware, there were no real hardware incompatibilities and abstraction layers to worry
11 3634_CH09
11:17 AM
Page 230
230 Chapter 9
This Is Now!
As computers grew more powerful and the industry prospered, we began to see programming teams that remained small enough to manage their projects fairly well and
that started to make reasonable use of already written code. However, the computer
games industry has always had a “pedal to the metal” attitude working against code
These days, except on the most underpowered platforms, such as the Gameboy
Advance, this attitude is all but extinct. It still exists in certain places, such as the
development of 3D engines, but for the most part, processor power has increased
enough for code “tightness” to take second place to code readability. In fact, today’s
optimizing compilers—to a large degree—do a fairly good job of optimization. Many
of the old tricks are no longer necessary.
Before the advent of the MS-DOS PC and Macintosh, it was nearly always possible
(and sometimes even necessary) to bypass the operating system completely and write
directly to hardware in order to gain the maximum speed possible.
Today, trashing the operating system and hitting the metal is certainly foolhardy, if not
downright stupid. It’s no longer that important to optimize every line of code. The
emphasis now is on writing “clean” code for that operating system. To write such
code, the operating system Application Programming Interface (API) enforces a certain
degree of standardization, and, consequently, with a little extra effort, code can be
made reusable and modular, making it ideal for team use.
As industries go, the games industry is still relatively young and developing. Compared
to other, more mature programming industries, it simply has not had the time to develop the advanced and proven techniques. It is because of the general nature of games
and game development that structured development techniques have become necessary
only in the last few years. Games are now being considered as small- to medium-sized
Hardware is also now much more advanced, with virtually unlimited storage space and
memory. Operating systems have also become much more prominent and need to be
accommodated. Whereas before, the operating system was the first thing to be dumped
when development started, it is now advantageous and even necessary to coexist with
the operating system.
11 3634_CH09
11:17 AM
Page 231
Current Methods of Team Management 231
These technological advances are something of a double-edged sword. The increase in
processing power and memory and storage space means that projects have become, over a
very short period of time, much larger and exponentially more complex to manage. It is
very difficult for developers who are used to the old way of doing things (which was not
that long ago) to adjust to the brave new world of advanced development techniques and
apply these “sensibly boring” techniques in the wacky world of game development.
Due to the increased project size, the number of people needing to be involved in the
development of a game has also increased. To handle the increased need for organization among team members, better-defined management structures and communication
channels are being developed. Teamwork is no longer a buzzword, it’s an absolute
The Trouble with Game Developers
The media has a lot to answer for, because they are responsible for the widespread
acceptance of programmers and other computer-based workers, elevating them from
the ranks of geeks and dorks to being acceptably cool. The media are also partially
responsible for the widespread use of computers in nearly every home, office, and
school. These are all good things.
However, every silver lining has a cloud, and in this case it is that the media have gone
too far!
The games industry is now considered the coolest of the cool. You can now introduce
yourself at social functions by saying that you develop computer games. Ten years ago,
you would have ended up sitting in the corner with all the other dorks, discussing the
merits of gameplay and assembly-language programming. Now even Dilbert, king of
the dorks, is cool.
So, it’s only natural to suppose that the industry attracts more than its fair share of
gargantuan egos. Does this sound familiar to you? How many of your coworkers have
egos the size of a minor planet? How many of them are convinced that they are the best
coders since Ada Lovelace? How many of them are childishly averse to criticism and protect their code and their ideas with a zeal beyond the rational, even when they are wrong
and have been presented with a clearly superior solution (such as yours, of course)?
11 3634_CH09
11:17 AM
Page 232
232 Chapter 9
Game developers also tend to be very individualistic and very bright. But these things
can be a problem as much as an asset. Such qualities can, if not managed, spark a team
dynamic (or more exactly an “antidynamic”) that is disastrous to your project.
Who’s the Leader?
When you get a group of egocentric and intelligent people in the same room together,
there is usually only one predictable result: trouble.
Competition exists elsewhere, but it need not be destructive. There is a cooperative
form whereby developers are mostly keen to learn from others and are open to criticism and new ideas. This results in the sharing of skills and ideas, raising the average
level of skill in the organization.
Microsoft has a name for this: Coopertition, or having the competitive instinct to gain an edge on
your colleagues, but always as part of a team.
So go ahead, gain that edge, but do so in a way that is beneficial to the team. No lonestar mavericks need apply.
Trying to impose more mature development techniques and standards on the games
industry is sometimes viewed as being stuffy and boring. We have found that it usually
meets some level of resistance at first. Many people who choose to become game developers did so because they perceived the industry as offering freedom and coolness factors. They may believe (mistakenly) that you cannot organize and manage a creative
process without losing the vital spark.
Therefore, game developers often try to resist the imposition of rigidity in technique.
Flying by the seat of the pants on the cutting edge of technology is far cooler (even
though it is usually as painful as it sounds). You don’t get this sort of problem with
spreadsheet programmers! Of course, not all rigidity is good, but in certain areas it is
absolutely essential to have rigidly defined protocols, or the associated risks to development will cause the eventual cancellation of the project, which is not very cool at
all. Think of the procedures and practices as padding for the seat of the pants. It makes
the ride a bit more comfortable in the long term, even though the extra bulk is a bit
more restrictive at first.
11 3634_CH09
11:17 AM
Page 233
Current Methods of Team Management 233
Let’s look at the classic stereotype. Mr. Hotshot Game Developer is an all-around
creative genius who has superior optimization skills and the ability to write fast, tight
code, and all before breakfast (which is usually taken at midday while hacking out the
greatest code in existence). By contrast, people who program databases are just boring
gray suits with no life. Strangely enough, we have found that usually the converse is
true. Database developers know that what they are doing is a job that has to be finished within a certain time limit. It is no less creative or difficult than writing code for
a game, and it usually is equally on the cutting edge of technology (just maybe not so
glamorous of an edge). More to the point, they have far more of a life than the game
developers because they know that what they are performing is just a job with standard hours and not an obsession that takes up all their time. Obsession doesn’t usually
make for better performance. A broader mind is often a more flexible and creative one.
So It’s War Then?
The clichéd cutting edge of technology is a fun place to be, and most game development teams feel as if they are right there at the sharp end, pushing the envelope, breaking the limits, and all the other macho stuff. It’s all true, of course. They usually are
writing for the latest technologies (although often through an insulating API layer, so
it’s not exactly uncharted territory).
Unfortunately, this fact has shifted the focus to the technology rather than the gameplay. This is the biggest problem facing the game industry today. There are dozens of
case studies to demonstrate this point, but there is no need to. Just take a stroll down
to your local software emporium and check out the games on the shelves.
As technology improves, game production mutates into a technological arms race: bigger, better, faster, more. If this trend continues, then games are going to become less
and less fun and more and more like very impressive technology demonstrations. (This
comment, originally written in 1998, seems born out by recent 2003 releases such as
Enter the Matrix and The Hulk.) If someone wants to see pretty moving pictures, they
go to a movie theater. If someone wants an entertaining interactive experience, then
they play a game. Quake suffered from this, being effectively a technology demo with
an afterthought of a game tacked on. Fortunately, Quake II rectified in great style any
deficiencies in the first iteration.
11 3634_CH09
11:17 AM
Page 234
234 Chapter 9
However, this “more technology, less gameplay” is not a trend that we see continuing.
Chances are there will be a backlash in the same way that there was a backlash against
the crop of interactive movies that seemed to pollute the software shelves at the start of
the 1990s.
It does not mean that games will become technologically challenged. The emphasis will
be focused more on the gameplay and stretching the technology will take second place.
(At least, this is what we as gamers would like to have happen.) It is interesting to note
that the biggest selling game of 1998 was Blizzard Software’s StarCraft, a technologically
competent game, but with outstanding and well-balanced gameplay and story lines.
Contrast this with two of the biggest selling games since then: The Sims and Grand Theft
Auto III. The first of these relied on originality over technology. The second provided the
holy grail—technological superiority combined with great gameplay.
The aim of this book is to narrow the gap between the games industry and the more
mature and reliable development techniques that are used everywhere else.
The Problem Developer
Clichéd though it may be, it is useful to describe a few of the classic problem types
that can mess up a team.
The problem types of game developer identified are the maverick, the prima donna, the
shy guy, the sleeper, and the jack. These guys are initially welcomed into the team, but
are capable of causing problems further down the line when their shortcomings
become apparent. It is not usually an issue with technical ability, because those with
insufficient technical skills are usually weeded out early. The problems that these guys
cause are usually more subtle and difficult to detect, which is also part of the problem.
We’re having some fun here. Not many people conform 100% to these stereotypes.
However, we can all run the risk of behaving like one of these types at times, and it
does seem that the games industry tends to attract these sorts of individuals.
Now on to the problem types.
11 3634_CH09
11:17 AM
Page 235
Current Methods of Team Management 235
The Maverick
The maverick is a talented individual whom everybody trusts and depends on. If you
have a tricky job, then the maverick is your man as shown in Figure 9.1.
Figure 9.1 The maverick.
Combining a high level of technical skill with a broad base of knowledge, the maverick
enjoys his position as top dog—until something goes wrong. The maverick takes complete ownership of his code and tries not to let anyone else near it. In the maverick’s
view, no one else can be trusted to touch his code. No one else would have the skill or
11 3634_CH09
11:17 AM
Page 236
236 Chapter 9
the knowledge to do anything other than sully its purity, so it’s best not to let them
touch it (or so the maverick thinks). This behavior is more appropriate for the lead
singer in a rock band than a coder in a software development team. Nonetheless it is
often tolerated, if only because the maverick is so competent and confident. Quite
often, the code is virtually unreadable by anyone except the maverick. The technical
description for this hoarding phenomenon is “job security.”
However, when something goes wrong, what happens then? Because no one else knows
the code, the rest of the team have to sit and sweat it out until the maverick can resolve
his difficulties. This means that everyone is now waiting on one person. This is not a
good situation to be in. If the maverick fails, he takes the whole team down with him.
And what happens if the maverick quits, leaving a vast quantity of source code that is
difficult to even understand, let alone maintain? The only option here would be to try
to assign other team members to reverse engineer the maverick’s code to try to understand it. There are other options, but they are so unpalatable as to be inconsequential
(such as rewriting the code from scratch or dropping an entire area of functionality).
Unfortunately, this may not always be possible because mavericks tend to be given
areas of critical responsibility. If it is not possible or desirable to drop the functionality,
then the only option is to stick it out and accept the delays caused by the maverick’s
departure. (Case Study 11.1 is an example of this choice in effect.)
The Prima Donna
He’s the bee’s knees. This guy knows that he is the best, and anybody who tries to criticize him will suffer his wrath. Everybody must know a prima donna. This is the guy
with the huge but fragile ego. His idea is always the best. His code is always perfect. (A
stereotypical image of the prima donna is shown in Figure 9.2.) Like the maverick, the
prima donna also does not know how to take criticism.
The prima donna is usually very intelligent and technically skillful. However, he lacks
interpersonal skills and has the knack of rubbing people the wrong way. He divides
other people into two classes: those who are less intelligent than he is and threats. He is
typically the leader on projects, because his many talents make him appear invaluable.
Unfortunately, it is from this position that the most damage to the team structure can
11 3634_CH09
11:17 AM
Page 237
Current Methods of Team Management 237
Figure 9.2 The prima donna.
The prima donna will attack anyone whom he perceives as a threat by insulting their intelligence and their skills. Anybody else who simply is not up to his technical level will be
made to feel completely incompetent. Not everyone (and particularly not introverted programmers) can to stand up to the attacks, which leads an uncomfortable tension within the
team. A prima donna is the biggest danger to a team. When a team breaks down as a result
of the prima donna’s antics, you can expect to lose two or three of your best members.
The prima donna is also basically a control freak who doubts everybody else’s ability
and looks with scorn at what they do.
11 3634_CH09
11:17 AM
Page 238
238 Chapter 9
The Shy Guy
The shy guy is the stereotypical computer geek as shown in Figure 9.3. You saw a lot
of these guys in computer films from the eighties. (You know the sort: personalitydeficient schoolboy/computer geek that saves the world and embarrasses the school
football hero.)
The shy guy tends to be very withdrawn, and communication problems of one kind or
the other are the order of the day. Taking criticism is also difficult for the shy guy. In
fact, taking any kind of advice is difficult, as is the basic communication of ideas.
These people seem to be comfortable only with their computers, so maybe the best
way of communicating with them would be to send email.
The danger that the shy guy poses to the team is fairly obvious, although there is no
direct threat and the shy guy doesn’t actively destroy the team. It is just that they never
really participate. The biggest problem with their lack of communication is that it drastically decreases the level of project visibility, and hence becomes a great risk to the
project at large.
Figure 9.3 The shy guy.
The Sleeper
The sleeper appears just like a normal (or even excellent) member of the team, at least
on the surface. Under the surface, he is not so useful, sowing dissent with the management among other team members as shown in Figure 9.4.
11 3634_CH09
11:17 AM
Page 239
Current Methods of Team Management 239
Figure 9.4 The sleeper.
There can be many reasons for this. One of the most common reasons is pure anarchy.
The sleeper simply has a problem with accepting authority and so subconsciously seeks
to undermine it in any way possible. Other reasons include attempts to poach team
members in order to form independent teams, or feeling as if they have been treated
badly and wanting to make sure that other team members feel the same way too.
The sleeper is a dangerous team member because he is also very difficult to detect.
The sleeper will not make himself known to management, and will often reveal himself
only to a trusted developer who he expects will keep quiet about his level of dissatisfaction. Anonymous reporting channels provide a possible method for detecting this
type of problem developer.
11 3634_CH09
11:17 AM
Page 240
240 Chapter 9
The Jack
The jack is the jack-of-all-trades but the master of none, as shown in Figure 9.5.
These are the most useful of the “negative” developer types and, in some cases, can be
viewed as positive. Jacks are the only type of problem developer that are worth preserving; as long as they can be taught to overcome their shortcomings, they will then
become a useful team member.
The jack’s main shortcomings are that they can be very sure of their own skills, which
sometimes develops into overconfidence.
The gap between their skills and their opinion of their skills is usually discovered
only when it is too late, after which their credibility is damaged beyond repair. This
is because they have managed to sell themselves (which they are very good at) into a
position that requires more skill than they actually have. Being unable to admit their
failure, they bluff and produce substandard code and technical excuses that tend to be
believed by the less technically experienced staff.
Figure 9.5 The jack.
11 3634_CH09
11:17 AM
Page 241
Current Methods of Team Management 241
Excessive Long Hours Mean an Unsuccessful Project
If you are expected to work long hours on a project at the expense of a social life, this
is an indicator that something is seriously wrong, and it’s usually for one of two main
reasons (or a combination of both):
1. The project has been badly scheduled. Too much work is being attempted in too
little time.
2. The project wasn’t badly scheduled to start with, but major developmental
problems were not detected and corrected in time, and now the project is attempting to play catch-up.
The first reason is inexcusable—there is no good reason why the development team
should have bought into a schedule that was clearly unrealistic. In real life, however,
things are not always as they should be. The usual reasons for impossible schedules are
pressures from the publisher or, believe it or not, the season.
Let’s say the publisher needs the release by a certain date (for any number of reasons: a
soccer title needs to released to tie in with the championships, a movie tie-in needs to
be released to benefit from the publicity, or the title is part of a lineup required for the
release of a new platform). Sometimes there is no good reason for a deadline other than
contractual obligations. One company signed themselves into a contract that obligated
them to produce an innovative AAA-quality product within nine months—and from a
standing start. Fortunately, this was quickly realized to be impossible, and involved
much hasty renegotiation of contracts and a change of publishers. In the end, the
product was late by more than a year (owing to the awful code that had been initially
hacked out in the effort to meet the nine-month deadline). This is obviously not a good
situation to be in.
Seasonal pressure usually boils down to one thing: Christmas. Everybody wants their game
to cash in from loving grannies buying little Johnny a game for his newfangled computer.
This also explains the glut of titles in the months before Christmas (and the late arrivals
following shortly thereafter) with nothing major appearing during the rest of the year.
In late 1998, a group of publishers got together to see if they could agree to a smoother
distribution of releases throughout the year in an attempt to avoid the annual silly
season. This seems to have worked fairly well, although we’re not sure whether it is by
11 3634_CH09
11:17 AM
Page 242
242 Chapter 9
accident or design. Either way, the annual nightmare of the Christmas release has been
alleviated to some degree (which is a good thing because nobody can delay Christmas
to get a release finished—although we’re sure some publishers have considered trying).
Exceptions to the Rule
Every rule, though, has some exceptions. In Case Study 9.1, we look at games studios
that have development methods that work. In several of the examples, the releases were
delayed. However, the delays were what we can call “positive” delays. They were anticipated. They were not caused through incompetence or coding problems or runaway bug
counts. They were caused by the developers refining the gameplay, or fine-tuning the
artificial intelligence, and (in one case) by replacing the entire graphics engine.
Some development companies are lucky. They are in the enviable position of being able
to define their own deadlines. They are in the position where they can say, “We are not
ready to release this yet. When we are satisfied, we will.”
These companies have won this freedom because they are the leaders in their fields,
the companies that produced the products that everybody else imitates. And this is
why they are the exceptions.
The games presented in Case Study 9.1 were either produced on schedule or were produced using good production standards with sensible change-control procedures to
ensure that the project was always under control. Why is it that these games are the
exception rather than the rule? All three examples were successful games in their own
right (particularly Quake and StarCraft). Is it because a race of superhumans have sent
forth their progeny to lead the way in developing games on earth using advanced
development techniques beyond the puny understanding of mortal man?
No. It is because they have a solid grasp of development techniques and the need for
good teamwork. Adequate development and change-control procedures were used,
which, in these cases, paid off handsomely.
If these techniques can pay off for these development teams, then there is no reason
that they can’t for you and your team on your next project.
11 3634_CH09
11:17 AM
Page 243
Current Methods of Team Management 243
Case Study 9.1 Quake, StarCraft, and XCOM: Interceptor
These three games are just a few prominent examples of many games that have had
successful developments.
Quake (id Software) The development of Quake is a perfect example of how every
company would like their development teams to be: no deadlines except those that
they imposed on themselves and publishers chasing after them with the scent of a
guaranteed hit in their noses. The team had time to research and perfect the technology. Also, due to the generous amount of funding (available from both the success of Doom and the distribution rights of Quake), the team was able to employ as
much time and as many resources as they needed in order to perfect the product.
The team structure placed Jon Romero in charge of the overall architecture and
design. He personally resolved any arguments, although there were few of these
due to the strictly delineated code ownership. This kind of arrangement wouldn’t
work, however, for just any team. id Software is truly an exception, and there have
been many failed attempts to emulate their style.
StarCraft (Blizzard Software) StarCraft was developed originally using a similar
graphics engine to that used in Warcraft II. During development, another division
of the company, Blizzard North, developed and released Diablo, which used a
graphics engine far more advanced than that used by StarCraft. The decision was
taken by those at Blizzard who were working on StarCraft to leverage the experience gained by developing the new graphics engine, so they could improve the
appearance of StarCraft. This decision was not taken lightly and added a significant
delay to the development of the product. Part of this delay was also due to
Blizzard’s tweaking the advanced game design and balancing the gameplay. The end
result of this calculated risk (as already stated earlier in the chapter) was the
biggest-selling PC game of 1998.
11 3634_CH09
11:17 AM
Page 244
244 Chapter 9
XCOM: Interceptor (Microprose Software) XCOM: Interceptor was a departure in style for
the XCOM series. The whole series revolved around the exploits of various aliens
trying to take whatever it is that aliens want from humans. The basic game structure
involved high-level strategy, researching new weapons and technologies to defeat
the alien threat. This was interspersed with varying missions of a more immediate
nature. This is how Interceptor differed from the previous iterations in the series.
(Interceptor replaced squad-based combat with space-based fighter combat. This
game is noticeable because it was produced to a high standard and to a tight schedule. (A boast to this effect even appeared in the end game credits.) The developers
were rightly proud of themselves for producing the game on time and on budget:
proof of the concept that it is possible with a dedicated and focused team.
12 3634_CH10
11:19 AM
Page 245
Chapter 10
• Assigning personnel
Roles and Divisions
• Improving morale and
the working environment
• Spreading the risk
ne of the central themes behind this book is that the
growth of the games industry has reduced the effectiveness of small development teams.
It is no longer possible for one person to design, write, and
produce all of the code, artwork, music, and other miscellaneous parts that a game requires. This means that a natural
process of specialization has taken place, with certain welldefined roles emerging from the previous uniformity.
There are exceptions to this rule, but they are mainly
constrained to niche games, such as shareware games—
old-school shoot-’em-ups for example. Games such as
Mutant Storm, from PomPom (
and CrimsonLand ( are specific
examples of small-team shareware games that excel.
Assigning Personnel
A key factor in a project’s success or failure is often the
personnel who are assigned to it and how those people are
divided into functioning groups.
Of course, the roles and divisions we present here are a “best
case” scenario. Very few companies in the games industry
have anything like this configuration, although it is fairly
common outside the games industry. Most game companies
have the main divisions, but some of the subsidiary roles do
not even exist. It is a sad fact that much coding still begins
12 3634_CH10
11:19 AM
Page 246
246 Chapter 10
before any detailed analysis has been made. (In many cases, the analysis is never performed at all.) Outside the games industry, a project such as this would never even get
the green light.
The five main personnel divisions each contain a number of roles, and Table 10.1 shows
some of the main roles and divisions that are used for project-based development schedules. Note that there are other roles in most companies, but here we are considering only
the main ones. (The other roles are mainly support roles that are not directly related to
game production, such as network manager, receptionist, and other ancillary staff.)
Table 10.1 Roles and Divisions
Management and Design
Software planner
Lead architect
Project manager
Game designer
Lead programmer
Lead artist
Music and Miscellaneous
Sound effect technicians
Assorted technicians
(such as motion capture)
Support and Quality Assurance
Technical support
QA lead
QA technician
Support technician
These roles are adaptable to the software-factory methodology, an efficient software
construction technique discussed in Chapter 11, “The Software Factory.”
12 3634_CH10
11:19 AM
Page 247
Roles and Divisions 247
We’ll discuss these divisions and roles in the next few sections, and then we’ll compare
them to the equivalent roles and divisions that are used in the software factory.
The roles described here are not necessarily positions that one person occupies permanently. Instead, these roles should be viewed as “hats” that someone will wear for a
while. People may wear one hat for a long time, while wearing other hats for shorter
periods. People can—and often do—fill more than one role (wearing two hats at once).
For example, the same person can act as both the software planner and the lead architect. In the extreme case, a shareware team of one or two programmers will often fill
all of these roles at various times during the project.
Management and Design Division
This division includes the levels of management that are directly involved with game
production. The lower levels of management tend to be blurred with the upper levels
of technical and design staff, especially in the smaller companies in which everybody
is often required to fill a number of roles.
A lot of this crossover, particularly into the area of game design, is mainly due to most
people’s unshakable belief that they know how to design games. Many managers and
CEOs believe that they are excellent game designers. A high proportion of these managers and CEOs are the original back-bedroom programmers, the one-man bands that
kickstarted the industry back in the 1980s. They designed the games and wrote them
as well, so it is only natural that they would want to keep their hand in the design
process as they develop their companies.
Of course, some of these guys really do know how to design games—the Gollop brothers of Mythos Development are good examples—but many others are in management
positions simply because they have been around the longest and their names are well
known within the industry.
During the development of a game on which we consulted, it was necessary to write
the script for the FMVs. The development team included a television scriptwriter and
an experienced game designer. However, instead of leaving it to them, the president
of the company decided to write the FMV script himself. His previous creative
credentials? He had once managed a rock-n-roll band!
12 3634_CH10
11:19 AM
Page 248
248 Chapter 10
The more technical management positions require concrete skills that can’t be filled by
just anybody. Unfortunately, it is not always the case that the right people are put in
the right positions. Programmers are too often put in management roles because of
their technical proficiency, regardless of their management skills—or lack thereof.
Software Planner
The software planner’s task is to break down a game design into a set of detailed technical requirements and to estimate the time and effort that are required to implement
those features.
The software planner usually works in conjunction with the game designer and the
lead architect to prepare the detailed specifications and work them up into a technical
architecture document that satisfies the needs of the project.
Lead Architect
The task of the lead architect is to work with the software planner to produce a set of
module specifications from the technical requirements.
The lead architect is responsible for the overall architecture of the project. The detailed
module design is usually left to the lead programmer. In some cases, this task is further
delegated to the programmers in the lead programmer’s team.
Note that a software architect is not necessarily the same thing as an experienced
programmer, although they are often technically skilled. Usually, the job of the software architect is full time, as modifications of the specs and reviewing the produced
code is a long-winded but very necessary job. There is rarely any time for
Project Manager
The project manager’s task is to arrange the workload produced by the software planner and the lead architect into an efficient and organized schedule.
The project manager handles interactions between team members and acts as an interface between the project team and the management and marketing departments.
12 3634_CH10
11:19 AM
Page 249
Roles and Divisions 249
Game Designer
The role of the game designer is to design the games that the team will be producing.
The game designer produces the initial game-treatment document, and then goes on to
develop this into the detailed game-design document.
The game designer works with the software planner in order to explore the feasibility
of the game design.
Programming Division
This division includes the team members who do most of the coding work for the
project. The programming division, on a project-by-project basis at least, tends to
be grouped into a small team of programmers working on the one game project.
This team is usually organized into a fairly simple structure, with one lead programmer
(responsible for the overall architecture) overseeing a team of programmers, each of
whom specializes in a different area of the program (such as the graphics subsystem,
the AI engine, or the control system). Some degree of crossover occurs, but, in most
cases, the areas covered by each programmer are pretty well delineated, even to the
extent that the lead programmer does not know what is going on in each of the
Lead Programmer
The role of the lead programmer is to coordinate and monitor the efforts of the
programming team to ensure that the schedule is being maintained.
The lead programmer interfaces with the project manager to ensure that the schedule
is being followed and to report progress so the project can be accurately tracked (and
any problems can be dealt with as expediently as possible).
The lead programmer is usually the most technically able programmer on the team and
is mainly responsible for the overall integrity of the software. Anywhere from half to
three-quarters of the lead programmer’s time is spent programming, while the rest is
spent dealing with administrative and personnel issues. Lead programmers are usually
selected based on technical prowess and not any other criteria.
12 3634_CH10
11:19 AM
Page 250
250 Chapter 10
The role of the programmer is to work in the trenches of the development process.
This involves following the detailed technical specifications that are handed down from
the lead architect and software planner.
The programmer is usually responsible for working on a particular subsection of the
main program, and this remains his or her area of responsibility for the duration of
the project.
The programmer is also responsible for all of the standard programming parts of the
development process, such as initial development, unit testing, integration testing, and
bug fixing.
The programmer is expected to know his or her craft well enough to implement the
most elegant solution without having to waste time on investigating how to do so. No
research is involved here: this is simple conversion of a detailed specification to code.
Art Division
This division comprises a pool of artists. These comprise concept artists, modelers
(often specializing in either characters or architecture), animators, lighting and texture
artists, and so on.
Lead Artist
The role of the lead artist parallels that of the lead programmer, although usually in a
much more loosely defined way. The output of an artist is more immediately obvious
in quality, and so does not need the close monitoring that a lead programmer would be
expected to perform for a programmer.
Consequently, the main role of the lead artist is to interface with the lead programmer
and the game designer to ensure that the artwork being produced for the game is
The lead artist is also expected to ensure that all the artists are producing artwork that
is in line with each other and with the overall project vision. The lead artist is usually
devoted to a single project so that there is at least one constant focus on the artwork
12 3634_CH10
11:19 AM
Page 251
Roles and Divisions 251
being produced for a particular project, even if some of the other artists involved might
be working on several projects at the same time.
A lead artist should probably be spending approximately 10 to 15% of his time on
administration tasks and the rest producing artwork.
The role of the artist is to produce the artwork required for one or more projects.
Artwork, in this sense, can be game artwork, background graphics, manual design,
advertising and packaging design, or any other associated tasks.
It is possible that the artist will be working on several projects at the same time. That’s
particularly true of concept and storyboard artists. Each of these projects has a lead
artist who assigns the work among the pool of artists. This means that each artist may
have to interface with more than one lead artist, unlike the programmer who is part of
a much more tightly knit team.
Music and Miscellaneous Division
This division includes the personnel that produce the miscellaneous bits and pieces
that are required to finish a game, such as the music, sound, and the development
Musicians generally tend to work separately from the main fold of the game
Creating music is a very individualistic activity, and, for most cases, the musician can
be given an animation or a demo, or even just a description of the mood required, and
he or she is then able to produce the music.
If interactive music (which changes in tempo or mood according to what is happening
in the game) is required, a closer interaction of the musician and the rest of the team
will be necessary. Interactive music usually involves much more detailed specifications
of interchangeable themed riffs and loops that can be swapped in and out as required.
12 3634_CH10
11:19 AM
Page 252
252 Chapter 10
The role of the musician has become much more prominent than it used to be. As
interactive music becomes more frequent in games, the musician’s role will become
more central to the game-development process.
Sound Effect Technician
All games use sound effects. The role of the sound effect technician is to produce all
the incidental sound effects and loops that are required to create the game environment, whether they are gunshots, button blips, or alien death screams.
This is usually a fairly autonomous role. The information required to produce sound
effects is pretty much the same as that for music. A list of effects can be created by the
game designer, and these can be produced to the required format by the sound effect
In many instances, the common sound effects will have been inserted by the programmers as test cases. (Many of these so-called test cases end up in the final game.)
Assorted Technicians
A variety of other technicians is always required to perform other tasks for a game’s
Some of these technicians will be directly involved, while others will be more on the
sidelines. Motion-capture technicians, for example, will be heavily involved with the
production of animations. Unless the company has a motion-capture studio onsite
(which is quite rare), the artists will produce a set of prototype animations that were
used during development, and the motion-captured animations will be farmed out to
a studio and inserted in the game at a later date.
Other external studios perform similar duties that may not be manageable in-house.
However, if these duties can be managed in-house, there will be a technician role in
Examples of these sorts of roles include software localization for other countries or
actors for FMV films.
12 3634_CH10
11:19 AM
Page 253
Roles and Divisions 253
Support and Quality Assurance Division
This division includes the testing team that ensures that the game is playable and of
release quality. Testing a game is both a qualitative and quantitative process.
It is qualitative in the sense that honing gameplay to perfection is a fine art.
The process is quantitative in the sense that the number of bugs can be counted and
prioritized. This is the main task of the quality assurance department during the earlier
phases of development.
Quality Assurance Lead
The role of the quality assurance lead is to supervise the QA team and to cooperate
with the project manager and the game designer to make sure that the game is
thoroughly tested, both from the point of view of gameplay and functional coverage.
The quality assurance lead will draw up test plans and assign different areas of
coverage to different QA technicians. The empirical results of the testing will usually
be reported to the project manager.
Quality Assurance Technician
The role of the quality assurance technician is to test the code written by the programming team. The QA technician is concerned with functional coverage, meaning that
the plan he is given by the QA lead should be designed to execute all code paths. All
code that is written should be tested. All paths in the code, no matter how simple or
trivial, should have a test case written for them.
The QA technician’s role is to interact with the programmer responsible for a particular
module to ensure that the test plan is written to test every code path.
The QA technician needs a good grasp of the technical details behind the code, so that
he or she can better understand exactly what they are testing. This is the most detailed
form of testing, and can also be performed by the programmer. This is called clear-box
testing because the internals of what is being tested are known. The alternative to clearbox testing is black-box testing.
12 3634_CH10
11:19 AM
Page 254
254 Chapter 10
Black-box testing tests the outcome, the visible results of the coding effort. An example
would be checking that a polygon-drawing routine draws polygons that are visually
correct. Black-box testing can be performed by any tester who has been given a suitable test plan to follow.
The role of a playtester is just that: to test how the game plays. Initially, the playtesters
are the programmers and artists working on the project.
However, towards the middle and end of the project, the importance of correct
playtesting increases. You generally have five options for playtesting, and the one you
choose depends upon a number of factors, such as the size of the organization and the
amount of money and time available for playtesting.
The first option is to use regular staff, which is not always suitable because they are
already too familiar with the project to be objective about it.
A useful alternative that many companies use is to pay college students a few dollars
an hour to come in and play the games for a while.
The third option is to have permanent playtesting staff. Although this option is most
suitable for larger organizations, it can be cost-effective for any company if enough
projects are simultaneously in progress. Even if there are not enough projects to keep
a team of playtesters constantly busy, the testers can be put to use in other areas of the
If a dedicated team of playtesters can’t be kept busy due to an inconsistent workload,
then the services of an outside playtesting agency can be sought (a fourth option). The
advantage of using outside agencies is that they can test across multiple configurations
with experienced playtesters who know what to look for in a game. The playtesting
agency also performs an excellent job of black-box testing; but, by this stage, the only
bugs you want to see are issues with different machine configurations or other such
obscure problems.
The fifth option—the public beta test—is the best for most organizations, whether
large or small. In a public beta test, a nearly complete version of the software is
12 3634_CH10
11:19 AM
Page 255
Roles and Divisions 255
released to general testing. Other than being a beta version, the software is usually limited in some way (say, by not including the full functionality).
The software to be beta tested can be released with a controlled distribution to a specific
group of chosen testers (such as Origin did with Ultima Online), or it can be released to
the general public for testing by Web or magazine-cover CD-ROM-based distribution
(as id Software did with Quake and Quake II). In the cases where beta testing is limited
and controlled, the company may choose to offer an incentive such as receiving a free
copy of the final product upon release.
Support Technician
The role of the support technician is to support and maintain the computing environment that is required by the rest of the company.
This responsibility entails maintaining the company network, ensuring that all
machines have the correct software installed, and performing hardware upgrades and
other such tasks that are required to keep things running smoothly.
Improving Morale and the Working Environment
It should be obvious to even the most antiquated manager that employee morale and
the quality of the working environment should be treated as valuable commodities.
No single larger factor contributes to the productivity of an employee. The message
here is simple:
Good Morale + Good Environment = Good Work
Morale Boosters
What affects the morale of an employee?
Like all good questions, there is more than one answer, and all are equally valid. The
surprising thing about morale is that it isn’t necessarily fostered by pandering to every
whim of an employee. In fact, doing so can hurt morale in the same way that giving a
child everything he wants is more likely to turn him into a spoiled brat than a future
Nelson Mandela.
12 3634_CH10
11:19 AM
Page 256
256 Chapter 10
The best approach to maintaining morale is through fairness.
Here is a list (in no particular order) of recommendations that we have found to be
beneficial to morale:
Good flow of information
Minimum level of internal competition
Realistic schedules
Fair pay
Good working relations
Ground rules
Pleasant working environment with up-to-date equipment and software
Regular working hours
Constructive benefits
Some of these may seem fairly unorthodox as far as the games industry is concerned, so
we will discuss each point in the following sections and give reasons for their benefits.
Good Flow of Information
The whole team should be allowed access to as much information about the project as
There should be no unnecessary secrets; these only breed resentment. All feasible
information should be made available either in printed or electronic form, and efforts
should be made to ensure that every employee knows how to find the information.
If some of the information may be considered sensitive outside the company, there are
two approaches. The first is to require that all employees sign cast-iron non-disclosure
agreements (NDAs). The second is to only reveal non-sensitive information and to
explain that the information not revealed is due to its sensitivity. Of course, with the
latter option you must guard against relying on it as a crutch. For example, using it to
hide information that is not sensitive, per se, but merely unpleasant would be an abuse
of privilege.
12 3634_CH10
11:19 AM
Page 257
Roles and Divisions 257
Minimum Level of Internal Competition
Internal competition between two factions—whether they be single employees or
entire teams—is more often destructive than it is constructive.
In most cases, simple competition degenerates into rivalry and sniping at others’ backs,
while the rivalry grows to overshadow the needs of the project.
Most rivalry arises through insecurities. If an individual or team feels insecure about
their position, then they will attack someone else in an effort to improve their standing. This is just human nature; man is a political animal. In some cases, this rivalry
can take the form of scapegoating, while in others it manifests as a propensity to be
combative and uncooperative.
If employees feel secure and comfortable, you are more likely to see a healthier competitive dynamic occur: friendly rivalries between groups and individuals to improve
the product. This “coopertition” is a great morale booster in its own right, but it also
has the added effect of increasing productivity, which further increases morale.
Realistic Schedules
This is a simple point and doesn’t need much saying about it here. It is covered in
other chapters.
A schedule must be realistic. Pitching an overly aggressive schedule at developers and
trying to persuade them that it is attainable sends morale sinking like a lead balloon.
Fair Pay
Fair pay means many things to many people. This is what it means to us.
The staff should be paid a fair wage that is based on their experience. When possible,
a staff member should receive industry-standard wages based on experience and skill
level. By “industry standard,” we mean standard of the computing industry and not just
the underpaid games industry. This can include royalty deals, but it is widely understood that most industry royalty “deals” are not worth the paper they are written on.
12 3634_CH10
11:19 AM
Page 258
258 Chapter 10
Staff should not be expected to work long hours for free. Each hour worked should
be paid for, either pro rata, or at an agreed overtime rate. Don’t subject an employee to
the indignity of working a bucket-load of extra hours to get a demo finished for an
important client only to be rewarded with a pizza of his or her choice!
If royalty or stock options are offered, there must be written royalty agreements that
are legally binding and that clearly state all deductions and clauses in plain language.
If you want to get your royalties, then get the agreement in writing, and always get it
checked over by a competent attorney before signing.
Good Working Relations
This is a simple point. Trust.
The different groups of employees must have a mutual trust and respect for one another.
In the situations where this is not the case, the resulting friction is enough to destroy
projects and make the working environment a living hell.
We believe in fostering a “bridge of the Enterprise” culture. If you look at how the
crew operate together on Star Trek: The Next Generation, you see that there are rivalries, but they are always channeled into the greater good. The working practices of
Starfleet constitute an ideal goal that possibly can never be achieved in real life, but
it is useful as a paradigm of what the perfect team should be.
Ground Rules
Morale can be improved by a clear, simply understood set of ground rules that cover
behavioral and professional expectations while on the premises.
These rules should not be draconian or pointless, or rules for the sake of having rules.
They should be grounded in basic common sense. They’re written down to ensure that
everybody has the same concept of what constitutes “reasonable” behavior.
These rules define the expected behavior of all employees while they are on the
premises and being paid by the company. These rules are the great leveler.
With a set of clear guidelines, no employee is likely to feel as if he or she is being
unfairly treated if someone is allowed to behave in a different manner (regular long
12 3634_CH10
11:19 AM
Page 259
Roles and Divisions 259
lunches or some other liberty) if it is against the rules. This may sound a bit schoolboyish, but it is much better than the anarchic alternative.
If the employees are responsible and mature, then they can be relied upon to produce
their own set of guidelines for the finer points. Even so, a recommended minimum
is general company-wide guidelines that define a start time, a finish time, and the
allowable range of break and lunch times.
Note that, if the start and finish times are specified, they should be respected. Staff
should be encouraged to arrive and leave on time.
A Pleasant Working Environment with Up-To-Date Equipment and Software
The importance of a pleasant environment should be self-evident.
Working in a cramped and dimly lit office is demoralizing. In a perfect world, offices
would be clean, airy, and spacious, with every employee receiving an office of their
own. Not only is this not always feasible, it is, in fact, rarely feasible. (Microsoft
apparently provides all its developers with private offices. The oft-heard counter to
this is that Microsoft is not exactly short of cash. But maybe the reason Microsoft
continues to thrive is precisely because of factors like this.)
At the very least, the employees should be provided with desks of their own that are
spacious enough to provide ample room for all the required equipment. We do not
believe that management should have big expensive desks (just because they are
management) when the big desk is probably more useful to a programmer with a
large computer, a stack of CDs, and two monitors on his desk.
It is not difficult to make an office pleasant to work in. Low-maintenance plants can
add ambiance, as can indirect lighting (not overhead fluorescent lights). Natural light
should be used wherever possible, but it should also be possible to shut it out. Nothing
is more annoying than working with the sun shining directly onto your monitor or
into your eyes, as Ion Storm Dallas undoubtedly discovered when they took possession
of their glass-roofed office in a Dallas sky-scraper.
Office design is easy to get right. Aim for an environment that looks professional and,
more importantly, makes you feel professional.
12 3634_CH10
11:19 AM
Page 260
260 Chapter 10
Regular Working Hours
In many parts of the games industry, there is still a tendency to work almost random
hours, and all-nighters are not uncommon.
These irregular working hours are, in some cases, the sign of a badly scheduled project. In other cases, programmers got used to working irregular hours on their own pet
projects, and they imported that culture with them when they joined the company.
Whatever the reasons for the lack of regular hours, the fact is that they represent a
symptom of an amateurist culture. Morale is higher among people who see themselves
as professionals, not amateurs, and will improve when regular hours are implemented.
The benefits are soon evident. If everybody works a regular set of hours, you can be
sure that all members of the team are available at the times when everybody else is.
There’ll be no more cases of missing team members sleeping off the effect of an allnight coding marathon, leaving the morning crew with convoluted code to decipher.
Having a rigidly defined set of core hours also reduces stress on employees. If it’s clear
that they are required to work only a set number of hours a day, then they can concentrate on working those hours, and leave work at a reasonable time.
By not working ridiculously long hours, the employees will not be fatigued for the
following day. In the long run, that means they will be more productive.
Constructive Benefits
If the company is going to offer benefits to the employees, these benefits should be
ones that the employee can appreciate and find useful.
Stock options, guaranteed royalties, free gym membership, training courses, and tickets
for relevant shows and free soft drinks are good benefits.
Cups, pens, beachballs, posters, and other such cheap detritus are not good benefits.
Issued in lieu of other benefits, these things send the message to your employees that
you think of them as amateurs. That doesn’t mean that the company should stop
providing employees with fun freebies —only that they mustn’t be a substitute for
real benefits such as stock options.
12 3634_CH10
11:19 AM
Page 261
Roles and Divisions 261
If you want an employee to feel appreciated, don’t make him or her feel as if you think
their worth is adequately expressed by the offer of a keyring.
Morale Booster Caveats and Warnings
There are some actions that appear to boost morale in the short term, but that have a
detrimental effect on long-term morale.
This has been touched on above, as it is a manifestation of the spoiled-child syndrome.
If you give employees enough rope, not only will they hang themselves, they will also
tangle up your company as well. This list covers a few of the common morale boosters
that are mistakes.
No dress code
Freeform hours
Overly casual working environment
These measures provide an initial boost in morale, followed by a rapid downward spiral.
No Dress Code
A complete lack of dress code is a morale killer.
At first, of course, it seems pretty cool. Employees feel free to express themselves and
their individuality. They feel relaxed and comfortable in the type of clothes that they
would wear at home in their free time.
The problem, however, comes in the attitude. Allowing an employee to feel as if they
are at home means they will be tempted to behave as if they were at home. For work
to feel like a place of work, it is important to differentiate it from home.
Freeform Hours
Some companies and organizations have a remarkably lax attitude towards working hours.
Usually, the situation is that the hours worked are a lot longer than the standard.
Twelve to 18-hour days are not unheard of, and these long days hurt the project in two
ways. The first is obvious: no one can consistently produce good work if they are
working that much. They will be produce substandard code, and they will be overtired.
12 3634_CH10
11:19 AM
Page 262
262 Chapter 10
Eventually they will burn out. When this happens, they will take more time off and
maybe even leave, which damages the team by leaving a lot of legacy code to be maintained. Worse still, they will have been perceived as being a hard-working member of
the team, and their departure will affect the morale of the remaining team members.
If they just become ill or overtired and take time off to recover, then this hurts the
team too. The entire team should be working in the same location at the same time
for at least a core set of hours. If not, then they are not really a team. They become a
collection of individuals working on a common project.
Overly Casual Working Environment
Many game companies have very relaxed rules of workspace and environment, with
posters, toys, stereos, inflatable toys, and regular lunchtime Quake sessions that last
into the afternoon.
In fact, all this does is make the office appear like a teenager’s bedroom, which violates
the rule about home and office separation. The games industry is viewed as being a fun
and cool place but, let’s face it, the object of working in the games industry is to create
games, not play them.
The office should be a professional environment because professional environments
create professional employees.
Spreading the Risk
The risk associated with specialized roles is the possibility of losing a staff member
who performs that role. The roles defined in this chapter are not job titles; they are
merely roles, and one person can fill one or more of these roles as a part of their jobs.
The roles should be viewed as hats that are worn at different times during the course of
an employee’s work.
This can be used to the company’s advantage as a risk-reduction mechanism.
13 3634_CH11
11:20 AM
Page 263
Chapter 11
The Software Factory
• Problems solved by the
software factory
• Setting up a software
• Applying a software
hapter 10, “Roles and Divisions,” covered the roles and
divisions that are needed for an efficient development
environment. These resources can be organized into many
different structures (not all of which are suitable for game
development). This chapter explains the roles and divisions
of an organizational structure that is particularly suitable for
established development houses and demonstrates how this
structure can increase performance to an optimum level.
The structure can also be applied to a startup organization, but
doing so causes the first project to take longer (or at least it will
seem to) as explained later in this chapter.
What Is a Software Factory?
The term software factory refers to a methodology of producing software with techniques similar to those used in a standard factory, such as one that manufactures cars. This doesn’t
mean, however, that all the software produced by this
method will be identical, churned out endlessly, and devoid
of imagination and flair. Enough companies are already too
good at producing creatively bankrupt software.
• Suitability of the software
13 3634_CH11
11:20 AM
Page 264
264 Chapter 11
As the computer games industry begins to mature, more sophisticated production
methods, such as the software factory, are being implemented to develop new products.
The software factory methodology centralizes and simplifies the production of specific
common modules that can be used across several products.
It uses the advantages of mass production to make specific tasks both easier and more
efficient, and it has the benefit of ensuring that a core set of tools and libraries are well
maintained and supported over a series of products. Thus, subsequent products
become easier to develop, as they will be supported by a more resourceful library and
useful tool set.
The software factory has been proven to work time and time again on projects with
common functionality. Keep in mind that this method may not be best suited to a particular product. (You or your co-workers may have a better methodology scrawled on
the back of an envelope!) However, software factories are particularly well-suited for
producing a series of products.
“But these are games we’re writing! Each one is unique! There is no way we can rehash
the code, change the graphics, and re-release the game!”
This objection is perfectly valid, but how many times do you want to write screen and
sound setup code, data file loaders, compression libraries, CD track playing code,
finite-state machine code, menu code, or any other chunk of potentially reusable code
from a long list of basic modules? Looking at it from a management perspective, how
much time and money do you want to spend for specialists to write code that you
already have? Not only does the code have to be written, it also has to be tested, integrated, and debugged.
Some common tasks that should be placed in reusable modules are as follows:
Data file loading
Hardware setup
Hardware configuration
Software configuration
13 3634_CH11
11:20 AM
Page 265
The Software Factory 265
Custom CODECs (compression and decompression)
Encryption/decryption code
Windowing and graphics features
Basic AI components
User input
The software factory concept eases tasks such as producing common libraries and tool
sets, leaving the developers to concentrate on the code bits that should be written from
scratch. The fastest work is the work that has already been completed and tested for
full functionality.
Why Use a Software Factory?
What advantages and disadvantages does the software factor have over other methods
of development? Although the answer to this question depends on the particular situation, Table 11.1 lists several key advantages and disadvantages. However, on the whole,
the advantages clearly outweigh the disadvantages over the course of several projects.
Table 11.1 Advantages and Disadvantages of a Software Factory
Average project length will be shortened
First project will take longer
Makes cross-platform releases easier
Reusable wrapper libraries must be developed for
each platform
Code will be more reliable
Code will be more generic, and more difficult to develop
Code is reusable and maintainable
Code takes longer to develop initially
Knowledge is spread out among all
New developers have to learn unfamiliar libraries
Increased visibility of project progress
More administration
Increased specialization of developers
Less skill flexibility
13 3634_CH11
11:20 AM
Page 266
266 Chapter 11
Solving Game Development Issues
The software factory methodology helps to solve a number of common issues in game
development, namely platform independence and risk reduction (knowledge spreading, code reuse, code maintainability, and so on).
Platform Independence
Consider the DOS-compatible PC—an expandable system with a wide variety of possible configurations. If, for performance reasons, you want to write directly to hardware
for this PC, how would you choose to do it? You would have to ask yourself what hardware you wish to support and in which machine. The choice would eventually come
down to supporting only a limited subset of available hardware, which would eventually
reduce your potential sales. You can also bet that the vast majority of support calls
would be due to conflicting or incompatible hardware—a logistical nightmare.
Let’s look at a specific example. Assume that you have written a 3D space combat game
for the DOS-compatible PC. You have been fairly sensible in your architectural design
and have isolated all platform-specific code behind interfaces. (Here, the word platform
indicates a unique hardware configuration for the PC.)
The code behind these interfaces must be rewritten for every platform you want to support. You have chosen to support three popular graphics cards and three sound cards.
Thus, you need to write, debug, and unit test six hardware-specific code interfaces, and
then system test nine possible configurations—not to mention the software-only solutions for everybody else who doesn’t have one of the chosen graphics and sound cards.
To make matters even more difficult, to get the speed required, you have written the
libraries in such a way that they must rely on internal knowledge of the game for full
functionality. That is a lot of work in anyone’s book.
Another common issue to consider is how can you be sure that you have written all of
the platform-specific code in the most efficient way possible. It is unlikely that you
were able to find enough developers with thorough, in-depth knowledge of each sound
and graphic card you wanted to support. Given the amount of code you had to write, it
is also unlikely that you have had the time to develop that knowledge.
Of course, this is a contrived example, and has no relevance to the real world. The real
world, in fact, is much worse.
13 3634_CH11
11:20 AM
Page 267
The Software Factory 267
Fortunately, as we predicted in the first edition of this book, game specific middleware,
pioneered by Microsoft’s DirectX, has stepped up to the plate and plugged the gap.
Tools such as NDL’s GameEmbryo and BioGraphic Technologies’ A.I.Implant are now
widely used to enable game developers to create cross-platform (PC, console) content,
reduce the development time substantially, and provide a well-tested, stable engine.
Among other things, these middleware libraries help solve the problem of hardware
independence and obviously prevent writing directly to hardware except through the
provided standardized mechanisms. (That would be contrary to the spirit of insulating
the developer from the underlying hardware.) In the case of DirectX, a common set of
features is provided that guarantee to be present in hardware or emulated in software.
The acceptance of DirectX by developers was initially slow, but try to find a PC game
today that is written without the aid of DirectX. In fact, it is increasingly rare to find a
popular game that hasn’t been written without the aid of middleware, such as the
Quake III engine or the Unreal engine.
Of course, the middleware libraries do have subtle differences across disparate platforms. To have true platform independence, you must create lightweight wrapper code
to present a uniform interface to hardware, no matter whether it is running on a DOScompatible PC or a Macintosh or console. Even if you are not planning to support more
than one platform, the wrapper code is still an excellent idea because it simplifies many
common tasks. Creating such wrapper code is one of the main tasks of the factory.
Risk Reduction
One of the problems discussed in Chapter 10 was the loss of key personnel. If one of
your team members has a specific, inflexible expertise in a specific area of knowledge
or skill set related to the project, then that person becomes a risk.
At this stage you should ask yourself whether you are willing to bet the whole project
on that one person. If they ran off to live in communion with the sky in the Tibetan
foothills (or even join another games company), would the project survive the loss? If
the project can survive, what are the costs in time, money, and functionality? Reducing
inflexible expertise is critical to the overall success of the project.
The software factory can help alleviate problems such as losing key personnel by encouraging multiple redundancy and the reuse of code. The scripting engine mentioned in
Case Study 11.1 was written specifically for the project. However, as an alternative,
13 3634_CH11
11:20 AM
Page 268
268 Chapter 11
it could have been written to an interface, and every scriptable game object would then
export methods to support that interface. If these interfaces are designed generically, the
ability to reuse and utilize them over many projects comes automatically.
Case Study 11.1 The Effects of Losing Key Personnel
According to Jason Regier at Bungie Software, the development of Myth: The Fallen
Lords suffered a setback when a key programmer left. The individual was responsible for the scripting engine, an important (but fortunately not essential) module
that then had to be scrapped. When the programmer left, the scripting code base
was incomplete, and no one had sufficient knowledge of the code to continue
developing it. Moreover, the Myth team was relatively small and consequently
was not even able to spare the manpower to learn how the code functioned.
The scripting engine was designed to allow custom unit and map behavior. It had
also already been publicly announced as an included feature.
This particular problem was solved simply by dropping the functionality. A scripting engine is a complex undertaking and is usually not a critical path module.
In this case, Bungie made the decision to drop it based on a cost/functionality
comparison. If the module had been the graphics engine, such a decision would
have been impossible.
Source: Game Developer Magazine (April, 1998)
If you are a true dyed-in-the-wool game developer, then you may feel slightly uneasy
after reading the last paragraph. Standardized interfaces? Won’t that slow things down
by adding an extra layer of indirection? Won’t it be harder to code and take much
Well, yes and no. If the interfaces are properly designed, you will be able to get performance at least equal to a custom integrated component. And, because the emphasis
of the software factory is on reusability, the scripting engine can be refined and modified over the course of several projects. For a more detailed treatment of how such a
scripting interface could work, see Chapter 19, “Building Blocks.”
13 3634_CH11
11:20 AM
Page 269
The Software Factory 269
In Case Study 11.1, the problem would have been a nonissue if more than one programmer had been familiar with the scripting engine and if the code had been fully
documented. Documentation is an essential part of the development process and is
often (actually, nearly always) overlooked. Why bother to document something if it is
only going to be used once? Why document procedures or code when you know the
code inside out and are the only one who is going use it?
These are acceptable responses in principle, but it begs the question as to why you are
writing throwaway code. If you assume that writing such code can be justified, what
happens when you feel that sudden inescapable urge to go to Tibet? How can you be
sure that you are the only person who will ever need to read that code?
The ideal path to follow is to reexamine the code being written to see if it can be
written with more than one project in mind, and with a full set of documentation.
Why Use A Software Factory?
A software factory is designed to produce a set of reusable core components and tools
that evolve over the course of several products. From the very beginning, these tools
and components are designed with reusability and future expansion in mind. As much
care goes into developing these tools and components as would into the game itself.
The development methodology used for these tools and components is rigorously
controlled and monitored, and full documentation is made easily available for use in
further projects. Any required updates to the tools or components undergo a full
requirements analysis and are performed to the same exacting standards. These tools
and components are the most important part of the whole process. They are designed
to have a shelf life lasting for several projects and are therefore of key importance, as
discussed in Case Study 11.2.
Case Study 11.2 Code Reuse
According to Wyeth Ridgeway of Zombie, the Viper game engine developed for use
with their game, SpecOps: Rangers Lead The Way, contains no game-specific code. It
is a module specifically designed for reuse.
The engine consists of a 3D renderer, an object-scripting module using a LISP-like
language, and a sound module, among other things.
13 3634_CH11
11:20 AM
Page 270
270 Chapter 11
Taken by themselves, these components are just tools. The game is composed of the
game engine and game-specific resources (graphics, sounds, and object scripts). It
is driven by a minimal amount of glue code, consisting mainly of the object scripts
and the small amount of game-specific code that is needed to support the scripts
and provide a general framework.
The engine has been designed in such a way that, with a different set of resources,
one could craft a completely new and exciting game. This ability was demonstrated
when a couple of nonprogramming team members were able to create a monstertruck racing demo over the span of a weekend. This shows that with proper planning, a minimal understanding of game development, and detailed and precise documentation, a fully operational and productive software factory can be a realization.
This is the sort of result that the software factory aims for.
Source: Game Developer Magazine (June, 1998)
The advantages are obvious: code that performs common tasks can be reused again
and again, thus saving development time and financial resources. The main disadvantage is that the initial set of tools and components must be developed in the first place,
and this will always be a slower process than just hard coding the desired functionality
into the project. The gains only begin to materialize on subsequent projects.
One other option is the possibility of incremental improvements to released projects by
distributing updated core modules. Use caution on this approach. You don’t get something for nothing in this world, and, although ideally all modified components should
be backward compatible, this approach may place a large burden on testing. If you are
confident that you can do it, then go for it! However, remember all the difficulty that
Microsoft had ensuring this sort of backward compatibility with releases of DirectX.
Up until version 5, new updates regularly broke what was on the system before the
update, rendering the system suboptimal at best, and unusable at worst.
An analogy would be building a sailing ship. If you decided to re-create everything
needed from day one, the team would have to go and fell the trees, shape the timbers,
13 3634_CH11
11:20 AM
Page 271
The Software Factory 271
season the wood, and then assemble the ship’s components. Doing this for multiple
ships would soon become overwhelming.
Rationally, you might arrange for the initial components to be developed by a productionline process (even from an outside source) and then use the team for assembly only.
This combines the advantages of the team model (enthusiasm and motivation) with the
advantages of the factory model (efficient use of labor and just-in-time processes).
Organizing a Software Factory
The structure of the software factory is different from what you may be used to. The
following sections describe how a software factory is comprised.
A Structural Overview
The core requirements for a minimal software factory are defined in Table 11.2.
Table 11.2 Essential Groups for a Software Factory’s Core Team
Game Designers
Create ideas and produce design documents
Software Architects
Oversee the project architecture and interaction among teams
Teams of level editors, and such
Produce low-level, platform-specific code
Produce actual game code using the output of the components
and tools teams
Keep abreast of new technological developments and research
new ideas for integration into low-level components
Table 11.2 is a basic list of the member groups of the factory. It is not exhaustive or
final. Other ancillary roles may be required, but this list defines the core requirements.
Granted, not all organizations can fulfill such lofty requirements, but, aside from the
software architects (which is a specialized, virtually full-time role), there can be a lot
of overlap among groups, as shown in Figure 11.1. Obviously, the more personnel
available, the less overlap there will be, but some degree of overlap is always desirable
to help the spreading of knowledge throughout the team.
13 3634_CH11
11:20 AM
Page 272
272 Chapter 11
Game Design
Figure 11.1 Group overlap in a software factory.
The functions and interactions of the core factory teams are described in detail. Figure
11.2 displays the hierarchical structure of the software core factory.
Game Design
Core Component Group
Figure 11.2 Hierarchy of a software factory.
Figure 11.2 is a rough guideline and not a rigid structure. In reality, the situation is much
more dynamic than can be shown in a static diagram. What is fairly rigid, however, is the
method of interactions among teams. These interactions (or interfaces) help provide
the high level of project visibility. Project visibility is the ability to accurately gauge the
progress of the project. This ability is very important to everybody concerned with the
product, including the investors, the publishers, and the teams themselves. Somewhat
surprisingly, a side effect of good project visibility is to reduce the number of team meetings. A lot of so-called “progress” meetings turn out to be anything but that, and, unless
projectwide changes are necessary, there is very rarely the need to stop development for
the length of time a typical meeting takes. Such meetings are one of the biggest time
killers and are disruptive to the team’s focus and work already in progress.
13 3634_CH11
11:20 AM
Page 273
The Software Factory 273
Group Responsibilities and Interactions
Each of the core groups in the software factory has clearly defined roles and tasks.
There is little, if any, task overlap, and all-important intergroup interaction takes place
through rigidly defined channels.
These interactions are based loosely on an internal market system. Each group is the
customer of each others’ groups and thus, should expect to be treated as valued clients.
This means providing exactly the same levels of documentation and product support
that would be expected if the software module in question had been sold on the open
market to an external client.
This “market system” structure is important to ensure that the progress of the project
is tracked correctly. It’s always a good idea for the left hand to know what the right
hand is doing. Figure 11.3 shows the interactions among groups in a software factory,
excluding outside influences. In the figure, the programming groups are positioned
together as one large group, because they often will all be the same people. However,
for interaction among the separate programming groups, they still have to go through
the proper channels. (These channels are described in more detail in Chapter 13,
“Procedures and Process.”)
Game Design
Thick lines
direct contact.
Thin lines
indicate indirect
contact through
project visibility
Programming Groups
Figure 11.3 Interactions among groups.
13 3634_CH11
11:20 AM
Page 274
274 Chapter 11
Figure 11.3 summarizes the main interactions. At first, this looks restrictive. Here is a
list of possible questions that you may have:
Why is direct contact allowed only with the architecture group?
How can the game designers influence how the game turns out if they can’t talk
to the programmers directly?
Why does the architecture group have to be in the center of everything?
Why is the research group directly connected to the architecture group?
To answer these questions, you should consider the typical scenario, as presented in
Case Study 11.3.
Case Study 11.3 Ineffective Problem Handling in Action
Andy, Barry, Chris, Dave, and Eddie are the programming team working on
FlyBusters III: Beyond The Flypaper, the latest masterpiece by Freddy, a veteran
game designer.
Eddie is the lead programmer, a technical genius but something of a loner. Andy is
an intern fresh out of college. Barry, Chris, and Dave are competent programmers
who have work experience in the games industry. FlyBusters III is their first project
together. Andy is new to the world of work, and isn’t sure how everything fits
While implementing a physics model for the player’s ship, Andy notices that at
some angles there is an unpleasant distortion of the landscape. Now, Eddie has a
reputation for being somewhat difficult to approach; his favorite response is to
imply that his time is being wasted, and that he has much better things to do than
listen to team members whining about little problems. (Andy has had his fingers
burnt like this before when he mentioned to Eddie that gaps were appearing
between polygons when they got close to the viewer. Barry and Dave, the programmers of the polygon engine, had taken great offense at this criticism of the code.)
This time Andy tries a different approach and mentions it directly to Chris, the
developer of the landscape engine. Chris, although busy, ponders for a few minutes
13 3634_CH11
11:20 AM
Page 275
The Software Factory 275
and suggests that if Andy restricted the player from looking directly down on the
landscape, the problem wouldn’t be so apparent, because there is no way that the
landscape engine can support that particular viewpoint. This is a small modification
that will take a couple of hours to put in place, and so Andy gets straight to it.
Eddie is particularly busy today and has been giving off a “Do Not Approach” aura,
so Andy makes a mental note to tell him tomorrow and gets on with the next task.
A few weeks later, Eddie decides that it is time to implement the bomb view—a
special viewpoint that is designed to allow easy targeting of superfly buzzbombs
(which have to be dropped with deadly accuracy onto several targets spread across
the landscape). When the view has been implemented, Eddie can’t quite understand
why the bombs do not fall precisely, even though he specified exactly where the
bomb was supposed to hit by setting the orientation and position of the player’s
ship using Andy’s physics API. A little investigation soon tells him that the physics
engine isn’t working properly. After grilling Andy as to whether he has checked his
code properly, and verifying that the problem indeed exists, Eddie makes a decision
to leave things as they are, seeing as a lot of code has already been written on top of
the physics engine, and changing it may break the dependent code.
Two weeks later, during routine testing, Freddy, the game designer, notices that it is
too difficult to target bombs correctly. In fact, it seems to be a very “hit and miss”
process. (Pardon the pun.) He immediately demands a resolution to the problem,
insisting that the bombing strategy is central to the game. The programming team
has no choice but to reimplement three areas of code, which delays the project by
three months. Andy feels unappreciated and not respected. Chris feels as if he is at
fault for not producing a good enough landscape engine. Barry and Dave feel
annoyed and irritable that Andy should have criticized their polygon engine. Eddie
feels stressed at being held responsible for the delays, and Freddy thinks that the
whole team is a bunch of idiots. Team morale sinks, and the game is released six
months late to awful reviews mainly because of the badly implemented bombing
mode and the outdated technology. In short, it bombs.
With quick and dirty, the dirty will remain long after the quick has been forgotten.
Now let’s consider the alternative scenario if a software factory structure is in place, as
in Case Study 11.4.
13 3634_CH11
11:20 AM
Page 276
276 Chapter 11
Case Study 11.4 Effective Problem Handling in Action
The same team from Case Study 11.3 is developing the same title, FlyBusters III, and
after initial resistance has agreed to adhere to the communication channels in place,
at least on a trial basis.
When Andy discovered the polygon-fitting problem, he logged on to the company
intranet Web page and submitted an anonymous report detailing the situations in
which they occurred. The architecture group examined the problem and discussed it
with both Freddy, the game designer, and Eddie. Freddy asked if it could be fixed,
and Eddie, after talking to Barry who explains that the symptoms are a deliberate
speed optimization, announced that it could be done, but only at the expense of an
extra check per scanline per polygon. Freddy asked what that meant in real terms,
and Eddie explained that it could mean a substantial drop in frame rate, depending
on how large the on-screen object was and how many polygons it contained. Freddy
didn’t want to lose the smooth-flowing effect of the engine, so he accepted that it
didn’t particularly affect the gameplay. The problem was resolved as “no action
required,” and this was logged on the Web site, with a brief explanation.
Andy then came across the problem with the top-down view, and again submitted
an anonymous report. This time, however, Freddy insisted that the bombing view
was very important to the game play. Eddie, after consultation with Chris, discovered that to implement a true top-down view would involve developing a separate
subengine, and that this would take a further three to six months of work. The
problem was left “open for investigation,” and the architecture group and Freddy
went away to think of a solution. After a couple of days of deliberation, Freddy
came up with an alternative bomb view that did not need to look directly down on
the landscape, and presented it to the architecture group. In the meantime, the
architecture group noticed that another project group had developed a top-down
map module that would be easy to adapt into the FlyBusters project due to the standardized interfaces, but some extra graphics would need to be generated using the
accompanying tools. Now, Freddy had two choices as to how to implement the
bomb view, and was satisfied.
13 3634_CH11
11:20 AM
Page 277
The Software Factory 277
The solution chosen was then logged onto the project intranet site, and the topdown view was smoothly integrated into the existing project base over a period of a
few days, after the architecture group had updated the architecture specifications to
include the new functionality. Eddie had been on enough traditional projects to
know what would have happened if this procedural approach had not been in
place, and was pleasantly surprised to see how quickly and easily the problems
were resolved. Andy was quietly pleased that his input was appreciated, even
anonymously. Chris was relieved that he didn’t have to make a difficult modification to his landscape engine, and Barry was pleased that he was right to implement
that optimization in the first place. Team morale improved, and the product was
released on time and became a roaring success.
The obvious advantage here is the increased visibility of the project’s progress.
Problems are dealt with as soon as they arise. The benefit of providing an anonymous
channel that can be read by anybody on the project is that it allows everyone to voice
their concerns without having to deal with personality clashes and other political problems that so often rear their ugly heads. The main concern should be the success of the
project, not the sensibilities of easily offended personnel. The anonymous channel
helps to ensure that this is the case. After all, it is not important where the information
comes from, only that it does come, and that it comes as soon as possible.
With a lot of developers, especially those in the games industry, there is initial resistance
to imposing a more defined team structure such as this, mainly because it’s viewed as
overly restrictive and creatively stifling. You may get developers saying that these procedures slow down and reduce their productivity. These will usually be the developers
who are used to cranking out (almost) working code at high speed and getting things
up on screen quickly. These are also the same developers who spend a couple of months
fixing bugs when the project slips due to the high defect count. In actuality, what these
developers are complaining about is having to do the work that is usually postponed to
the end of the project and well past the shipping date. By enforcing the procedural
approach, most defects will be caught at an earlier stage.
13 3634_CH11
11:20 AM
Page 278
278 Chapter 11
It is a well-known statistic in development circles that the earlier a bug is caught (and this can be
in any stage of the project), the less time and money it will cost to fix. A bug caught during the
architectural stage can cost up to 200 times less to fix than if it were caught during the release
preparation phase.
The perceived extra work and delay caused by the extra procedures is only an illusion.
For most projects today, this work still has to be done, but it is never scheduled in the
project plan and ends up being done later in the development process. Meanwhile,
management anxiously awaits the project’s release, which has already slipped past shipping date by six months. Late projects damage the morale and credibility of the team
and company. Any measures that can detect and correct these problems as early as
possible become indispensable. After all, how many times have you heard after a year
and a half of solid programming that a project is 90% complete, only to have to wait
another year or so for what should have been the remaining 10%?
The following are arguments against the introduction of architecture design procedures: there is no time for all of the delays and paperwork that it will cause, the project
is already on an aggressive and tight schedule, and adding all this overhead will only
slow things down in the long run. Having seen the poor states of projects that have
advanced these arguments, the answer is that there is no time not to implement the
measures! Saving time in the long run makes up for any slowdowns in the earlier
stages of the project. By developing at a scheduled pace (and foreseeing problems and
fixing them as soon as they are discovered), time is saved throughout the remainder of
the project.
For more in formation on communication mechanisms, see Chapter 13.
Now that you know the reasons for restricting the interaction among the various
groups, the following sections will briefly explain the mechanisms and channels that
are in place. Note that information will still be transferred in other ways: impromptu
corridor meetings and discussions during lunch or breaks are an important mechanism
for information transfer. The primary concern is that any issue brought up in these
informal meetings must go through the proper channels before any solution or action
is performed. This gives everyone concerned the opportunity to have their say.
13 3634_CH11
11:20 AM
Page 279
The Software Factory 279
The Game Designer Group
The game designer group is responsible for producing the initial game designs and subsequently refining any gameplay issues that arise during the development and testing
process. As such, members from other groups can be co-opted into the game designer
group to contribute ideas and refinements to the game design. However, as discussed
previously, game design is generally more complex than it appears, so the generation
of the initial designs and specifications are best left in the hands of experienced game
designers (or people who have read Part I, “Game Design,” of this book). Usually, this
group is also responsible for end-user documentation, such as the manuals for the
The game designer group interacts mainly with the software architect group, and
generally tends to rely on the project visibility measures to follow the progress of the
project rather than directly approaching the other groups for information.
The majority of this group’s time is spent researching new designs and refining designs
currently in development. As the tools and components groups get up to speed, the game
designers may find themselves being even more “hands-on” in tweaking and programming. For example, if a simple AI scripting engine is developed, the game designers will
be able to contribute directly to writing scripts. If the project has been developed using a
database to store all commonly tweakable parameters, the game designers also have the
opportunity to tweak parameters to their hearts’ content. These parameters are deemed to
be soft architecture and can be modified without changing the structure, the hard architecture. The domain of the game designer group can include direct modification of the soft
architecture, but this has to wait until the tools are in place.
The game designer group also interlaces with the ancillary sound and art groups in
order to specify the look and feel of the game.
The Software Architect Group
The software architect group acts as the central hub of the factory and is the main
interlace between the game designer group and the programming groups (tools, components, projects, and research). The members of this group represent the traditional
“lead programmer” for all the programming groups, although they will be doing little,
if any, actual programming.
13 3634_CH11
11:20 AM
Page 280
280 Chapter 11
The primary responsibilities of this group are translating the game designs into technical architecture documents, making use of the components and tools developed by the
tools group, writing the documents for any necessary modifications to these tools and
components, and initiating the development of new tools and components from the
results of research. The group members also provide feedback to game designers to
ensure that the design is feasible, and to advise on technical issues related to the game
envisaged by the game designer. The specification of the architecture is an ongoing
process. The architecture is progressively refined as the projects progress. Most of the
major shake-ups will have been sorted out in the feasibility study and in the prototyping
stages, and any further modifications should ideally be only refinements and clarifications. It is important to realize that—it is impossible to complete more than 80% of a
given architecture before coding begins. It is this final 20% that is worked out during
the lifetime of the project. This is known as the 80/20 rule.
The software architect group is in complete control of the overall company standards for
architecture and coding. This control encompasses diverse issues such as file formats,
interface definition standards, documentation standards, directory structures, source control standards, and other common areas where reuse and maintainability are paramount.
Code reviews are performed by a member of this group, which ensures that the architectural standards are maintained across projects. The group also oversees the separate
system testing of components, tools, and projects.
The software architecture group is generally composed of the most technically skilled
members of the company. They are skilled in architectural design and are able to produce detailed specifications that can be followed without ambiguity by a member of
one of the three programming groups. Members are technically skilled enough to
answer any questions and explain any technical points to the programmers who are
following the specifications. If necessary, training should be provided to achieve this,
or skilled software architects should be brought into the company.
The Tools Group
The tools group’s primary function is the production and maintenance of the tools that
will be used by all other development groups. In some cases, the tools will be written
in-house, and in other cases the tools will be more general off-the-shelf products such
as 3D Studio MAX and Microsoft Visual C++.
13 3634_CH11
11:20 AM
Page 281
The Software Factory 281
When possible, these tools are designed to be used for more than one project. This
concept is not as difficult as it sounds. For example, a properly designed map editor
can be generalized quite well. After all, a map is a map. In cases where more specialization is needed, it can be defined by providing the information in data files. If true specialization is needed, custom output modules can be written. Some projects may need
genuine throwaway tools, but they should be considered carefully. Maybe they will be
useful for future projects, or even as expansion packs for the project that the tool is
being written for. In general, however, throwaway tools are rarely necessary, and, even
if they are, the coding and documentation standards should remain the same as for the
tools designed for reuse, as shown in Case Study 11.5.
Case Study 11.5 The Benefits Of Tool Reuse
According to Swen Vinke of Larian Studios, the game L.E.D. Wars was written in a
five-month period during the course of another project, The Lady, The Mage,
and The Knight.
These two game styles are fundamentally different (the former being a realtime
strategy game and the latter a multiplayer role-playing game), but they both had a
tile-based map structure. Because of this similarity and some careful framework
design, the two games were able to use a lot of common code. Swen developed an
application framework that could be used across two completely different games.
As a result, he was able to use the same editors to create the levels for both games,
thus saving on the development of separate tools.
Source: Game Developer Magazine (October, 1997)
The Components Group
The components group generates the low-level modules. These are the interfaces to
third-party, platform-specific libraries; common modules such as compression libraries;
and other similar modules.
Initially, the modules developed will be obvious and straightforward, but, as the experience of the teams grows, suggestions for extra functionality that can be moved into common libraries will become commonplace. Looking at Case Study 11.2, it is not hard to
imagine a situation whereby an entirely different graphics engine (for example, Isometric
3D) could be slotted in without too much difficulty just by rendering the graphics again.
13 3634_CH11
11:20 AM
Page 282
282 Chapter 11
The most important design goal for the components is that they should do one task
and do it well. This does not mean that they should be designed to work for only one
specific project. For example, if you have a full-perspective 3D game engine, it should
just provide that service. A 2D map view should be provided by a different module.
Each module will have an identical interface. This way, if at a later date it is decided to
replace the 2D map with a 3D isometric view, then in theory, only the graphics will
need to be re-rendered, and the new library will link straight in.
Another example is the sound engine. Under most circumstances, this module should
not have both 2D (standard pan left/right) and 3D positional functionality. The 3D functionality should be a separate module that interfaces with the 2D library. If necessary, the
2D module can be modified to support the additional functionality required by the 3D
module, such as recognizing different flags passed to the sound object creation methods,
but this should be backward compatible with previous versions of the 2D library.
Remember the overriding design goal of the components is to do one job, and to do
it well. This will make it easier to mix and match components during the project
building phase.
The Project Group
When a project is first started, the project group is tasked with producing prototypes
of the technologies to be used. Beginning a project before prototyping is not advisable.
Unless all the components have been developed by the components group, how can
the project begin?
The first task once a project begins is to ensure that all necessary components have been
developed. This is constant for every project and could mean that, for the first three
months after architecture design, nothing but component development and a small
amount of prototyping is taking place. This situation has to be handled very carefully,
because even the most proactive managers will start to feel a little twitchy if there is no
actual game of some form after this period. The development of a prototype can sometimes allay these fears, but not always. The prototype is also used as a feasibility study.
Do the ideas for the game work? Does it look like it will be fun? Are there likely to be
any technical difficulties at a later date? If so, how will you prepare for them?
13 3634_CH11
11:20 AM
Page 283
The Software Factory 283
After the development of the prototype and components, and assuming that the feasibility studies do not uncover any anomalies that need reworking, the work on the
project can begin in earnest.
The job of the project group is to coordinate the slotting together of the components,
the writing of the glue code, and the provision of art and sound effects from the appropriate groups (see “The Ancillary Groups” later in this chapter). This task is also controlled by the game design group. The project group should ensure that the output
of the art and sound groups is in a suitable format.
The project group coordinates with the testing group to ensure that the project is
always shippable. What does “always shippable” mean? It means that the project
should always be in a buildable, releasable state. Granted, not all functionality will be
there, but it should not crash if the user selects an unimplemented option. The functionality that exists must work properly and be stable. The project is tested constantly,
and any problems are dealt with immediately (with the architecture group being notified if necessary).
The game design group is restricted to interact with the project group through the
architecture group and the defined mechanisms discussed in Chapter 13. This restriction prevents feature creep and ensures an attainable schedule.
The game design group is free to tweak the game, and it is possible through the exposed
soft architecture. Only changes to the actual program architecture itself require interaction via the architecture group. Changes to the actual program should be rare, and so
the project group should be shielded from the wild excesses of the game design group.
The Research Group
The research group is loosely directed by the architecture group. The duties of the
research group (among other things) include investigating and prototyping new techniques to be included in the core libraries for use by other projects.
These research projects should be carried out with exactly the same thoroughness
and attention to detail and procedure that you would find in an important scientific
establishment. Detailed journals must be kept of findings, and these should be
regularly reviewed and checked. It should also be realized that, by its very nature,
13 3634_CH11
11:20 AM
Page 284
284 Chapter 11
it is impossible to plan research around timelines. To quote Dilbert, “It is logically
impossible to schedule for the unknown.”
The purpose of the research group is to make the unknown known, and remove the
single greatest risk from the development paths of the other groups. Because all of the
other groups are using known, documented modules, all open-ended research is
removed from the critical development path. If a project actually requires such
research, then do not commit to the project before the research has reached a successful conclusion and an accompanying module has been produced by the components
group. Either that, or change the project so that it does not depend on the outcome of
the research. For example, use a less powerful graphics engine, and then, if the
research version is finished in time, switch over. Then, if the graphics research proves
fruitless or is not finished in time, you will still have a product to release.
The size of the research group will vary considerably depending upon the requirements
of the other teams. Except in the case of dire emergencies, there should always be at
least one or two members, and, if you find yourself with any spare programmers (after
all other group requirements have been considered), you should assign the spares to
research projects.
Note that research doesn’t necessarily have to include just researching new techniques.
It should also be looked upon as a means of increasing the average skill level of your
programmers. Much research implies that the quality of programmers’ output tends
toward the average skill level within an organization. Improving the average by concentrating on individuals will produce improvements across the board.
For example, a programmer could be given the task of understanding and improving
his knowledge of how some of the core modules function. Or he could be set to read
and work through a technical book on a specific area. Thus, time is not wasted.
Anything that can raise the bar of experience for the company is worthwhile. The
research group is the ideal proving ground for such endeavors.
The Ancillary Groups
The ancillary groups are critical to the success of a project. They comprise sound, art,
testing, marketing, and management and will usually have less work for any one particular project than will any of the other groups. This does not mean that they are any
less important.
13 3634_CH11
11:20 AM
Page 285
The Software Factory 285
The art and sound groups will be providing art and sound or music for the games. This
group is directed by the architecture group, and also receives direct input from the
project and game design groups, relating directly to the project being worked on. This
is mediated by the architecture group if necessary, but it rarely is, unless the art group
is being asked to perform a lot of redundant or nonscheduled work.
The testing group is closely linked to the architecture group, is directly controlled by
them, and reports to them. Every piece of released software—component, tool, or
game—is thoroughly tested, and any obvious defects or faults are submitted via bug
reports to be tracked and fixed. Because of the stringent testing requirements on
releasing software that is built into the development process, the main functions of
this group will be integration, systems, and compatibility testing.
For more information on these functions, see Part III, “Game Architecture,” of this book.
Marketing and management structures already exist in your company. The only reason
they are mentioned here is that they need to be considered when using the software
factory method. Results are measured differently, and the price of an increase in technical visibility and knowledge of your projects is the lack of “wow” factor. There will not
be so many sudden leaps in functionality to amaze and astound upper management, as
this development method uses steady, incremental techniques. The lack of apparent
progress during development can sometimes panic marketing and management personnel who are unaware of the method. Management must be prepared for the differences
between this program and standard game development techniques.
Applying the Software Factory Structure and
You now know everything you need to know about how the software factory is
structured. We guess that by now you’ll be itching to know how to get it going. The
following sections cover the initial stages and the ramping up processes required.
13 3634_CH11
11:20 AM
Page 286
286 Chapter 11
Getting off the Ground
For the purposes of this discussion, assume that you wish to move from a more
traditional development environment with the minimum of disruption—and be able
to back out if things are looking shaky or don’t go according to plan.
The software factory can be incrementally introduced by creating just the architecture
and research groups. You may already have similar groups, particularly a research
What you need first is companywide development standards and guidelines for coding
and documentation. This is mainly common sense. Code and documents are easier to
read and maintain if everybody is whistling to the same tune. This is also the time for
implementing some of the project visibility procedures, but at a limited scope. Provide
an intranet site so people can see how progress is going on the pilot scheme.
See Chapter 13 for more suggestions on achieving consistency.
Once this is done, these groups can begin investigating and prototyping the initial set
of component libraries. To begin, the architects should review the projects currently
under way, or soon to be under way, to decide what functionality would be suitable as
a core component. Concentrate especially on functionality that is yet to be written in
ongoing projects.
After the feasibility study and testing of the prototype components, work should begin
in earnest on core components. Form the components team and begin developing the
components researched by the research team. The code should not be developed from
the prototypes, because prototype code is no basis for a solid code base. Prototype code
is just what it says it is: prototype. Resist the temptation to use it. Start again using the
lessons learned while prototyping to develop some killer components.
On completion of the initial set of components, begin leveraging them into the ongoing projects to implement unwritten functionality. It is not wise to try to replace functionality that is already written, as it will probably involve code rewrites and cause
delays to existing schedules. If, on the other hand, an ongoing project has yet to implement a CD-playing code, then producing a fully documented and tested CD-playing
13 3634_CH11
11:20 AM
Page 287
The Software Factory 287
module has got to be a good thing. Make sure the team understands that all support
requests go to the architecture group, and not to the programmers in the components
group. This shouldn’t be a problem anyway, because the components group won’t make
any ad hoc modifications without going to the architecture group for confirmation.
Think and think again. Think small at first, and think on your feet. Start by providing components
that are going to be immediately useful. Once project teams start coming around to your way of
thinking, work can start on the more fundamental and forward-thinking components for subsequent
As personnel become available, draft them into the tools group and start work on the
tools to support the more adventurous components.
Now you are halfway there, and the rest of the process follows naturally, as projects
wind up and more developers become available for the other teams, which are incorporated into the communication framework. Also, by now it will have become clear
whether the software factory will work for your organization, or whether another
approach or combination of approaches would be more suitable.
Knowing When to Use Each Team—a Parallel Development
Each programming group is assembled on a just-in-time basis. This ensures maximum
flexibility and makes efficient use of resources.
It is very important to make sure that each group has the correct number of members
to support the needs of the dependent teams. These numbers will need to be tweaked
according to the needs of your organization. Figure 11.4 shows the group dependencies. If any of the supporting structures are weak, the whole structure will collapse. Be
aware of the strengths and weaknesses of each group, and avoid overburdening groups
by making the structure top heavy. This will only cause trouble further down the line.
13 3634_CH11
11:20 AM
Page 288
288 Chapter 11
Game Design
Figure 11.4 The group dependency structure.
So how do these groups work together? Figure 11.5 shows a sample workload against
time for a small to medium-size organization.
Game Design
Start Date
Ship Date 1
Ship Date 2
Ship Date 3
Figure 11.5 A graph of average group workload against time.
When a group has less work, personnel can be moved to groups that require extra staff
to encourage knowledge transfer. One of the stranger findings of research into this sort
of group dynamics is that the information doesn’t always spread as well as one might
expect. This only usually becomes an issue when a member is added late in a project,
but, for larger groups, the team identity should still be intact. Each member should be
13 3634_CH11
11:20 AM
Page 289
The Software Factory 289
made to realize that they are all part of one team. The logical divisions are in place
only to make sure that project visibility is maintained to the highest possible level.
For larger organizations, it is possible—and indeed sometimes necessary—to run
projects side by side, As long as the supporting infrastructure is strong enough, this
will cause no additional problems.
Rotating and Reassigning Team Members
One of the strengths of the software factory is that you can spread knowledge among
the team members. This is not without its risks, but most of you will already have
experienced the greater risks of not doing so, as shown in Case Study 11.6.
Case Study 11.6 The Indispensables
There was a developer who had an “unusual” naming convention for variables. He
would start with the letter a and work his way to z. Then he would begin again at
aa, and work through to az, and so on. The amazing thing was that he remembered
what they were all for. Even without any comments in his code, he could come
back to it months later and modify it without difficulty. When he left the company,
all his code was thrown away and rewritten from scratch. No one had realized that
he coded like this until it was too late. They had only been interested in the results.
By rotating team members, the kind of problem described in Case Study 11.6 can be
avoided before it becomes a major risk. Regularly rotate members around the different
core programming groups but pick your time carefully—adding a new member
towards the end of a project tends to slow things down.
Although this sounds dangerous, it ensures that more than one person has knowledge
of any particular area of code, and, because of the documentation procedures, the
lead-in time for new developers is minimized. What would you rather have: slightly
increased development times, or the risk of not being able to complete a project if a
key member leaves? Spread the risk wherever you can.
13 3634_CH11
11:20 AM
Page 290
290 Chapter 11
The Suitability of a Software Factory
The software factory is suitable for a medium- to large-scale organization. In this
example, you need five or more programmers and accompanying technical staff so
that teams can be allowed the partitioned in an even and balanced way, and members
rotated to good effect
The software factory also scales down easily. You will lose some of the parallelism and
not be able to have separate ancillary teams, but the advantages for the second and
subsequent projects will still be evident. This method is the technique that you can use
for developing software and has been honed out of five years of solid C++ development
experience for various clients around Europe in situations where speed of development
and performance was of great importance. Whenever possible, use software factory
construction techniques, and you will be pleased at the consistent results, even if the
developers have to get used to being more restricted, and managers have to stop
expecting early results and to fix the bugs after the shipping date.
Smaller Teams
However, there comes a point where the overheads of implementing the software
factory outweigh the benefits.
For small teams of around two or three developers, much has been written of the
benefits of Extreme Programming (XP).
Personally, we feel that the benefits of XP have been overtaken by the hype, but there is
no disputing the wisdom of the basic ideas: test early, test often, and have someone
looking over your shoulder while you code.
In combination with some of the other techniques mentioned in this chapter,
XP can be a win for small teams. For more information on XP, visit
13 3634_CH11
11:20 AM
Page 291
The Software Factory 291
The Final Word
The methodology in this chapter is not a universal panacea. There is no silver bullet in
software development. If you have fundamental problems with your development
teams, such as a lack of skill and experience, no amount of methodologies will help.
Don’t feel that you have to take this methodology as the letter of the law, Try it out.
Experiment with it. See what works for you and what doesn’t. Keep the bits that do
work and toss the bits that don’t. With determination—and a lot of effort—the benefits
will pay off.
Many other methodologies and techniques are suitable for games design, most of them
(including this one) lifted straight from more traditional programming endeavors. This
is a system that has been proven to provide quality software, time and time again over
a series of staged releases, with each subsequent product cycle converging towards a
smaller and smaller release interval.
13 3634_CH11
11:20 AM
Page 292
14 3634_CH12
11:21 AM
Page 293
Chapter 12
• Scheduling of projects
Milestones and Deadlines
• Avoiding arbitrary
• Defuzzification of
n Chapter 11,“The Software Factory,” we discussed the software factory method. In this chapter, we will examine how
to set this in motion, Keep in mind that the ideas presented
here are not restricted to just the software factory, but can be
used in any project and team structure.
However much we’d all like to write software without
constraints and pressures, some milestones and deadlines are
a necessity. They are, indeed, a common factor for all nontrivial software projects and, as such, are the bane of many
developers’ lives.
Given the universal nature of milestones and deadlines, it is
surprising how most milestone specifications are set in an
arbitrary fashion, and how, in many cases, deadlines are
defined on a “wish list” basis rather than realistic estimates.
This peculiarity can be attributed to a number of factors
such as marketing influence, inexperienced software managers, and, in some cases, sheer stubbornness. Most people
simply do not know how to estimate schedules for software
development. It is a difficult procedure, and it involves much
analysis and calculation.
Outside the games industry, software companies use a slightly
more rigorous approach than “pick a date and hope for the
best.” Even so, precious little training and guidelines are
available on how to define milestones in a sensible manner.
• Starting the project
14 3634_CH12
11:21 AM
Page 294
294 Chapter 12
Milestones divide projects into manageable chunks. The alternative is a huge list of
desired features that the developers check off as each is implemented. This, of course,
is not recommended. Can you imagine how demoralizing it would be, in a project that
lasted 18 months, to check off only a few items every week? How would the team
maintain focus and enthusiasm? As ridiculous as it seems, some projects are still developed this way, and they are most likely one of the most joyless experiences in software
The method presented here is applicable to all types of software, whether spreadsheets
or shoot-’em-up arcade games. This chapter presents a set of guidelines based on these
“good practice” techniques for defining milestones and deadlines in a more realistic
fashion. Later in the book, we’ll explain how these estimates are progressively refined
to become increasingly accurate as the project continues. This phenomenon of increasing accuracy is referred to as the convergence rule. As well as discussing the overall
structure of a project, this chapter also serves as a road map for the chapters ahead.
So, this chapter is your first port of call if you are searching for information on the
development process.
How Milestones Currently Work
How many times have you had to work to a “fuzzy” milestone? For the uninitiated, a
fuzzy milestone is one that has many interpretations and x possible ways to achieve it,
where x is any number you like.
And how do you know if you have really reached a fuzzy milestone? If the milestone
goals are fuzzy, they are open to interpretation. Whose interpretation counts? If it’s
your boss who makes the “right” interpretation, the chances are it won’t be the same
as yours.
Try this simple test. Imagine that your boss shows you a piece of paper with five dots
arranged in a pentagon and labeled A, B, C, D, and E, as shown in Figure 12.1.
14 3634_CH12
11:21 AM
Page 295
Milestones and Deadlines 295
Figure 12.1 The task.
Now imagine he says, “I’m very busy today. I’ve got lots of people to see, so I can only
explain briefly what I need you to do. I’d like you to draw a line between A and B. I’ll
be back in an hour.”
Three hours later, your boss staggers back and demands his completed document.
If you’ve drawn a straight line directly between A and B (as in Figure 12.2), then I’m
sorry, but you lose. If, however, you have drawn a line between A and E, and have also
made sure that it goes through C, D, and B to form the outline of the pentagon (as in
Figure 12.3), while leaving the space between A and B blank, then you have hit the
milestone correctly.
Figure 12.2 The wrong solution.
Figure 12.3 The right solution.
14 3634_CH12
11:21 AM
Page 296
296 Chapter 12
Okay, so your boss wouldn’t give you a task as trivial as this. The point is that a vague
milestone is a useless one.
To arrive at your solution, it could be that you have worked too much or too little. In
cases of the former, it is also usually the wrong sort of work, meaning that the programmer has overengineered the solution, providing unneeded or unnecessarily complicated functionality.
Ideally, milestones should not give this much room to maneuver, but this sort of situation is unfortunately at the heart of most scheduling problems. On a fair proportion of
projects, the milestones were spaced too far apart, and thus allowed a fair degree of
drift, during which unnecessary work was done due to a lack of direction and focus.
A milestone should be clear and concise. It should also be obvious at a glance whether
it has been achieved or not: black and white, failure or success. It should not accommodate or encourage any ambiguous results or shades of gray. No one should be able
to say a milestone is 90% complete: a milestone that is 90% complete makes as much
sense as a light bulb that is 90% on. This point is critical. The fuzzy milestone is the
single most common problem with projects.
Your boss’s pentagon-dot task should have been specified in the following manner:
1. Prerequisite tools: ruler, and black pen capable of producing consistent thin lines.
2. Connect point A to point E using a thin, straight, black line.
3. Connect point E to point D using a thin, straight, black line.
4. Connect point D to point C using a thin, straight, black line.
5. Connect point C to point B using a thin, straight, black line.
Note: Do not connect points A and B.
This list is comparable to a detailed analysis document. The analysis should be detailed
enough to leave no room for doubt as to how the problem will be solved. In this case,
there is one optimum solution, and producing that solution is just a case of connecting
the dots (pun intended). You may feel that this takes the creativity out of programming, to which we respond that programming is the wrong place for creativity.
14 3634_CH12
11:21 AM
Page 297
Milestones and Deadlines 297
Another word for “creative programming” is hacking. The time and place for creativity
is in the design and architecture. By the time we get around to actually programming
the system, the game and architectural designers have finished all the creative parts.
This may dent egos, but the job of a programmer is that of simple translation: Take a
detailed design and turn it into a working computer implementation.
The actual programming of a design is really just the last small stage in a long and
complicated process. It’s the finishing touch, a small but important detail. However,
programmers usually have more responsibilities than just writing code. In most cases,
they are also responsible for the detailed module design as well as the implementation.
This allows the creativity to be shifted to where it belongs, into the documented code
design, and not directly into the code.
All the good projects we have worked on have deliberately reduced the programming
to the level of a simple translation. This is the way to make sure that a project runs
smoothly and is completed on time and within budget. All the bad projects we have
worked on have gone straight from overall architecture design to coding. Without
exception, they have overrun their budgets and were completed late…if at all.
As shown in Case Study 12.1, fuzzy milestones cause delays to the project, and sow
confusion and distrust among the parties waiting on those milestones. By not specifying an exact point at which to stop and take stock, you decrease project visibility,
which means no one is exactly sure where the project is. Delays and problems become
endemic. No one intentionally sets out to create a situation such as this. It just happens, unnoticed, one day at a time until it becomes too great to ignore. Statistically,
most projects that lose time to slippages are unable to make it up: slippage begets
slippage. Such problems need to be caught early.
Case Study 12.1 What Fuzzy Milestones Can Do to
a Project
For one of the very first game projects Andrew worked on professionally, fresh out of
college, he was given what should have been a very simple assignment: Write a level
editor using Microsoft’s Visual Basic 4.
This is fine for a project title or mission statement, but unfortunately it also became
Andrew’s project milestone.
14 3634_CH12
11:21 AM
Page 298
298 Chapter 12
It is difficult to see how you make a milestone any fuzzier than that. At the time,
however, he was inexperienced enough not to see any problems with this.
He was given a deadline of two weeks. Having a good knowledge of the previous
version of Visual Basic, he got cracking.
The first obstacle was a lack of knowledge of the problem domain. Andrew didn’t
know exactly what was wanted, even though he and his boss thought that they both
had the same end product in mind.
The project that Andrew was thinking of (and that he knew he could produce in
two weeks) was a tile-based landscape editor that was capable of generating tessellating tiles and a height map for these tiles.
The project Andrew’s boss had in mind required a landscape editor with the above
capabilities, as well being able to automatically generate landscapes both randomly
and from an imported height map, allowing the import of data from VistaPro (a
commercial landscape-generation package), placement of game objects on the landscape (including autogeneration of formations if groups were placed), automatic
import and categorization of the output of the artists, and generation of a complete
set of level files with an optimized palette for each level.
Both of the above descriptions are descriptions of level editors, but only one is
possible within two weeks, and only one (in the boss’s opinion) is the right one.
Unfortunately, they aren’t both the same one.
Consequently, after two weeks, Andrew had a working landscape editor according
to the first specification, hence hitting the project target (as he saw it). He had
given no thought to any of the other issues. And given the tightness of the deadline,
he had not written the program to take any of these things into account, nor had he
even provided for the option of inserting them at a later date. (In other words, he
had used the “code like hell” approach.)
The end result was a further ten weeks of development in order to hit the milestone,
and a program that ultimately ran intolerably slowly because it was trying to do
some very processor-intensive palette calculations using VB4.
14 3634_CH12
11:21 AM
Page 299
Milestones and Deadlines 299
Fuzzy Milestones
Just as mountaineers are warned not to eat yellow snow, you have been warned not to
tolerate fuzzy milestones. Why try to work towards a goal within a fixed deadline
when you are not even sure where the goal is, and there is likely to be an argument
about the position of the goal posts when you get there? Fuzziness is the enemy of
software scheduling.
Fuzzy milestones are also the result of completing milestones and schedules before fully
defining the specifications of the project. Because not enough is known about the problem domain, milestones are defined with sweeping generalities, and deadlines are divided fairly arbitrarily across the available time using a gut instinct for the difficulty of any
particular milestone. Usually too much work is covered by a single milestone, and it is
simply too difficult to complete all the separate tasks to hit the milestone. It then
becomes a fuzzy milestone, and can be legitimately described as being, for example, half
completed (if it depends on two streams of work and only one has been finished).
Cramming too much functionality into a single milestone is counterproductive. Apart
from the obvious confusion factor, it can cause other, more subtle problems. The team
should get a sense of achievement from completing a milestone. A milestone that drags
on for an extended period has a detrimental effect on morale, and the attitude of the
team transforms into that of trench warfare: us versus them, in a daily slow grind to
advance the lines. A team should know at all times how much progress has been made
on the project to the nearest percentile.
This is not an impossible task, but it does require more planning than is usually
associated with the average game project.
Milestones and Mini-Milestones
As a rule, a good length of time for a milestone is between four and eight weeks. This
provides a finite goal and allows the milestone to be reasonably focused. Another problem with large fuzzy milestones is lack of focus; who cares about a deadline that is four
months away? This situation normally degenerates into three months of laid-back
programming, followed by a month of coding like hell.
14 3634_CH12
11:21 AM
Page 300
300 Chapter 12
The solution (apart from spacing milestones closer together) is to divide each milestone into mini-milestones, or checkpoints, each lasting approximately a week.
Obviously, to be able to do this, you have to have specified the architecture fairly well.
If you find yourself in a situation where you cannot set fairly concrete milestones and
mini-milestones, the architecture design is not complete enough to begin coding. In
this way, we can help ensure that good practices are followed during the critical period
of project startup. If things start well, it is easier to keep them going well than it is
repair a situation that had started badly. A good way to divide tasks between milestones
and mini-milestones is to specify module completion as mini-milestones, and module
integration as the main milestone. (It may sometimes be necessary to divide the minimilestones even further, but this will depend on the team and the task at hand.)
With a good team, mini-milestones are not always necessary, for instance, if the team is
experienced enough to know the best way to tackle problems and mature enough to
stay focused on an approaching deadline. However, for new teams or for new and
unfamiliar projects, mini-milestones can serve two purposes:
Keeping the team focused on the task at hand
Providing increased levels of project visibility during critical times
The first of these is important to prevent the aimless drifting of projects that can occur
when milestones are too widely spaced. A good analogy is to imagine that you have
been given written directions to a place you have never visited, but you have a rough
idea of the general area. You have been asked to draw a map of the route on your way,
and to arrive within a certain time limit. As you walk, you spend most of the time drawing your map. Every now and then, you stop and look around to get your bearings.
In this analogy, the milestones are the instances when you stop and check where you
are, and the project is the map. The more often you stop and look, the less likely you
are to make as many wrong turns or mistakes on your map. Obviously, this method is
subject to the Law of Diminishing Returns: If you stop every two steps, not only will
it take longer to arrive, but you won’t be able to concentrate on drawing the map.
Finding the balance is the key. The milestones should not be so frequent that they
detract from the amount of useful work done, but not so far apart that the impetus,
momentum, and focus of the project are lost.
14 3634_CH12
11:21 AM
Page 301
Milestones and Deadlines 301
A good rule of thumb is to scale the amount of mini-milestones to the level of experience of the team and the familiarity of the subject matter. The absolute minimum spacing of mini-milestones should be one to two days. Any less than this and the amount
of administration required would outweigh the benefits.
When to Use Milestones
In general, every program or module is a separate project, and, for all but the most
trivial programs, they should be treated as such and specified, written, and documented accordingly. What exactly makes a program trivial is open to debate, but the
recommended guideline is to consider writing a throwaway program only if you can
categorically state that there are no other feasible uses, and that developing it in a
more general way would generate a ridiculously large project.
An example of a program that fits these criteria is a look-up table generator that is
used to create files of pregenerated values to reduce the amount of calculation needed
at runtime. (With the advent of faster and faster processors, look-up tables have begun
to fall out of favor, but they still have their uses on more-limited systems.) To develop
this in a general way would involve writing an expression parser that could read userdefined formulas to produce the tables, In this particular case, it is clear that it would
be far better to hard code the required tables.
Making Your Milestones Accurate
In order to improve the accuracy of schedules, we need to remove as many of the
uncertainties as possible. Not all can be removed, but one way to minimize the number
of uncertainties is to concentrate on avoiding fuzziness in milestones.
In Case Study 12.1, the problem was caused by a mutual misunderstanding and the
lack of an in-depth explanation as to what the requirements were. No specification
document was followed because no analysis had been done of the problem domain.
Given the tightness of the deadline, it is clear that no thought had been given to gathering the requirements. The boss had just assumed that the knowledge of what he had
wanted would be transferred from his head to his team’s.
14 3634_CH12
11:21 AM
Page 302
302 Chapter 12
This clearly doesn’t make sense, but you would be astounded at the number of projects
that are run on this basis. Vast sums of money are spent on back-of-the-envelope
estimates that do not take into account all the possible factors that could affect the
outcome of the project. In some cases, deadlines are set on the basis of demands made
by the marketing department, which is probably the least qualified group to make
such decisions.
Marketing pressure is obviously important to the release date, but it should never be
the deciding factor in scheduling issues. Trying to get teams to respect marketingdefined deadlines is an impossible task, particularly when the team is fully aware that
these deadlines are arbitrary and unrealistic, and do not take into account the technical
factors involved in the development. If there is a marketing-defined deadline (the infamous Christmas rush), then scale the functionality of the project back to meet the
deadline. This reduced functionality must be made completely clear to all concerned.
The usual response given in such situations is to ask for time to be shaved off the
schedule but for the same amount of work to be done anyway. (The assumption here is
that all developers pad the schedule to make their life easier. This is rubbish, and if
you succumb to this argument then you are just asking for trouble.) It must be made
perfectly clear that the schedule is the optimum time for producing the requested
amount of work. If the work must be done in a shorter time, functionality must be
reduced. We have seen too many situations in which project leads have accepted such
situations and assumed that they can accomplish 18 months of work in 12. Admittedly,
this is just about possible, given enough overtime. Trying to get a team of developers
to work solidly for 12 months with tons of overtime is not an easy task and most certainly not an advisable one. The damage to morale would be very high, and you would
stand a good chance of losing most of your skilled developers at the end of the project.
The deadline for the program in Case Study 12.1 was clearly too short under any circumstances. The requirements gathering for this project would have taken at least a
week, and, if requirements had been gathered, it would have predicted that a further
six weeks would be needed. Of course, even this estimate depends on the progress
made by other team members in this period and their availability to document and
explain various areas that the editor needed to take into account. This may or may not
have been acceptable to the boss, but at least this knowledge would have allowed him
to pursue other options, such as assigning more people to the project, reducing its
scope, extending the schedule, or giving the project to someone with experience in
that area.
14 3634_CH12
11:21 AM
Page 303
Milestones and Deadlines 303
Before you can even consider planning and scheduling an entire project, certain steps
must be performed. The first three milestones of a project (shown in Table 12.1)
should cover these steps. These will allow a more detailed and accurate analysis of how
the time will be spent over the months to come.
These milestones effectively encompass the initial stages of specification and feasibility
studies. The cost of these stages is a small part of the project’s total cost, and it helps to
provide good information on the total cost, duration, and technical feasibility of the
overall project.
Table 12.1 The First Three Milestones
General requirements gathering
Technological requirements gathering
Resource requirements gathering
General feasibility study
Technological feasibility study
Resource availability study
Draft architecture specification
Project initialization
At each of these first milestones, the decision makers are in a good position to cancel
the project, or allow it to continue to the next milestone. After the third milestone
has been completed and the project is given the go-ahead, then the project should be
allowed to run to completion (unless it falls into dire straits). The cancellation of a
project after a successful feasibility study should be treated as a failure of the feasibility
study process and investigated accordingly.
The initial investigative phases of a project are obviously a better place to make this
decision than 15 months into development, when a lot more money and effort has
been spent and everybody has begun to assume that the project will run to completion.
(Does this situation ring any bells?) The effects on morale of canceling a project
increase with the length of time that the project has been in progress. If, however,
it is clear to the team members that they are performing feasibility studies to check
out the technology and playability requirements, the damage is minimized: everybody
understands that all they are doing is investigating whether the project is worth
14 3634_CH12
11:21 AM
Page 304
304 Chapter 12
attempting. This is most effective if the final go-ahead can be relied upon totally. It
loses its potency if, even after all the feasibility checking measures have been taken,
there is still the risk of random cancellation.
People naturally work better if they feel comfortable and secure in their position. Of
course, the risk of a project’s premature ending cannot be eliminated entirely, but it can at
least be minimized by solving any major issues as early in the process as possible, before a
lot of money has been spent and a lot of people have bought into the project. Not only
will you lose money and time, you will also lose the respect of the people involved.
In Case Study 12.2, the projects were canceled at different stages in their development.
Fortunately, some of these were still at the prototype stage and therefore had not gone
too far. Even so, an average of $450,000 would be a lot of money to spend on a prototype at 1997 prices. In these cases, the “prototypes” were not really prototypes in the
classic sense, but were merely partially developed versions of the full game. In most
cases, the games were being developed as normal, and no effort was being made to
separate the prototype (such as it was) from the main development. In these cases, the
prototype code would be the actual game code.
Case Study 12.2 The Costs of Canceling Projects
One large games company announced in 1997 that they had recently canned almost
a dozen projects that weren’t measuring up to expectations over a six-month period.
The cost of canceling these projects varied between $150,000 and $750,000 per
project. If you consider that this money could have been used to completely fund up
to four other projects (if the early prototyping stages had been performed properly),
the value of proper planning and prototyping becomes obvious. The wasted effort
was not only expensive in terms of money, but also adversely affected the morale of
the development teams.
However, a prototype should be just that: quick and disposable. It is a mechanism to
test ideas and concepts, so don’t confuse it with technology research. The purpose of
the prototyping stage is to investigate the feasibility of the game’s concept. By the end of
this stage, the required technology should already be developed (or substitutes should
be available). This is not as restrictive as it sounds because these components will have
14 3634_CH12
11:21 AM
Page 305
Milestones and Deadlines 305
been developed to standardized interfaces. If (or when) a new technology is developed
during the actual course of the project, the new component can always be slotted in.
The prototype doesn’t even need to be written in a compiled language. Often a series of
successive prototypes—starting with pencil and paper and progressing to a high-level,
rapid-development language such as Basic—are enough to be able to test the rough
After all, it’s a prototype and not a finished product, and, if it is written in a different
language than the final product, it removes all possibility and temptation to use the
prototype as a code base for the final product.
Using prototype code at production level is a recipe for disaster, because prototypes
usually, by definition, do not feature production-level code. This does not mean that the
code is not well written; it just means that the coding priorities are different for a prototype. For example, speed is not a priority in prototype code, but it is for some production
code. Don’t underestimate this temptation to use prototype code for the main product.
The prospect of saving time by using already written prototype code is a dangerous trap
and usually results in significant delays and problems further down the line. It is also not
worth attempting to write prototype code to production level, which would be an impossible task and would ultimately impede the development of the prototype.
The milestone schedule presented in Table 12.1 attempts to encompass all the above
points and covers the initial stages of a project up to the conclusion of the feasibility
The following sections cover each of these initial milestones in more detail.
Checkpoint 1.0 General Requirements Gathering
This checkpoint answers the following questions:
What Is the Project?
Obviously, it’s a game, game component, or a tool to facilitate the production of a
game. But what type of game? What is the target audience? Why will it appeal to
them? Where and how does the game fit into the company’s overall vision? The
answers to these questions help to provide a larger focus for the project.
14 3634_CH12
11:21 AM
Page 306
306 Chapter 12
There should be no doubts as to what type of game is being written. For example,
the look and feel should be explored, as should all the other issues and concerns
that affect the project at its most general level. We personally abhor the term mission
statement, but if ever any of our projects have them, this is where they need to be
What Functionality Do the “Customers” Expect?
In this case, customer can be defined at several levels. The first customer is the company management, the second is the publisher, the third is the press, and last—but most
certainly not least—is the buying public.
Each group of these customers must be individually catered to. Unfortunately, not all
of them have the same needs and desires. Each is arguably of equal importance, so
the task of pleasing them all becomes a precarious balancing act. For example, the
management and the publishers may require regular demonstrations and interim
releases for shows and previews. The press will need preview versions, and the
buying public needs a finished product! These needs all must be anticipated and
built into the schedule (especially the last one).
What Is the Minimum Functionality Required?
The minimum functionality, surprisingly enough, is the most important consideration.
While we do not want to dwell on gloomy thoughts, it must be realized that development, like true love, never runs as smoothly as expected. From the earliest stages in
the project, several tiers of functionality must be defined. (These are introduced in
Chapter 14, “Troubleshooting” and discussed in more detail in Part III, “Game
The lowest functional tier defines the minimum release requirements: the point at
which the software is “complete enough” to be released publicly. The game may not
be very good at this point, but at least the core is functional. Reaching this point is
the primary goal of the schedule. Once this point has been successfully achieved, additional functionality is added tier by tier until the ideal level of functionality is reached
and the product can be released. Of course, if the project has experienced any delays,
it may be necessary to release the game before all the desired features have been
implemented. In this situation, you will be glad that you took the tiered approach.
14 3634_CH12
11:21 AM
Page 307
Milestones and Deadlines 307
In this stage, you’ll gather all the overall project requirements, which will serve as the
general guidelines for the entire project.
The prerequisite for this stage is the game treatment document, as defined in Part I.
To recap, this is a small document that describes the essential feel of the game and the
overall idea behind it.
From this initial treatment, you’ll develop the game design document, as described in
Part I. Once the game design is fully specified (subject of course to the 80/20 rule), and
has been approved by the powers that be, we are ready to proceed to the next phase.
Checkpoint 1.1 Technological Requirements Gathering
This checkpoint answers the following question:
What Techniques and Processes Are Needed?
The ideal situation is to have all the components already developed. This question then
becomes much easier to answer. If the components aren’t yet available, can they be
easily constructed to a fixed timescale, or is research needed in order to define the
functionality required?
If research is needed, then the project cannot proceed past the prototype stage. It is
very important that you then consider putting it on hold until the research is complete.
As soon as you break this rule, you are assuming an unacceptable risk. Remember, the
object of these development processes is to reduce the ever-present risks as much as
possible. Relying on research to a fixed timescale undermines every other measure
taken. Don’t do it.
If you are sure that the technology to be researched is feasible (and you have a fallback position), then you may wish to continue developing a prototype. An example of
a fall-back situation would be to have a software-based 3D engine that you would like
to extend to allow hardware support. Even if the hardware support research fails, you
can still use the software engine, so failure, although inconvenient, in this case is not
fatal. However, you should proceed with caution.
14 3634_CH12
11:21 AM
Page 308
308 Chapter 12
Checkpoint 1.2 Resource Requirements Gathering
This checkpoint answers the following question:
What Resources Will Be Needed for the Project?
Prepare a list of all the software and literature that you don’t already have that will be
needed to complete the project.
Don’t go overboard on this; it’s easy to buy everything including the kitchen sink. You
may have a limited budget, and it’s Murphy’s Law that, as soon as you deplete it, an
essential piece of software is released and you can’t purchase it. Here is a list of recommended useful items (and examples of them):
Compiler (Microsoft Visual C++)
Automated error-checking software (NuMega BoundsChecker)
Profiler (NuMega TrueTime)
Source-control software (Microsoft Visual SourceSafe)
Scheduling software (Microsoft Project)
Word Processor (Microsoft Word)
3D modeler (Kinetix 3DS Max)
2D paint software (Adobe Photoshop)
Format translation software (Okino Polytrans)
(The contents of this list are obviously just suggestions based on the software we are
most familiar with. It is similar to that used in most development houses. Some may be
a tad expensive, especially for those on a limited budget. Fortunately, there are always
other options.
Would you like some useful advice, hard won from bitter experience? Never use the latest version
of software until it has been out for a while. If you switch versions halfway through a project, it can
cause delays. Also, if things do not work out well, “rolling back” the change can be a time-consuming and expensive task. They don’t call it the bleeding edge for nothing.
14 3634_CH12
11:21 AM
Page 309
Milestones and Deadlines 309
Checkpoint 2.0 General Feasibility Study
This checkpoint answers the following questions:
Is It a Worthwhile Project?
The gameplay and story should be investigated and polished. Obviously, this will not
be final, but subject to tweaks and modifications as the development progresses. This is
where old ideas are discarded and new ideas discovered.
A word of warning here: Beware of feature creep. Many games are delayed or canceled
due to underestimating the ease of adding a new feature. Feature creep—and preventive measures—are discussed in Part III.
Is the Market Ready for a Game Such as This?
Does the game fit into the current market? This does not mean that the game should
be a derivative of one that is already available. (If this is the case, then you need to
rethink the project. Enough companies are producing uninspired and lifeless games
without you adding another. Go back and read Part I until you feel more inspired!)
Okay, assuming you are still here, then your game design has passed the first acid test.
The next question is whether the proposed design is completely new or whether it
adds and expands to a current genre. If it is completely new, you could carry out a
study on how the market is likely to accept it. Present it in the form of use cases to a
group of unbiased people and listen to their impressions of the game concept. Ask
them if it is the sort of game they would play, and then ask them what they like and
what they don’t like. It is important to get truthful answers, which rules out family,
friends, and colleagues. A common piece of advice is to design a game that you would
like to play. There is nothing wrong with that, but you run the strong risk of ending up
with a game that only you like playing. Of course, you must design a game that you
would enjoy playing, but this cannot be your sole guideline.
Originality poses a dilemma in the entertainment industries. The tension is perceived
to be between art (creating for yourself) and craft (creating for a customer). This in
particular becomes an issue when an industry is moving from being a niche or hobby
interest (where it is legitimate to assume that most consumers are people like yourself)
to being mainstream (where your consumers may have very different tastes from you).
14 3634_CH12
11:21 AM
Page 310
310 Chapter 12
But is there really a difference? Shakespeare, da Vinci, Bach, Dickens, Dylan, Raimi,
Fincher—these are all strikingly original artists who found a market. Will Wright spent
two years trying to explain to game industry executives why The Sims would be the
fantastic success that it has proven. In the end, the only way to convince them was to
make the game. Market research would have been of little use in cases like that, for
how could a player say in advance whether they would enjoy a kind of entertainment
software product they had never experienced before?
An artist is someone who creates something new out of his or her imagination.
However, contrary to widespread belief, a craftsman is not a person who creates something to a brief supplied by his patron—whether the patron be Cesare Borgia or a publisher’s marketing department. The term for that is hack. A craftsman, in fact, is one
who creates art that will appeal to a market. If Cesare or the marketing director had any
idea what that was, they would design it themselves. If the market knew what they
wanted, they’d be able to ask for it. The job of the craftsman—whether writer, painter,
film-maker or game designer—is to imagine a new thing that the market will want when
they see it. The craftsman does this by using his or her creative genius to think as the
intended customer does. You could call it the “method acting” approach to creativity.
This is, of course, completely different from catering to the market’s requirements in an
industry like pharmaceuticals, for example. There, you can question the public and get
a list of features that they want in a new product. But for goodness’ sake never try creating entertainment that way—the result would be abysmal! Your company’s marketing
department should be providing you, as designer, with information about the market.
This will help you to think yourself into your archetypal customer’s head. The marketing department should not, however, be providing you with a list of features they think
would be good in the game. That’s the designer’s job.
If the game is an expansion of a currently existing genre, then the designer’s job is a
little easier, but the above advice still holds. Try to define exactly how your game will
improve the genre. It can’t just be more of the same with a list of extra features; it
needs to advance the genre somehow. A classic example is Valve’s Half-life, a firstperson shoot-’em-up built atop the Quake II engine. It was the best FPS game released
in 1998 even though there were other contenders released at roughly the same time
(including Sin and Unreal). Half-life considerably added to the genre by including a
strong story, incredible adversary AI, and a very carefully constructed playing environment. In Half-life, there were usually two solutions to every problem. You could go in
14 3634_CH12
11:21 AM
Page 311
Milestones and Deadlines 311
with guns blazing and kill everything in sight, or you could think about your next
step. This is an example of excellent gameplay. (Admittedly the thinking element
seems to diminish towards the end of the game, relying more on fast reflexes and
a high ammo count. But no game can ever be perfect!)
When you consider releasing a game in an already existing genre, try to improve upon
the existing content, rather than following the more conventional “more weapons, more
enemies, bigger explosions” route. For examples of how not to expand an already existing genre, take a look at the realtime strategy genre. As of December 1998, it was already
stagnating, with only a few gems (such as Blizzard’s StarCraft) adding something new.
Use your imagination. Try to exploit every possible opportunity to gather opinions and
ideas for the design. The general feasibility study is the last point in the development
process where major restructuring of the game design can be done cheaply and efficiently, so make the most of it. Don’t be afraid to scrap a design if it doesn’t make the
grade. This may be difficult due to personal attachment and internal politics, but it is
better to admit failure privately at this stage than to release a poor game and damage
the reputation of your company.
Checkpoint 2.1 Technological Feasibility Study
This checkpoint answers the following questions:
What Are the Performance Requirements?
Is the technology required for the game feasible for the configurations and machines of
average consumers?
Game developers are lucky. They usually have access to the best equipment, the fastest
machines, and the latest hardware. Games really fly when presented on such hardware.
The end user, of course, is not always so lucky. Research indicates that consumers buy
a completely new PC at most once every year. (This is usually the gamehead who
always requires the latest and best hardware.) The average consumer probably replaces
their PC once every two to three years, making do with upgrades in the meantime.
When you target a base machine, you should bear in mind that there is a two- to threeyear spread of technology. To maximize potential revenues, you should aim for the center
of this spread as the base configuration. If you specify a higher minimum specification,
14 3634_CH12
11:21 AM
Page 312
312 Chapter 12
then you should have a good reason (for example, you are id Software, and are just about
to release the Trinity engine), or accept the potential loss in sales. Only the most groundbreaking of games can push the envelope and set a new standard base configuration. In
general, for the average development timescale (18 months), the top machines of today
will be slightly above the average specification by the time the product is released.
Checkpoint 2.2 Resource Availability Study
Checkpoint 1.2 defines the software and literature required for the project. The purpose
of checkpoint 2.2 is to consolidate the results of the earlier checkpoint and to choose
the initial members of the project group (as defined in Chapter 11). If possible, the project group should be selected from personnel who are most skilled in the type of project
being written. The size of the team depends on the size of the project. A group of two or
three may be ideal for a small project, but a larger project will require a larger group.
Although this seems obvious, assigning a group to a project is always trickier than it
appears. Too large a group and you risk inefficiencies and extra expense; too small and
you risk overworking the team members and causing delays. The trick is to try to get
the balance right the first time. Changing things later on is usually more difficult and
always more expensive. (Chapter 21, “Development,” discusses this in more detail.)
If it was decided in checkpoint 1.1 that research would be needed before the project
can be started, then this is the point where you need to decide whether or not to go
ahead with the research project. If not, then this is the end of the project. Tough cookies. Look on the bright side—it was most likely ahead of its time anyway (where are
holographic displays when you need them?), so maybe it can wait for another day.
If it is decided to proceed with the research, then the project spawns one or more
research projects. Checkpoints 3.0 and 3.1 (following) still apply, but from there on in
we are waiting for the results from the research project(s). (These are discussed in
Chapter 11 and again later in this chapter.)
Checkpoint 3.0 Draft Architecture Specification
The first thing to realize is that the architecture is constantly evolving. It is impossible
to completely specify an architecture before a project is well under way. Our old friend,
the 80/20 rule, rears its head yet again.
14 3634_CH12
11:21 AM
Page 313
Milestones and Deadlines 313
The best that we can reasonably expect is to complete approximately 80% of the project architecture before programming begins. Even this will come as a surprise to most
developers in the games industry. Although it is more common to see projects defined
to this level outside the games industry, these are generally the projects that succeed,
and succeed well. Once the architecture is felt to be approximately 80% complete, the
project can get under way. The remaining 20% will be completed during the course of
the project. (This 20% usually consists of things that cannot be worked out fully in
advance. This is covered in more detail in Part III.)
Checkpoint 3.1 Project Initialization
The 80%-complete architecture is enough to allow a preliminary schedule to be
created. The work is divided into logical modules and divided among the groups as
appropriate. The module is then broken down within the groups into a further
schedule. This subschedule is then split into a series of interdependent tasks that
are assigned to individual group members.
Development is an intricate set of interdependent tasks that need to be organized to
minimize critical path problems. The analysis of these tasks and interdependencies
comprise a major proportion of the project management time through development.
Like the architecture, the schedule will be constantly updated during development.
(This may involve rescaling individual schedules and shifting individual tasks between
group members.) The important point is that this is a continuous tracking process that
runs throughout the project. It is not to be used just when problems arise. As will be
explained in Chapter 14 in Part III, the best troubleshooting is proactive, not reactive.
One of the major mistakes made in development is to treat the schedule as if it were
etched in stone. How can anyone predict the events of every working day for the next
18 months? Usually, the only time a schedule is modified is in emergency mode: Major
readjustments are made when the project is in crisis, a case of “too little, too late.”
As with navigating a ship, you will achieve better results if you make small corrections
to your course during the voyage rather than wait to make one large adjustment when
you’re getting near the destination.
14 3634_CH12
11:21 AM
Page 314
314 Chapter 12
The Next Steps
At this stage, the project is ready to start, so you need to consider where to go next.
The next sections explain how to define the milestones and checkpoints for the rest of
the project.
Defining Milestones
A typical game project can be partitioned into various subprojects. Each of these subprojects will, in turn, have interproject dependencies and (on a more detailed level)
intraproject dependencies.
A common mistake made with projects of all shapes and sizes is to assume that a milestone list is just that: a two-dimensional list that charts your progression from milestone
A to milestone B, and so on. You should realize that the list is really the collapsed form
of a multidimensional web of dependencies. When this web is collapsed into the form
of a linear list, you lose all of that interdependency information, and only the end
result of all the dependencies remains visible. This can reduce flexibility, so it can be useful to keep these complex interdependencies visible and up to date as long as possible.
Trying to visualize the complex web of information from a mere list of milestones is
comparable to trying to construct an intricate, three-dimensional object from looking
only at the shadow it casts. The shadow has lost some of the object’s dimensionality.
Consequently, important information is lost. It’s therefore crucial to have access to the
project interdependency information while working to a milestone list.
The first step in discovering and defining these dependencies is to create milestone
schedules for each of the subprojects. In this way, a multilevel hierarchy diagram of the
project can be developed. Remember, a game project is a large and complex problem,
and the obvious way to solve such a problem is to break it down into a number of
smaller—and simpler—problems. Drawing a hierarchy of the project (and including
the subprojects and their respective components) helps you get a good overall picture
that you and the team can refer to.
14 3634_CH12
11:21 AM
Page 315
Milestones and Deadlines 315
Figure 12.4 shows a template of the overall hierarchy chart format we use. You can
easily customize this chart for other structures, but the basic form allows a quick
understanding of the position and status of a project within the hierarchy. A chart
such as this placed on the company intranet site, as explained in Chapter 11, could
also contain hyperlinks to the relevant documentation for each section. Being a
hierarchical diagram, it can also act as a link in a bigger hierarchy that connects all
the projects in the company, and thus help to define strategies for future development.
By acknowledging these dependencies and creating the hierarchy diagram, it is easy
to see the big picture and decide how to draw on already available resources, thus
reducing unnecessary effort.
Basic Modules
Overall Company Architecture
Project Specific Modules
and Design Guidelines
Initial Design
Level of
Increasing Module Complexity
Increasing Experience and Skill-Level Required
Figure 12.4 Format of a hierarchy chart.
14 3634_CH12
11:21 AM
Page 316
316 Chapter 12
The chief benefit of making this diagram publicly available is that it effectively
eliminates the communication barriers in a company. Every member of every team is
capable of knowing exactly what every other member of every team is doing. The
ability to view this information at a high level, drilling down in detail if necessary,
makes the visibility of all progress of the entire project easily available to everyone.
Figure 12.5 shows the hierarchy diagram for the Balls! project.
File Access
Figure 12.5 The Balls! project.
Large projects need to be broken down into milestones. This is usually not too difficult
when there is clearly defined functionality, but it is crucial to avoid milestones that
encourage cutting corners in order to reach them. Instead, milestones should be
defined around architectural points, so that the architectural integrity of the solution
is maintained.
To better understand what makes a good milestone, we should pay some attention to
what makes a bad one.
Bad Milestones
Bad milestones are those that do not provide any information about the progress of
the project. These milestones can be described as fuzzy or are specified to such a
superficial level as to be essentially meaningless.
Examples of fuzzy or superficial milestones are:
Write a Visual Basic level editor to design worlds.
Convert the graphics engine to 32-bit.
14 3634_CH12
11:21 AM
Page 317
Milestones and Deadlines 317
Integrate two graphics systems so that we can see both on screen at once.
Have a playable level up and running.
Have five playable levels.
Go into alpha testing.
These are taken from various real-life projects. Some are obviously bad, but others are
not so obvious. We’ll examine each in turn and explain the good points, the bad
points, and how the milestone should have been specified.
“Write a Visual Basic Level Editor to Design Worlds…”
This (the milestone specified in Case Study 12.1) was for a flying shoot-’em-up game
that featured a fractal-generated landscape. The milestone is far too general; it doesn’t
provide enough information. It should have been changed to something similar to this:
Do the initial requirements gathering and feasibility study for a Visual Basic level editor to
design worlds suitable for our game.
From there, a schedule and a further set of architecture-based milestones should have
been developed.
“Convert the Graphics Engine to 32-Bit…”
This is one of the better milestones from the list. The full specification was to convert a
fully 16-bit set of C and assembly-language modules from a 16-bit architecture to take
full advantage of a 32-bit processor. This is no real problem. In fact, it is also fairly
clear and concise.
However, the problem is that it covered too large an area of work and didn’t specify to
what level to convert the code. For example, 16-bit code can be “converted” to 32-bit
code by the simple application of a thunking layer to call the 16-bit functions from 32bit code and handle all necessary conversions in the process. Alternatively, the milestone
could mean a ground-up rewrite of all the code so that the engine was entirely 32-bit.
So we have two solutions, each with advantages and disadvantages. The advantage of
thunking is that it is reasonably quick and straightforward to implement. The disadvantages are the extra layer of overhead necessary to convert arguments and return codes
for every function call and the fact that the code is still not fully 32-bit (and so may
have to use slow memory-management calls to work around 16-bit limitations).
14 3634_CH12
11:21 AM
Page 318
318 Chapter 12
The advantages of the second method are the general improvements that can be
achieved by such a conversion. The primary disadvantage is the amount of time such
an extensive conversion usually takes.
This milestone should have specified the type of conversion required. In fact, it is also
possible that not all the code would need to be converted. (For example, only the most
time-critical code would need to be converted, whereas other code—such as loading
code—could have used a thunking layer.)
This milestone should have been split into a number of smaller milestones that dealt
with the task module by module, Doing so would have allowed a more sensible decision as to how to tackle the conversion, and would also have drastically increased
visibility of the progress. In reality, this milestone disappeared from radar for a period
of four months while it was being worked on. This is clearly an unacceptable risk, and
hence would have been well served by the use of checkpoints. In this case however, we
were lucky. The programmer working on the engine was very competent and knew the
engine inside and out. However, that much reliance on a single team member is a very
bad thing (as discussed in Chapter 11).
“Integrate Two Graphics Systems So That We Can See Both Onscreen at Once…”
This is a similar problem to the previous one: There is simply not enough information.
However, this milestone involves the source-level integration of two complex modules
(meaning that standardized information has to be passed from one engine to the
other), and the interfaces for this need to be defined. This milestone describes an
entire project in its own right, rather than a small task in a larger project.
Think about the number of steps that are required to perform this task. First, you have
to analyze what information you need to export from each engine to allow the other
one to function. Which engine is going to have dominance? Is one meant to be a subset of the other? How will they interact? Do they have a shared screen buffer? What
modifications will need to be made to the way the engines use the screen buffer? What
impact will this have on other modules? How is this likely to affect future integration
of modules?
Already you can imagine how the list of milestones and checkpoints could be built up
using the guidelines in this chapter. By dividing the problem into atomic chunks, we
can increase both the visibility and the accuracy of the project progress indicators. The
14 3634_CH12
11:21 AM
Page 319
Milestones and Deadlines 319
idea is to attack the problem with enough questions so that the software planner’s
intent is perfectly clear.
“Have a Playable Level Up and Running…”
This is so fuzzy as to be virtually useless. What is meant exactly by “a playable level?
We can’t imagine this milestone as being useful in any way. It is not based on any technical or architectural checkpoints; it is instead a visual check to chart progress. This is
like an auto mechanic checking out your car by looking at the closed hood. To truly
know how things are going, he needs to open the hood and take a look at the engine.
Milestones such as this one encourage shortcuts in order to get a visible result.
Milestones that require visible checking such as this one are prone to undermining.
Also, how on earth do you estimate how long such a milestone would take?
The obvious solution to this is not to use this type of milestone at all. Make it so that
the visible features are the results of the milestone and not the aims of the milestone.
Applied to this example, this would result in milestones based on architectural goals.
The combined effect of completed milestones would result in a playable level.
In this case, the following milestones may allow this to be achieved:
If the level-editing software is complete, start to load and convert the level files
into the internal data format (including all subsequent loading of referenced
objects in the level data file and the handling of any unknown objects).
Correct display of level based on the level data that is loaded from file, including
handling of any unknown objects.
Implement user control using keyboard and/or joystick, including simple menu
navigation according to the design document and basic player control consistent
with the physics system described in the architecture document.
This certainly is not all that would need to be achieved, but it should give you a rough
idea of the difference between the two styles of creating a milestone.
14 3634_CH12
11:21 AM
Page 320
320 Chapter 12
“Have Five Playable Levels…”
This milestone directly followed the previous one. The idea was to ensure that the software was capable of loading generically defined levels (rather than being locked into
the single test level of the previous milestone).
Unfortunately, the net result was that the previous milestone contained more hardcoded assumptions than would normally be present, mainly because the programmers
could all see that they had been allowed time to clear up their own mess for the next
milestone. So they could take all the shortcuts they wanted to hit the original milestone. Needless to say, doing so involved reworking a lot of code that could have
been just written correctly in the first place.
The aims of this milestone should have been combined with the previous one and then
split into a series of smaller checkpoints based on technical goals, with the net effect of
these being to allow the loading of levels.
“Go Into Alpha Testing…”
This milestone is not specified in enough detail. What exactly is “alpha testing?” How
do we distinguish “alpha testing” from the standard developer and integration testing
that we have been performing throughout the project?
We would define alpha testing as the stage in which all components are “interface complete” (all the interfaces to all modules used in the project are implemented). It may
not be a useful implementation yet, but at least the components compile without any
undefined-link errors. Obviously, there has to be a certain level of functionality present
by this stage, but—assuming the development has progressed in a method similar to that
described in this book—it is likely that there will be a useful level of core functionality.
Why These Milestones Are Flawed
The main problem with these example milestones is their fuzziness. Most are not
well specified, and they can be achieved in too many differing ways. Short cuts can be
taken, thus propagating mistakes and misunderstandings around the project and even
14 3634_CH12
11:21 AM
Page 321
Milestones and Deadlines 321
As already stated, these example milestones are okay for general project waypoints, but
they leave a lot to be desired as detailed, payable-on-completion milestones. They
would need to be split into many micro-milestones, each with a binary, on/off completion criterion.
The next sections cover some ways to define better milestones.
Good Milestones
Having examined what makes a milestone bad, we now turn our attention to what
makes a milestone good. Much of this can be deduced from the previous discussion—
what isn’t bad must be good—so we’ll just summarize the features of a good milestone.
In general, a good milestone leaves no room for doubt. It is clear, concise, specified in
technical and architectural terms, and is binary. It is either completely finished or else
it is not finished.
Good milestones should be specified in detail. Each should specify in great depth,
exactly where the project should be at a specific stage. Estimates should not be based
on wishful thinking, but on brutally realistic assessments of development time that
are based on experience and not on marketing requirements. It doesn’t matter how
important it is to get something out before Christmas. You cannot change the laws of
physics, and equally you cannot change the length of time required to write good solid
code. You can, however, attempt to predict and schedule it. This is much more successful than trying to rush the coding to meet an arbitrary deadline, a gamble which in
most cases backfires.
Good estimates are achieved by first creating a hierarchy chart that tracks the complete
project and all its interdependencies, and then tracing a two-dimensional route from
start to finish, minimizing the delays due to dependency interlock on the way. Good
milestones are specified at the technical level, with each milestone or checkpoint specifying a stage in the completion of an architectural module or submodule. The crisper
and more defined the milestone, the better the project visibility. In order to accurately
estimate the amount of time required for a milestone, you must divide the milestone
into a list of sub-tasks that are small enough for you to reasonably and accurately
estimate the time required. This usually means breaking the tasks down into those
14 3634_CH12
11:21 AM
Page 322
322 Chapter 12
that will take a day or two, providing you with a good basis for submilestone checkpoints if required. This is a skilled process. As you gain experience in the estimation
process, you will find it’s becoming easier to subdivide the milestones and provide
accurate estimations.
A good milestone enhances project visibility and, upon completion, can provide a
profound sense of achievement for the individuals involved.
In Case Study 12.3, this particular milestone had been inserted into the schedule after
the architecture had been defined. It was a modification to the architecture to increase
screen update times for systems with slower graphics cards. (Consider this modification as part of the 20% of the architecture that can’t be known or guessed.) As you can
see, it is unlike the usual sort of project milestone, because it is based on the architecture and is specified in plain English with a minimum of technical jargon. (Any jargon
that is there can be easily explained.)
Case Study 12.3 A Real-Life Milestone
Balls! was used as a proof of concept for all the development methods presented in
this book.
This case study is an examination of one of the milestones set for the project and
details how it was subdivided into checkpoints.
The milestone:
Overview: Implement an object-level Z-buffer to increase speed update by minimizing unnecessary overdraw per frame.
Note: When performing this milestone, allow for reversion to the standard painters algorithm currently used by surrounding all Z-buffer code with conditional compile directives. When these directives are defined, the Z-buffer code will be activated, or else the
code will perform exactly as it did before.
Checkpoint 1.0. Take the standard test levels and activate the timing code in order
to time how long each frame is taking to draw. (0.5 days)
14 3634_CH12
11:21 AM
Page 323
Milestones and Deadlines 323
Checkpoint 2.0. Categorize the objects into the minimum number of sets. Each
set should contain those objects that completely overdraw the others when the twodimensional screen coordinates are the same. Flag exceptions to the rule (for example,
transparent objects should not cause objects behind them to disappear). Provide a
separate Z-buffer for each set of objects. (1 day)
Checkpoint 3.0. Implement “check” and “set” Z-buffer members in each of the effective classes. Each game object will have this member, so implement them as pure
virtual members of the graphic object base class. Provide a standard implementation
that takes a pointer to Z-buffer of the correct object set as a parameter as high up
the hierarchy as possible, and override this for specific cases (such as transparent
objects) where necessary. (1 day)
Checkpoint 4.0. Insert Z-buffer check and sets in the base class Prepare to Draw and
Draw members. Aim to fail the Z-buffer test as soon as possible in the processing
stage, so that the minimum amount of work is done on objects that are not going to
be displayed. (1 day)
Checkpoint 5.0. Produce new timing information for the test levels and compare
with the results achieved in checkpoint 1. Fully update the module design document to reflect the changes, and insert the new timing information. (1 day)
This is a direct transcript (although edited for brevity) of the milestone and checkpoints. We have modified checkpoint 1.0 as it also concerned reviewing and tidying
up the code to ensure that all comments were up to date and that all dead code had
been removed (because we hadn’t worked on that particular module for sometime).
This doesn’t matter here, but it accounted for an extra day in the schedule.
Because the example project was highly modular, the milestones were usually spaced
approximately two weeks apart, and the checkpoints one to two days apart.
Research Versus Deadlines
At this stage, it is worth digressing into a quick discussion of research milestones.
Let’s get the first thing out of the way: Research milestone is an oxymoron.
14 3634_CH12
11:21 AM
Page 324
324 Chapter 12
A deadline specifies a fixed date and time that were presumably decided upon by a
careful and informed consideration of the tasks to be performed. Research, on the
other hand, is an investigation of things unknown. How is it possible to accurately
schedule for the unknown? It’s not, but you try explaining that to your average project
manager and see what sort of response you get.
Research in this case should not be confused with requirements gathering. Research is
the process of seeking out new techniques and algorithms for implementing an idea.
Requirements gathering is the process of finding out what is required to build the
product specified in the design document.
The main job of a vanilla programmer, as mentioned in a previous chapter, is to translate a set of detailed technical specifications into a program or program module. It is
not to perform on-the-job research to meet a deadline. Research and deadlines are
often mutually exclusive entities. If research is needed at this stage in a project, something has gone very wrong with the initial feasibility studies (which should have investigated these possibilities beforehand). All research is a risk: you risk getting no results.
The later you are in the project when you take a research risk, the more unacceptable
that risk is. In short, avoid all unnecessary research during the course of a project.
Evaluation of Milestones
When it is time to evaluate a milestone, a number of people need to be able to sign it
off as being completed. At the very least, these are the publishers, the company, and
the project managers. Because of this, milestones tend to be based around results that
can be visually demonstrated. As discussed previously, this is bad. Basing milestones on
visual results means you often get just what you deserve: a program with a beautiful
exterior and a hollow interior. Like a swan on water, it appears smooth and graceful on
the surface, but paddles vigorously underneath. At the moment, this is how milestones
are often evaluated in the games industry.
The whole development life cycle is based around external appearances rather than
internal structure. Although this is fine for a painting, it is certainly no good for a complex technical undertaking such as a computer program. Imagine what would happen
if these criteria were used to build a bridge: It would be declared complete as soon as
the architect liked the color of paint that had been used. So what if it collapsed the first
time it was actually crossed?
14 3634_CH12
11:21 AM
Page 325
Milestones and Deadlines 325
It is in the best interests of all those who are concerned with evaluating milestones to
accept a schedule based around the architecture specifications.
Milestones should be as technical as possible, and act as an integrated statement of the
current architecture status. The architecture is the single most important entity in the
game development—hence, milestones should be based around it. It’s comparable to a
skeleton: it needs to be properly developed or the creature will be deformed, no matter
how attractive the flesh covering it.
Even if you cannot avoid using technical terms when writing your milestones, use
as much plain and concise English as possible. If you write carefully, even the most
nontechnical person can understand the increased benefit and visibility that concrete
technical milestones provide to a project. In general, the architect, the game designer,
and the publisher are the most important milestone reviewers. It is important to be
impartial, and the milestones should also be evaluated primarily from a strong
technical standpoint.
The members of the group responsible for the milestone should present their case as to
why the milestone is completed, and it is the task of the internal reviewers to act as a
devil’s advocate and try to prove that it is not complete. Be strict on this: an incomplete
milestone that has been flagged as complete tends to rear its head later in the most
unexpected and inconvenient.
Gaining a complete understanding of what a technical milestone represents is not
always immediately straightforward, especially for a nontechnical person. It is the
task of the group responsible to present and explain the milestone in such a way as
to enable the company management to understand the significance and how it fits
into the grand plan.
14 3634_CH12
11:21 AM
Page 326
15 3634_CH13
11:22 AM
Page 327
Chapter 13
• Procedures
Procedures and “Process”
• “Process”
• Source control and code
n this chapter, we’ll discuss the procedures and practices
that are required for the efficient and cost-effective use of
the software factory concept.
These procedures and practices are not restricted to the
software factory, as they produce benefits for all nontrivial
development. Much software in development (and game
software in particular) has no real procedural approach to
development. The growth of the average game project tends
to be only somewhat organized and very organic in nature.
This chapter covers the aspects of the development process
that are relevant to management, the areas in which
management is going to want status reports and control.
Management’s interest is the primary reason for procedures
and practices.
Managers want to be able to get accurate information and
status reports, and to act on this information without disrupting development. In the same manner, developers don’t
want managers indiscriminately wading in and disrupting
the development process by asking lots of pointless and
time-wasting questions.
These procedures and practices will cover the interfaces and
areas of interaction so that information can be transmitted in
a timely fashion and defined control paths can be used to
steer the course of development.
• The importance of
information transmission
• Proactive and reactive
15 3634_CH13
11:22 AM
Page 328
328 Chapter 13
The “process” doesn’t need to be as bad as you may think, as it is mostly a matter of
common sense. However, the best sort of common sense is that which is written down
so that everyone is clear as to what the local idea of “common sense” is. Common
sense breaks down when everyone has a different idea of what it is.
The extent to which these procedures are implemented will depend on the needs of the
project. A critical project will need a rigid approach, while a less critical one can afford
to be slightly more relaxed.
The types of procedures discussed in this chapter are basically controls and measures to
minimize wasted effort and to weed out low-quality work before it pollutes the project.
The following list of procedures is recommended to maintain accurate project information and control during the lifetime of the project. Each of the procedures is described
in the next few sections.
Design reviews
Document reviews
Technical reviews
Code reviews
Unit testing
Integration testing
System testing
Configuration testing
Regression testing
Overall, reviews are better at finding defects than is just testing alone, as long as the
reviews remain focused on detecting defects rather than correcting them. An added
benefit is that reviews also tend to find different types of errors than those found in
15 3634_CH13
11:22 AM
Page 329
Procedures and “Process” 329
The format of a review is roughly the same regardless of the material being reviewed.
The two classes of review are formal and informal. Which class you’ll use depends on
the nature of the work being reviewed. This is context sensitive, and the ratio of formal
to informal reviews will also depend on the situation (all of which is usually decided
by the project manager when assigning tasks, but individuals may also be able to contribute to the decision). After all, the project manager knows how complex the work
really is, and so he or she should be able to say which type of review would be ideal.
The important point is that no work should go unchecked. Reviews are usually the
first thing to fall by the wayside if the project is suffering from time pressures. This is
invariably a mistake. Short-term savings will add up to long-term pain.
It’s not good enough to just insert unchecked work into the project and assume that
if the whole thing doesn’t crash then the work must be okay. This “method” does not
provide any information on the impact the work will have on the project schedule, or
how well the work has been performed. Reviewing weeds out obvious errors at an
early stage—before they can get into the project—and it is one of the best preventive
measures for keeping out substandard and shoddy work.
In projects without review procedures, substandard work can go undetected for a long
time. This kind of problem shows up much later, usually when someone else has to
maintain or update the affected area. By this time, who knows how far the poison has
spread? Later work may have depended on the original, quirky work, and the evil was
thus perpetuated.
Reviews have other benefits, too: they allow the sharing and dissemination of information. Because the work is being discussed and reviewed in a group, the information is
naturally transferred among the team, which greatly reduces risk.
Informal Reviews
Informal reviews are very simple, with no real paperwork and no complex and timeconsuming meetings.
An informal review involves gathering a couple of team members around your monitor
and talking them through your work, giving a demonstration where applicable.
15 3634_CH13
11:22 AM
Page 330
330 Chapter 13
These reviews can be done in several ways. For example, in the case of an informal
code or design review, you can choose from a number of approaches, such as walkthroughs and inspections, although these can also be used for formal reviews as well.
Walkthroughs are directed by the author of the work and are mainly used when the
work to be reviewed is fairly simple and easily understood. A walkthrough has the
advantage of being quick and easy, as its purpose is just to bash out the technical quality of the work. It is not really an assessment; it is more of a request for comments. The
walkthrough is also a great opportunity for sharing skills and experience. If a reviewer
has a better idea for how to implement the work, it can be passed on to the author.
The converse is also true: sometimes the reviewers will learn some new tricks.
Information transmission helps to spread the load of knowledge more evenly, thus
reducing the reliance of the team on one person. In theory, every one of the employees
should be working to make themselves dispensible. No single person should be critical.
Obviously, for this to work, there should be no undue pressure and stress from management. The belief that one can get more work out of people by putting them under undue
pressure is false. For people to work their best, they must be as relaxed and secure as possible, while still realizing the importance of what they are doing and the need for deadlines to be met. This is enough pressure without extra arm-twisting by the management.
However, all is not rosy with walkthroughs, as they are potentially the least effective
way of presenting a review. According to statistics, the effectiveness of walkthroughs
alone varies wildly, finding only between 30 and 70% of errors.
One of the reviewers is usually the immediate technical manager, and he or she will be
able to give the work the seal of approval. Once this is done, and any issues or questions have been discussed and resolved, the work can be inserted into the project’s
main repository. The important thing here is that the technical manager should not
exert managerial authority, or it soon changes into something else, with the author
feeling as if he needs to glorify his work in order to prove himself. It shouldn’t be like
that. The purpose is just to discuss the work and to detect any obvious defects.
On the other hand, inspections are almost the reverse of a walkthrough. Rather than
have the author talk through the work, the technical manager takes the reins and performs the walkthrough himself. This has the advantage that a fresh eye examining the
work may spot problems that the author had missed through sheer overfamiliarity with
15 3634_CH13
11:22 AM
Page 331
Procedures and “Process” 331
the material. The disadvantage of this type of inspection, however, is that it can take
longer, but at least it stands a greater chance of finding defects.
If, during the review, someone detected an issue that was later resolved, the work
needs to be reviewed again by the original reviewers, including the one who requested
the change. Whether the subsequent review is formal or informal very much depends
on the nature of the change. Large changes require formal reviews, but minor tweaks
and clarifications merit only another informal review. Change control, a large topic in
its own right, is discussed in Chapter 14.
The danger of informal reviews is that they can become too informal. Thus, it is always
a good idea to make sure that responsible people are in charge of these reviews, and
that informal reviews are also combined with a fair proportion of formal reviews. An
ineffective informal review can be worse than having no review at all, instilling a false
sense of security in the team and causing no end of trouble.
Formal Reviews
Four or five people are usually involved in a review. The members of this review team
are chosen by an appropriate group lead and should include the person whose work is
being reviewed. Anything less is ineffective and anything more is inefficient.
The author’s responsibility is to ensure that all reviewers have the information they
need before the review begins, which may involve distributing printed copies of work
and answering questions.
The responsibilities of the reviewers are to ensure that they are familiar with the
material to be reviewed. The purpose of a review meeting is not to review the work;
this would take up too much time. The purpose of the review meeting is to discuss
the reviews done before the meeting.
So, before the meeting, the reviewers should complete a review form that lists the
points they intend to raise. A sample review form is given in Appendix A, “Sample
Game Design Documents.”
Basically, the review form is a reminder to the reviewers, so that they don’t forget any
observations. The comments are noted and a severity code is given, with A being the
most serious defect and E the least. (Of course, you can choose your own grading
scheme, but this is one of the most common.)
15 3634_CH13
11:22 AM
Page 332
332 Chapter 13
The work should be reviewed by a number of criteria, of which “fitness for purpose”
is the main one. The work should be what was asked for. Conformance to project and
company standards is also important, but functionality is obviously the top priority.
In the meeting, each of the points on the review forms is discussed. The meeting starts
with a presentation by the person whose work is being reviewed. This presentation will
be an overview of the work and the various implementation decisions.
One point that many people seem to miss is that criticism in reviews is not a personal
attack on either the individual or the work done by them. The object of a review is to
get second opinions in an attempt to weed out defective work. Focusing a group of
minds on the same problem may spot errors more easily. The focus of the reviews
needs to be on finding the errors; this meeting is not the place to discuss solutions.
Doing so would be a distraction, and solutions do not need to be immediately
The actual review meeting itself is rather insignificant in the search for defects. It just
serves as a common point for sharing information from the independent reviews done
before the meeting. Ninety percent of defects found in the review process are found
before the meeting.
After all the points on the review forms have been discussed, they are given an “action
status.” Basically, the points are either accepted or rejected. If accepted, the author has
to correct the work to the satisfaction of the person who raised the point. In cases in
which more than one reviewer raised the same point, only one of them takes “ownership” of it. All the points are then consolidated into an electronic review form, which is
stored with the work to be checked. If the changes affect the functionality in a significant way (say, if it affects other areas), then a change control meeting needs to be convened to control and limit any adverse effects.
Depending on the nature of the required changes, another formal review may be necessary. The same reviewers should be used. If the required changes are minor, informal
reviews should suffice. The author needs to have each reviewer sign off each of the
points to be addressed. Once this is done, the work can then be inserted into the main
body of the project.
15 3634_CH13
11:22 AM
Page 333
Procedures and “Process” 333
This may sound long-winded, and it is to tell you the truth. However, we have always
found that the time saved by preventing substandard work from entering the project
outweighs the time spent on reviews.
We have also found that if someone knows that their work will be reviewed, they will
work with more care and attention to detail and take more pride in it. This has a positive
benefit for morale. People know that the work going into the project is checked, and
the project is therefore of a higher standard (and the risk is also substantially reduced).
Reviews have been shown to increase productivity by approximately 20%. Never underestimate the positive effects that this could have on employee morale and output.
Because of the nature of game development, there is a natural resistance to the imposition of reviews. The most effective way to combat this is to collect statistics about the
number of defects found by reviews, and the average time spent finding them. Compare
them with the amount of time (and money) that would have been spent fixing these
defects later in the development cycle, and they cannot really be argued with.
Testing in General
While we have seen projects that have no form of review system, we have never seen
a project that had no form of testing system.
Projects with ineffective testing are usually easy to spot. When a game is released full
of bugs, it is usually for one of two reasons. The testing was not up to scratch, or the
development team had simply run out of time—though the publishers had had enough
time and published anyway, bugs and all. The latter is a scheduling issue. Here, we are
concerned with the first scenario: substandard testing.
Testing is the process of finding errors. How many errors can you expect to find in
code? The jury is out on the exact figure, but the industry average is roughly 15 to 50
errors for every thousand lines of code.
Testing looks for a number of errors that can have occurred at any of the stages of
development, but, in general, testing can detect errors at only the prototyping and
code implementation stages. Bear in mind that the longer a defect lies undiscovered,
15 3634_CH13
11:22 AM
Page 334
334 Chapter 13
the more expensive it will be to fix. If a defect occurs in the design phase, fixing it
when the implementation is half complete is going to be much more expensive than
if it had been found when it occurred back in the design phase.
In fact, this is the main reason to have a process. It allows you to attempt “phase containment.” No, that’s not something out of Star Trek; phase containment means to
detect and correct defects in the phase in which they were created. A standard industry
statistic holds that a defect can cost up to 200 times as much to fix if it is not discovered in the same phase it was created in. Obviously, it makes good sense to detect and
fix the defects as soon as possible.
Also remember that testing alone has no direct effect on the quality of the software.
Trying to improve the software by testing alone is ineffective; it just acts as an indicator
of the quality. It must be followed up with decisive measures to improve the quality.
Testing in this context generally refers to code, and so the following sections are geared
towards testing code.
Informal Tests
Informal tests are not in the same league as informal reviews. There’s no need to get a
group of peers crowded around the monitor, because an informal test is simply any test
that a developer does while developing a code module.
Every developer periodically compiles and tests code while developing a program.
This is a good start, but it is not sufficient to prevent bugs creeping into a program—
particularly at the module integration stage.
Many game development projects use only this sort of testing, along with a healthy
dose of playtesting towards the end of the project. This is usually only adequate.
In order to improve informal testing, a number of automatic error-detecting applications (such as NuMega BoundsChecker), and programming techniques can be used to
catch the obvious errors.
It can work, but, with the addition of at least some formal testing procedures, the risk
can be reduced substantially. Although the extent to which formal testing is implemented is dependent upon the nature of the work, all the testing stages that are
discussed here should be implemented in one form or the other.
15 3634_CH13
11:22 AM
Page 335
Procedures and “Process” 335
Formal Tests
Formal tests are fairly similar to informal tests except that they are more structured.
A test script is provided for a module to exhaustively test all features and functions of
that module. The script is usually written by the software planner with the module’s
Test scripts are written in a particular way. (A sample test script is given in Appendix
A.) The tests should exercise all the code paths in the module. Because this is quite
difficult to do correctly, the test script itself needs to be reviewed.
Three main types of testing can be used to do this: positive, negative, and ad hoc.
Positive testing checks for the intended behavior, so these tests are geared to produce
the expected results from the module. An example of positive testing is passing sets of
valid coordinates to a triangle-drawing module, and expecting to see the correct triangle produced.
Negative testing checks for unexpected behavior, and these tests are geared to produce
exception and error-handling conditions. An example of negative testing is to pass a
triangle-drawing module three points with the same value, or three points in a straight
line. Boundary conditions should also be tested. What happens when you pass very
large values (positive and negative)? What about very small values? Or zero values?
When planning tests, try to consider every type of input into the system, both valid
and invalid. The object is to stress the system as much as possible; throw everything
you can at it. You may not find all the errors, but you’ll weed out all but the trickiest
Ad hoc testing is effectively random. The tester plays with the module and checks
whether he or she gets the expected results. These tests are just to throw at the module
whatever input the tester feels like. An example is to pass sets of random points to
the triangle-drawing module, and check each output triangle to ensure that it appears
Many organizations use only one of these methods, but, for truly effective testing,
all three should be used. If a test script is written so that the three types of test are
segregated, then the tester has the choice of which tests to perform.
15 3634_CH13
11:22 AM
Page 336
336 Chapter 13
When a test script is executed, the results should be recorded in electronic form. The
tests should be worded so that the result can be expressed by true or false responses.
In some cases, further details are required to clarify results (especially if a test fails).
If a test fails, further action has to be taken. A test can fail for two reasons: a problem
with the code, or a problem with the test. The action in both cases is the same. The
tester raises a problem report, and the problem-fixing cycle begins again.
Unit Testing
Unit testing is the first wave of testing that the developer performs, and it is a type of
clear box testing, in which the tester has full knowledge of the internals of the material
being tested. Statistics show that, on average, unit testing can find approximately half
of the errors in a code module.
Unit testing is usually done with a test harness, a simple program designed to exercise
every feature of a module. This may be simple and noninteractive, or it may be more
advanced and allow the configuration of input parameters.
Unit testing means exactly what it says. It is testing in isolation. This prevents any
possible interactions with other, possibly error-riddled modules. This is why the test
harness is used, and is also why the test harness is as simple as possible. You don’t
want to spend time debugging the test harness! In a number of cases, errors in the test
results turn out to be due to errors in the test data being used! Having spent hours
searching for a phantom problem in the code, finding that the problem is in the test
data can be incredibly frustrating!
Unit testing can operate at several levels, If we take C++ as an example, unit testing
can be performed at the method level or at the object level. For those unfamiliar with
object-oriented technology, an object is a logical package of data and code that acts on
that data. The code is divided into methods that are analogous to functions in procedural programming. Object-level testing is black box testing, whereas method-level
testing is clear box testing.
Of course, testing cannot prove that a module is free of defects. A test can prove only
that a module has errors. The important rule to remember is that if a test indicates that
a module is free of defects, it is much more likely to be due to an inadequacy in the test
itself. Maybe not enough cases are being covered, or maybe there is an error in the test.
15 3634_CH13
11:22 AM
Page 337
Procedures and “Process” 337
It is difficult for developers to get themselves into the mindset required for testing. A
successful test finds broken code—but how many programmers do you know who
actively want to break their code?
Actually, this is a loaded question; the answer should be all of them. If not, the developers are not being thorough. Developers should be actively looking to break their own
code. They need to approach the testing assuming that the code is riddled with bugs. It
is like the modern scientific method, whereby a scientist propounds a theory and then
does his level best to think of experiments that will disprove the theory. Unfortunately,
some developers write code and immediately declare it finished. Finding errors in your
own code is not a sign of weakness but of thoroughness.
Integration Testing
Integration testing is the next level of testing done by the developer, although it can
also be done by a member of the test team. If a test team member does it, it is a black
box test. If a developer does the test, it is a clear box test. (It is preferable to have
both the developer and the tester do the integration testing, but sometimes this is not
possible. However, integration testing is a halfway house between unit testing and
system testing, so this is not a hard-and-fast rule.)
Integration testing is the act of testing the integration of a new code module with the
existing code base. Does it cause compile problems? Are there namespace clashes?
Does it even work?
The integration test is—by necessity—less detailed than the unit test, because it is a
test of how the module interoperates in the code environment. You could almost say
that it is a “field test” of the module.
The focus of the integration test is to resolve technical issues with the integration of
the new module. In theory, the module itself should be fairly free of error due to the
unit testing. During integration, problems always show up in great style that did not
manifest themselves during unit testing. These can be the most frustrating and difficult
to find, and they should serve as a good incentive to catch as many errors as possible
before this stage. You definitely want to try to find them before the system test stage. If
you think that diagnosing errors during integration is difficult, then just try finding the
little buggers during a system test!
15 3634_CH13
11:22 AM
Page 338
338 Chapter 13
The test used should be similar to that used in the unit test. A script should be written
that exercises each of the code paths.
The one hard-and-fast rule for integration testing is to integrate only one module at a
time. It’s a simple and obvious rule: It is exponentially more difficult to locate the
source of an error if you are integrating two or more untested modules. They may both
be full of errors, or they may interact in unforeseen ways. Either way, it’s territory you
don’t want to get into.
The good news is that system integration using object-oriented architecture is nowhere
near as difficult as it used to be in the bad old days of procedural programming. Objectoriented techniques can make all this much easier. Game development has been slow to
catch on to these techniques, mainly because object-oriented applications were considered slow bloatware, and the compilers were considered to create inefficient code.
This may have been valid in the bad old days, but because we now have to cooperate
with an operating system, consistently wringing every last ounce of performance out of
a system is more difficult. Besides, with object-oriented APIs such as DirectX acting as
an insulating layer between the game and the underlying hardware, it makes more
sense to use a programming language that makes development easier.
After the integration test is complete, the module can be signed off. It is then considered part of the code base.
System Testing
This should be done at least daily. If this is not possible, the absolute minimum interval
is once a week. Any more and the system test turnaround becomes ineffective. The
code will have changed so much in the one week that is becomes difficult to apply
fixes to broken code that may already have been modified or had other code built upon
it. The frequency of system test turnaround means that, in some cases, not all of the
system test can be performed. This usually applies only to the largest projects, and
most games should be able to be system tested daily.
In order to facilitate this, the software needs to be buildable every day. In fact, this is a
major requirement, irrespective of whether system testing is performed daily or not.
The software should always be in a buildable state. This doesn’t mean it should be fully
15 3634_CH13
11:22 AM
Page 339
Procedures and “Process” 339
functional; subsystems may not be present, in which case they should be stubbed out.
Stubbed out is where the interface is defined, but the functionality is not implemented,
so it is replaced with “stub functions” that output a debug message, or throw an error
exception, depending on the importance of the interface, The architecture has been
completely defined for the most part, so this is within easy reach of the developers.
The build is initially given a smoke test by the test team to see if it is stable enough to be
tested. A smoke test takes its name from electrical engineers’ method of testing newly
constructed equipment: They switch it on, and if it starts smoking, it doesn’t work.
The build is rejected if it fails the smoke test. The developers’ highest priority then
becomes fixing the build. The longer the build remains broken, the more time is wasted.
Once the build is working, it is labeled within the source control system. All good
source control software allows you to take a snapshot of the state of the archive so it
can be recreated later, even if further files have been added afterwards. Creating at least
a daily snapshot is important, because it ensures that the testing team is working with
a build that is (at most) one day old.
A sensible daily scheduling system for this is to ensure that all developers have
achieved sign-off by mid-afternoon. The resulting code and modules are then checked
into the source control system, and a complete system build is done.
The smoke test is done the same afternoon, and the system test is done by the testers
the following day. The frequency of building the game is limited to the length of time
required by the system testers to perform all the tests.
The system test should be conducted at two levels. The first level is to run test scripts
that check whether all the completed functionality is working as advertised. (This is
effectively the negative and positive testing.) These test scripts are most likely written
by the software architect; at this level of testing, it is mainly architectural design errors
rather than low-level coding errors. Of course, some integration issues that slipped
through integration testing are also likely to show up here.
The second level of the system test corresponds to the ad hoc testing. The testers
should play the game and test all the features of the program as they would expect an
end user to. They should use the manual to install and configure the game, and follow
the game instructions to see if they can play it.
15 3634_CH13
11:22 AM
Page 340
340 Chapter 13
Doing so serves two purposes: trapping any errors lurking in the code and providing
preliminary playtesting information. Obviously, this may not be much help if the game
is at a very early stage, but as the project progresses, it will become increasingly useful.
The results from the system test are recorded electronically, and any defects are
reported to the project manager.
Configuration Testing
Configuration testing is an extension of the system test. It is the testing of the
application on a range of different hardware configurations.
This is particularly important for games, as they are usually more dependent on hardware than are standard applications. The range of machines tested should encompass a
fair representation of what is in the marketplace: cutting-edge power monsters, weak
and feeble Pentiums with low resources, and everything in between,
And, yes, we know that DirectX and other APIs were supposed to solve all the hardware-compatibility problems. But if you believe that, you’ll believe anything.
To be frank, this level of configuration testing is usually beyond the resources of the
average development studio. However, independent testing houses can do this sort of
testing for you.
If resources are not available in house, independent testing houses could be your best
option (and it is a lot better than releasing a game that works only on your development machines and not on anyone else’s).
Regression Testing
Regression testing is not really a type of test in its own right; it is instead a technique
that is applied to all the other tests presented here.
Regression testing is the act of re-running the tests using the same test data used on
previous revisions of the module.
The idea is to ensure that the module hasn’t regressed to an earlier form. One of the
most common types of error is the reintroduction of previous defects. The regression
test is applicable to all the previous test types.
15 3634_CH13
11:22 AM
Page 341
Procedures and “Process” 341
To begin with, we will examine a general overview of development. At each stage, you
need to generate accurate status information. Hence, you need a set of procedures such
as those described to make this happen.
Figure 13.1 shows the main phases of development in a vague fashion. In reality,
things are more complex, with multiple interweaving strands of the main development
phases occurring simultaneously. The diagram does not attempt to show this, but
features only the points where an information-transmission interface—and therefore,
“process”—is required. The rest of this chapter is devoted to the discussion of how to
implement these measures, making method out of the madness.
Personnel Scheduling
Reviews and Approvals
Defect Tracking
Change Control
Quality Assurance
Development Planning
Figure 13.1 What is “Process”?
Process is usually viewed with a high measure of disdain among the development community at large. In the game-writing sector, it is viewed with even more scorn. Process
is seen as a complete waste of time (WOT) that subtracts from the amount of really
useful work (RUW) that can be done.
Examples of WOTs include rewriting old code to be compatible with new code,
rewriting of architectures, and uncontrolled revisions and modifications.
These developers believe that process is purely unnecessary overhead. They do not
take into account the amount of extra work that is prevented by the processes.
Figure 13.2 shows how they believe the way that work is distributed on a standard
project. Project work consists solely of writing new code and fixing old code (RUW),
with a small amount of overhead caused by having to deal with team dynamics:
meetings, duplication of other people’s work, and other WOTs.
15 3634_CH13
11:22 AM
Page 342
342 Chapter 13
Waste Of Time
Really Useful Work
Figure 13.2 Erroneous view of effort expenditure on a project without any process.
Figure 13.3 shows what effect that these developers believe process has on a project.
They believe that the process subtracts directly from the RUW in a one-to-one ratio.
For every hour of process endured, the project loses one hour of RUW.
Waste Of Time
Really Useful Work
Figure 13.3 Erroneous view of effort expenditure on a project with process.
15 3634_CH13
11:22 AM
Page 343
Procedures and “Process” 343
This can be true if the process used is badly suited to the project, but, in cases in
which it is well suited, the situation portrayed in the diagram is false.
Figure 13.4 shows the true work distribution on a project with no process. As the
project becomes bigger and closer to completion, the amount of WOTs increases. The
rate of increase depends on the competence of the people working on the project and
a fair degree of luck.
Waste Of Time
Really Useful Work
Figure 13.4 Accurate view of effort expenditure on a project without any process.
Figure 13.4 is shown for the average project. The amount of WOTs increase in proportion with the number of developers on the project. The more potential for interaction
between team members, the more potential there is for foul-ups.
The effect of process (Figure 13.5)—if the procedures in use are well chosen for the
project—is the increase the efficiency of work and reduce the amount of WOTs.
Obviously, these cannot be completely eliminated—there will always be unproductive
work—but the verification and cross-checking (as well as the visibility) provided by
good procedures will reduce this to the minimum.
15 3634_CH13
11:22 AM
Page 344
344 Chapter 13
Waste Of Time
Really Useful Work
Figure 13.5 Accurate view of effort expenditure on a project with process.
The process must scale with the amount of staff on the project. If this is done effectively,
the ratios of process, WOTs and RUWs remain fairly constant.
The trouble with process is that it is really difficult to achieve a balance between the
amount of process and the amount of work. In many cases, rigid formulaic procedures
are used without any account taken of the type of project and the needs of the developers. In such situations, the process is worse than useless, because it has a demoralizing effect on the developers. The process should be there to serve the project team; the
project team should not be slaves to the process.
The “knee-jerk” problem in Case Study 13.1 is common. A project that starts without
process and then tries to curb the ensuing problems by adding it at a later date is
subject to this common ailment.
15 3634_CH13
11:22 AM
Page 345
Procedures and “Process” 345
Case Study 13.1 Process Gone Mad
While working on a large-scale banking project, one of the authors discovered that
too much process can be as destructive as too little.
The development procedure for the particular suite of applications was fully
documented. There was an intricate set of rules to follow, laid out in a 300-page
This is ridiculous! Three hundred pages? What sort of process requires that much
detail to tell a developer how to write code for a specific project? Worse still, the
process tracking system had been written in-house using WordBasic (the programming language that comes with Microsoft Word), which was not up to the task.
Maintaining a code module sometimes required more than 20 pages of paperwork,
even for a small one-line change. Forms had to be completed in duplicate and
sometimes in triplicate. These then had to be printed, distributed to at least five
people, and the electronic copies placed in a disparate set of directories, the names
of which were constructed using some arcane rules that were defined somewhere
in the 300-page manual.
This was just to make a small change to a code module. You can imagine the red
tape required to create a new module. Let’s not even go into the red tape that bound
the testing team.
This was simply too much process. The rules were so complex that no one person
had a complete grasp of the entire procedure. Mistakes were made, shortcuts were
taken, and procedures were ignored.
The situation rapidly became worse than if there had been no procedure, because
the state of the project was unguessable. Because of the complexity, the status
documentation lagged behind the code by approximately three months. Modules
would be changed, and documentation would not be updated to match, causing
problems for the next poor soul. These problems rapidly compounded to make
the whole situation unmanageable.
15 3634_CH13
11:22 AM
Page 346
346 Chapter 13
The history of this project reveals what had happened to make it like this. It began
with a small core team requiring minimal (if any) process. As the complexity of
the project increased, more staff had been added, but the structures and processes
required to support them had lagged behind. After a time, the number of new staff
members reached critical mass and caused a crisis whereby the bug count was
reaching the high two-thousands, provoking a knee-jerk response: the instigation
of the draconian procedures.
The belief was that these procedures would ensure that no new bugs were introduced while the number of bugs already present were reduced. This may have been
true, but not for the right reasons! The procedures were so complex that they
slowed development to a crawl so that no code got written.
Unfortunately, it was not until new management was brought in that the situation
changed. The amount of process was slashed, and each unnecessary and overblown
procedure was pared down to the minimum needed to maintain control.
After that, the project started to get back on track.
This situation is shown in Figure 13.6.
Waste Of Time
Really Useful Work
Figure 13.6 The effect of late introduction of process on effort expenditure.
15 3634_CH13
11:22 AM
Page 347
Procedures and “Process” 347
The late introduction of the process in combination with the backlogged work brings
the project grinding to a halt. The project is usually cancelled at this point, but, in
Case Study 13.1, this was not possible, due to the critical importance of the project.
Game development projects are not usually so critical, so such a project would most
likely have been cancelled at this point.
Note that Figure 13.6 shows the project actually reaching its release date by the point
that process and WOTs consume 100% of the expended effort. In this sort of situation,
this point can be reached before the projected release date, resulting in a high probability of cancellation unless some fairly major remedial action is taken. This sort of drastic action is covered in Chapter 14, “Troubleshooting.”
In general, the introduction of process to a “virgin team” is likely to cause problems,
The main problems are going to arise from a restricted understanding of exactly what
process is and what it is for (a mental inertia that takes some time to overcome). There
is likely to be resentment and resistance to the enforcement of “unnecessary” procedure. Process is going to be viewed as simple overhead, which, at first, it will be. The
benefits of the formal procedures come into their own only when the project has been
under way for some time.
The solution to the resistance problem is to ask for a little trust. The alternative
system, with no real control or procedure, led to an inefficient working environment
and poor project visibility. The developers cannot be a law unto themselves. They are
working for a company, and the management of that company (the development team’s
“customers”) have every right to know exactly how the project is doing.
By the same token, the development team has the same right, as well as the right to
have the management know exactly how the project is going. In this way, preventive
steps can be taken if problems arise. If there is no visibility, the only steps that can be
taken are usually remedial after the problem has occurred, and are a lot more drastic,
in terms of scope, risk, and cost.
15 3634_CH13
11:22 AM
Page 348
348 Chapter 13
Experience has shown that what generally tends to happen with the introduction of
process to a new project is a temporary dip in productivity. It is important to recognize
this; otherwise, it could be seized upon as ammunition to argue why process is a bad
There are two main reasons for the dip. The first is that anything new involves a learning curve. The team has to become familiar with the procedures to be able to use them
efficiently. The procedures must become second nature to the entire team. This in turn
implies that the procedures must be simple and clear enough to be easily memorized.
The purpose of each procedure should be obvious to all. Even if the team is not initially enthusiastic about the procedure, they should at least be able to clearly see the
potential benefits.
The second reason for the dip in productivity is the nature of the initial phases of a
project: everything is new, so there is no great capacity for error. Process at the start is
effectively pure overhead. The main reason for the procedures is to prevent the sort of
errors that occur once a project progresses beyond a certain complexity threshold, and
this does not occur in the early stages. (Refer to Figure 13.5 to see this represented
Procedures: Where to Use Them?
Figure 13.7 shows the main phases of the module development (left), and the activities
associated with each phase (right). Note that the two boxed activities in the diagram
are at phase-transition boundaries. It is possible to insert procedures to control every
point of the design, development, test, and release phases, but this is seldom necessary.
15 3634_CH13
11:22 AM
Page 349
Procedures and “Process” 349
Initial Concept
Overall Design
Module Design
Detailed Technical Design
Unit Test
Integration Test
Sign Off
System Test
Quality Assurance
Figure 13.7 Breakdown of project module phases and activities.
A good rule of thumb is that the amount of process needed and the number of points
in the development cycle for which it is required scales in proportion to the size of
the project. The following sections discuss suggestions for where and which types of
procedures to implement.
The Design Phase
The design phase covers the project from the point where the game design is formalized by the game designer up to the point where the overall design for a module is
For a single project, this is a one-to-many relationship. There is only one game design,
but this leads to many module designs that all have to be consistent with the overall
architecture. Figure 13.8 demonstrates this concept.
15 3634_CH13
11:22 AM
Page 350
350 Chapter 13
Initial Concept
Overall Design
Figure 13.8 One-to-many relationship: initial concept to module design.
Initial Concept
There’s not much procedure that can be implemented here. The initial concept is too
far into the realms of imagination to be controllable. The initial concept for a game
(unless marketing gets too involved) is pure creativity.
Fortunately for management, the initial concept tends to be a kick-off point for a
project, and there is no shortage of good ideas—only of great ones.
Output: Ideas and notes describing the game. Basic concept sketches and diagrams.
A one-page “pitch” document.
Recommended Procedure: Presentation of the idea to stake-holding parties (the management, the development team, the publisher, or even just your colleagues).
The treatment is a proto-manifesto that defines the gameplay, a document that defines
the major features of the game and attempts to paint a picture of the game that sets the
mood for the team to follow.
Output: A formalized document that describes the game (the story, the look and feel,
and the basic mechanics).
15 3634_CH13
11:22 AM
Page 351
Procedures and “Process” 351
Recommended Procedure: Document review by relevant parties (the same parties who
reviewed the initial concept). The game design should be thoroughly thrashed out as
part of the review procedure, and it should be finalized before the next phase.
Overall Design
The overall design is the first draft of the detailed game design. This constantly
evolves during the lifetime of the project, but the baselining of this document is
required before any project-specific technical work can begin in earnest. The overall
design document is semi-technical and it specifies how the game works, looks, and
plays, and how all the game rules and units work in detail.
Output: A detailed specification of all units, characters, plots, physical appearance,
mood and setting, controls, and all other details related to the game. This document
could be used as a basis for the game manual.
Recommended Procedure: Document review by the entire team, or at least a representative
The architecture document is the initial technical document. This document describes
how the project will be constructed down to the module level. This should include
how the project fits in with the overall company architecture guidelines, including a
reuse plan to make optimum use of components that have already been developed.
Output: A document specifying the modules that compose the game and the connections
between these modules.
Recommended Procedure: Document review by the programming team, or at least a
representative sample.
Module Design
This is the hand-over document between design and development. One is written for
each module. The initial draft is written by a software planner or the game designer,
and the subsequent revisions are usually handled by the developers. The document is
a fairly high-level technical specification on how a particular module functions, and
it is used as a manual for the use of a particular code module. The module design
15 3634_CH13
11:22 AM
Page 352
352 Chapter 13
document is used chiefly to provide information for a library of modules, and is an
important part of the reuse plans. Note that a module can be of many types, each with
varying importance and reusability. For example, code module can be reusable, but an
artwork module or a game data file module is not easily reusable.
Output: A document containing detailed instructions on how a particular module
functions and what it is used for, containing examples of intended use. It must be
maintained along with the module.
Recommended Procedure: Technical document review by the lead programmer (where
possible), the software architect, and/or planner (if the document has been completed
by a programmer), and at least two other programmers. Where applicable, one of the
reviewers should be appointed as a reuse officer to make sure that all opportunities for
software reusability are being taken.
The Development Phase
The development phase doesn’t just include the writing of code, although some programmers seem to feel this way. Even many of the more enlightened programmers
believe that commenting code is sufficient documentation.
The development phase is the most critical phase of the entire cycle, but writing code
makes up only a small part of it.
For the purposes of this discussion, we are concentrating on the code development.
We are not including the development of artwork and other modules necessary for
the game, although the principles are the same as for code modules.
Detailed Technical Design
Every module is written in tandem with a detailed technical design document. This
can be viewed as an extension of the module design document, but with much more
detail. This document’s aim is to explain to other developers exactly how a module
functions, including the reasoning behind design choices and other salient details.
This document effectively becomes a journal of the development of the module, and
is considered to be as important as the module itself. The module and the technical
design document must always be maintained in tandem.
15 3634_CH13
11:22 AM
Page 353
Procedures and “Process” 353
Output: A document containing detailed technical specifications of a particular module,
such as interfaces, algorithms, design choices, test script, and test harness details.
Recommended Procedure: Design review conducted by developers and the software
The module is usually developed from a design document of some form. In the case of
artwork, this will usually be the game design in combination with a style guide. Code,
on the other hand, will be developed from a detailed technical design document.
Output: A module for the project. This could be a code file, a 3D model, 2D artwork,
a text configuration file, or anything to do with the project.
Recommended Procedure: Code review conducted by developers.
Unit Test
Unit testing is the testing of the code written by a developer, and usually by that developer. The unit test follows a script written by the developer as part of the technical
design document.
Output: Unit test script results.
Recommended Procedure: The errors found by the unit test are reported to the project
manager for assignment. Depending on the nature of the error, errors will usually be
assigned back to the original developer.
Integration Test
The integration test is usually the last round of testing that the developer does. The
developer tests the completed module to see if it fits in with the build. Integration
testing is considered to be an extension of the unit test, and the tests used are part of
the unit test script in the technical design document.
Output: Integration test script results.
Recommended Procedure: The same as for the unit test.
15 3634_CH13
11:22 AM
Page 354
354 Chapter 13
Sign Off
The sign off is the last phase of the developer’s work on the module. Sign off is
dependent on all the tests being completed successfully.
Output: A tested error-free project module.
Recommended Procedure: Check in to source control.
The Testing Phase
The testing phase involves three critical levels of tests: the system test, the quality
assurance test, and playtesting.
System Test
System testing is done by the testers as often as possible. It produces much more general information than would be produced by unit and integration testing due to its
black box nature.
Output: A system test log.
Recommended Procedure: Errors are reported to the project manager for assessment.
Quality Assurance
Quality assurance is a higher level of testing. This ensures that the program is artistically pleasing. It is not meant to find defects in the program, as they should have all
been weeded out by this stage.
The purpose of quality assurance is to make sure that the aesthetics (game atmosphere,
menu screens, manual, and so on) are user friendly and consistent, while conforming
to the game style section of the game design document.
Output: QA report.
Recommended Procedure: The results are reported to both the project manager and the
game designer for assessment.
15 3634_CH13
11:22 AM
Page 355
Procedures and “Process” 355
This is the final stage of testing. The total gaming experience is tested. How does it
play? Is the manual adequate? How is the learning curve? Are there any gameplay
Output: Gameplay report.
Recommended Procedure: Issues are raised with the game designer and the project manager for further assessment.
Source Control and Code Reviews: A Synergy
Synergy is defined as the sum of the parts producing a greater output than the parts
separately. The combination of source control with code reviews is a perfect example of
a synergy.
Source control is a software application that administers a centralized database of file
revisions. It allows a level of control to be maintained over who is working with which
source code file and which revision level of the code is released to testing. If a developer is working on a source code file, all others should be prevented from working on it
at the same time. This prevents those horrible configuration errors we used to get,
because two people were editing the same file in a mutually incompatible way.
Source control has become an indispensable part of development, acting as both an
automated project history log and a regulating control system for tracking the progress
of the review cycle. Case Study 13.2 provides an example.
Case Study 13.2 Source Control? We Don’ Need No
Steenkin’ Source Control!
Julian, a seasoned technical lead, had seen how development techniques were maturing outside the games industry before he started work at a fresh young
development studio.
15 3634_CH13
11:22 AM
Page 356
356 Chapter 13
When he arrived, he was put in charge of one of the two development teams,
working on a team-based sports game. The project had been under way for a few
months, and there had been no design phase, just coding, so already a substantial
spider’s web of badly organized code had been written.
“Okay,” said Julian to the team, “Let’s talk about source control.”
The team were highly dubious. They had never heard of such a thing. It was something that boring database programmers in gray suits used. Source control didn’t fit
in with their cool, seat-of-the-pants image of game coding.
“We don’t need source control,” one of the team members objected. “We’re doing
fine, and we don’t have time for the extra work. It’ll just slow us down.”
“But surely you can see the benefits? We’ll be able to keep an annotated archive of
the source code. We’ll know exactly how the work is progressing and who has done
what, when, and why,” explained Julian.
Now another member of the team had a comment. He had been a bit insecure about
the quality of his code, and didn’t like the idea of being “found out.”
the management wants to use this software to spy on us and check up on how
much work we are doing, is that it?” he asked, to murmurs of agreement from the
rest of the team.
The discussion soon degenerated to the point where the programming team
absolutely refused to have anything to do with source control.
In the interests of being diplomatic, Julian decided to let the issue go. Instead, he
went to a higher level to find out how the management wanted him to proceed.
The general consensus was pretty much the same as in the development team. “Why
do we want to buy software that the team feels is unnecessary? Surely they know what
they are doing, and if they say it will stifle their creativity, then let’s ride with that.”
15 3634_CH13
11:22 AM
Page 357
Procedures and “Process” 357
Julian realized he was facing a losing battle, and decided not to press the point.
“Without the support of management, there is no point in me trying to force them
to use source control,” he surmised.
Development continued as normal until one day a junior programmer shuffled up
to Julian. “We may have a tiny little problem with some of our source code,” he
mumbled. “What sort of problem?” asked Julian.
“Well, I was checking up on one of the core AI files, and I found that there were a
few bits that needed tidying up. So, for the past few days, I’ve been working on
that,” explained the programmer.
on,” said Julian, a growing sense of fear crawling up his spine.
The programmer gulped. “Well…it seems as if somebody else on the team was also
working on the same area. He started the day after me, but was away from his desk
when I asked if anybody was working on that file.”
Julian’s knuckles whitened, but he didn’t say anything.
The programmer continued. “It turns out that he was, and he’d optimized the core. His
new version ran at the target speed we were aiming for, but it still had the same algorithm flaws. He’d spent about a week working on it, and I copied my file over his. He’s
lost his work, and he’s furious with me. The last backup he has is over a week old.”
Julian sighed. “This is why I wanted source control. This wouldn’t have happened if
we had used source control procedures.”
Lack of code review procedures is not uncommon in the games industry. Until fairly
recently, even source control was an unfamiliar concept. When we began working in
the games industry, source control was viewed with distrust. It was considered restrictive and unnecessary, an invasion of privacy that tracks who wrote what code.
Code reviews were discussed in detail earlier in this chapter, so this section concentrates on source control and using it correctly, particularly in conjunction with code
15 3634_CH13
11:22 AM
Page 358
358 Chapter 13
Like any tool, source control is effective only if used correctly. If used ineffectively, it is
worse than useless: source code can be lost, archives can become confused, and no one
will be sure what the latest version of the project is.
What Should Source Control Be Used For?
The short answer: Everything!
The development as a module should be viewed as a package. All electronic material
relating to the module—such as design documents, problem reports, and review
reports—should be archived. All source control packages allow you to attach comments to the revisions. These should provide a summary explaining the changes made
and references to other related areas. An identity code that is unique to every code
review could be used to track changes across several files. If this was used, a simple
archive search will extract all information relating to a single piece of work.
Not all organizations need such a detailed audit trail, but the small amount of extra
effort that this requires can reap rewards later when the data is examined.
The Importance of Information Transmission
Information transmission is a sadly overlooked area of development and is usually
never singled out for special attention. The assumption, if it is even considered, is that
information transmission simply happens.
However, the reality is quite different. If information could just be transferred from
brain to brain, things would be simpler. Until then, we are going to have to rely on the
good old-fashioned methods of speech and writing.
Most teams have some form of information transmission, but it tends to evolve as
part of the team culture rather than being specified, and hence is often not as effective
as it could be. Most information is transmitted verbally and is about the day-to-day
concerns of the project. The state of the project is effectively maintained verbally, and
everybody has his own idea of the project’s status. This mismatch of information can
cause problems.
15 3634_CH13
11:22 AM
Page 359
Procedures and “Process” 359
In some cases, a token effort is made to keep track of the project status with documentation and regular status meetings. These can vary in effectiveness, depending on the
experience level of the project manager who implements these measures. The main
problem here is that attracting such skills into the games industry tends to be difficult
owing to the lower wages.
A common management theory, the Peter Principle, relates to this: An employee is
promoted to his or her level of incompetence. If an employee is good in a role, he or
she is likely to be promoted. Sooner or later, the employee will reach a position in
which the job no longer matches his or her skills. That’s the point at which they cease
to be promoted, so that’s where they remain. Most of the people promoted to project
management tend to be taken from the ranks of programmers. By the law of averages,
some of these people make excellent managers. However, a developer’s skill as a manager is not necessarily proportional to their skill as a programmer. Hence in many cases
the team can end up with a bad project manager just because he or she happens to be a
superb programmer. The important point to realize is that the team members must
look to the team leader for how to conduct themselves, and the team leader sets the
tone for any intrateam communication. In the best-managed teams, communication
works efficiently on several levels. Not only are the details of day-to-day events spread
among the entire team, but skills and experience are shared. The net effect is that, as
the project progresses, the entire team is raised to a new common level of skill. A fair
amount of this is project specific, but a significant portion is an improvement to the
general abilities of the individual.
This most effective form of information transmission—the spread of knowledge—relies
on a number of factors. The first, and most important, of these is the use of all the
communication methods mentioned previously. The second of these is glasnost—a
sense of openness—among the team. Before we discuss specific recommendations in
detail, here’s a digression on the relative efficiencies of communication.
Figures 13.9 and 13.10 demonstrate the factorial nature of communication lines
between team members.
15 3634_CH13
11:22 AM
Page 360
360 Chapter 13
Figure 13.9 Communication lines between three team members.
Figure 13.10 Communication lines between six team members.
When there are three members in a team, there are three lines of communication. A
can speak to B, B can speak to C, and C can speak to A. This is easily manageable.
Each developer needs to communicate with only two other team members, which
won’t take up much of his or her time.
However, with four, five, or even six team members (as shown in Figure 13.10), the
situation becomes horribly complicated.
Figure 13.10 reveals 15 such communication lines. This is substantially more than the
original three, and hence will take up more of an individual developer’s time.
15 3634_CH13
11:22 AM
Page 361
Procedures and “Process” 361
Of course, this is a worst-case example, and it doesn’t take into account the fact that
communication, even in the most lax organizations, is usually more structured than
Figure 13.11 shows the sort of situation that usually arises.
Team 1
Team 2
Team 3
Team 4
Figure 13.11 Communication lines between four teams.
One of the reasons for dividing large teams into functional subteams is to optimize
communications. With n team members, there will be n(n-1)/2 communication links.
In Figure 13.11, a team of 24 employees is divided into four subteams of six members
each with a team leader (as shown in Figure 13.10). Interteam communication is
handled via each of the team leaders, and this helps to reduce the nightmare of
communications that would occur otherwise. With this configuration, there are six
interteam communication links, and four sets of 15 intrateam links, totaling 66 official
communication links.
Compare this with an undivided team of 24 members. There would be 276 lines of communication. It’s clear that this would take up a major portion of an individual’s time, very
little work would get done, and soon the whole thing would collapse into anarchy.
This system can be used quite efficiently for the transfer of information, but there are
more efficient means. This sort of verbal communication is essentially one to one, and
there are many cases in which one-to-many communication methods are far more
effective: meetings, documents, and internal Web sites. Each has pros and cons.
15 3634_CH13
11:22 AM
Page 362
362 Chapter 13
Meetings, for example, are not usually very efficient. (Some managers seem to have a
peculiar fondness for meetings. It’s almost as if they feel that their lives would not be
quite complete if they did not have at least one meeting a day, although preferably
more often.)
Our view is that meetings should generally be called only in exceptional cases, and
even then not everybody needs to be automatically included. Meetings tend to waste a
lot of time, and, although they are more efficient than telling the individuals one by
one, simple status reports can be done much more effectively by the use of an internal
Web site or weekly newspaper sent out by email.
If everything is going smoothly, meetings are not generally necessary. If there are problems, or changes ahead—for example, news that requires widespread action—then a
meeting can become necessary.
It is not only clear written communication with other team members that can pay dividends. Documentation prepared for the purpose of communicating with your future
self is often very useful too:
Most programmers would say that when updating code they wrote a few months before, their first
instinct is that it would be easier to rewrite it from scratch. This is usually because they don’t want
to make the effort to refamiliarize themselves with the old code. But if the code is commented well
and the associated documents are in line with the code, it is much less work to modify the old
code than to rewrite the whole thing from scratch. This shows the importance of documenting
and commenting code effectively.
Proactive and Reactive Information Transmission
Information to be transferred may be proactive or reactive. What exactly do we mean
by that?
Proactive news is information that directly affects the future of the project. An
example would be a large change request. This potentially affects all personnel and
may require the gathering of opinions in order to decide further courses of action.
15 3634_CH13
11:22 AM
Page 363
Procedures and “Process” 363
Proactive information is usually best disseminated at meetings. In fact, most of the
meetings in a well-run software project will be review meetings and change control
meetings. Except in some specific instances, everything else can be transmitted more
efficiently by other means.
Reactive news is effectively everything else. Examples would include project status
reports, results from reviews, information on competitors’ projects, or any of the 101
humdrum occurrences that are part of any project.
We have found that the two best routes to transmit this sort of information is either
an email newsletter or a linked web of documents on the intranet Web site (better
Each project should have its own home page, such as that shown in Figure 13.12.
Figure 13.12 Example home page for the fictional Project X.
15 3634_CH13
11:22 AM
Page 364
364 Chapter 13
The sample page shown in Figure 13.13 would be used with the software factory
model described in Chapter 11, “The Software Factory,” but the exact structure is, of
course, down to personal taste. Note that only the links from the main project home
page are shown for clarity.
Figure 13.13 Example Web links for software factory model.
In the ideal system, a weekly email newsletter should be sent to all employees, containing a digest of updated material. Of course, it is the responsibility of the employee to
check out the information that applies to him or her.
The one main drawback is that larger organizations may have to employ a dedicated
Web page designer to perform this work. Fortunately, most of these companies will
already employ such people to maintain their Web sites anyway.
Moving all the company documents and process records onto an intranet is quite an
undertaking. It needs to be carefully designed, and the intranet should have search
capability. If this is done properly, anyone in the company will be able to view the
project status.
15 3634_CH13
11:22 AM
Page 365
Procedures and “Process” 365
We implemented a system such as this on one of the development projects we worked
on and extended it to include a scheduling system so that tasks could be assigned to
employees by the management using a simple Web-based interface. The employees
were then able to access a customized page that detailed the tasks they had to do, as
well as a personalized schedule of their review and meeting assignments.
Assuming that the home page is carefully designed and laid out, it should nullify the
classic “But I couldn’t find it anywhere” excuse used by nearly every programmer
(including one of the authors of this book!) to avoid reading documentation.
15 3634_CH13
11:22 AM
Page 366
16 3634_CH14
11:24 AM
Page 367
Chapter 14
• Risks
• Proactive and reactive
• Change control
n this chapter we are going to look at what can go wrong
with your project. The average project is a catalog of errors
waiting to happen: scheduling errors, coding errors, obsolescence, personnel problems, and illness. As John Lennon said,
“Life is what happens to you while you’re busy making other
Troubleshooting (noun). Tracing and correcting faults in
machinery, etc.
This chapter may be titled “Troubleshooting,” but we’ll
mostly be suggesting methods of preventing the need to troubleshoot. The most cost-effective projects are those that
build high-quality software from the outset.
Even in the best-run projects, unexpected problems will crop
up. What differentiates the successful teams from those that
fail is how well they plan for and handle these problems.
How can you plan for the unknown? If you don’t know what
is going to happen, how can you take measures to fix it?
No one can read the future, so there is no way of knowing
what troubles will beset your project. However, you do know
that something is bound to happen. Only the most naïve project managers assume that a project will run smoothly and
free of trouble to completion. Believe it or not, there are
managers like that. We know of one who maintained a belief
that an 18-month project could be completed in a 9-month
• Contingency planning—
the “Plan B” scenario
• Metrics and information
• Disaster recovery
16 3634_CH14
11:24 AM
Page 368
368 Chapter 14
period. He believed this up until the seventh month, at which point a hasty renegotiation of the contract with the publisher was required. The game in question was
released one year and two resignations too late. This is, incredibly, not an isolated case.
Why does this sort of thing occur? It’s because of lack of planning. An ounce of
preproduction is worth a pound of production.
The games industry, as a whole, was still cutting its first teeth when the mainframe
world was working on huge projects involving millions of lines of code and vast teams
of people. As a testament to the strength and cohesion of these teams, many of the
results of these large-scale projects are still in use today. (Admittedly, some of the more
cynical would attribute this to the fact that the cost of replacing them is more than the
cost of keeping them running. This may be true to some extent, but at least they are
still working.)
However, games are generally not meant to have 20-year lifespans, so planning and
scheduling hasn’t evolved as fully as they have in other disciplines. In fact, owing to the
short lifespan of the average game, such organized practices have not been as necessary.
But, as the song goes, the times they are a-changing. Games projects have become bigger and more involved. It’s no longer a small undertaking. Few platforms remain where
the lone developer can produce anything to match the big teams. Today it’s all motion
capture, FMV, movie-quality soundtracks, movie-quality artwork, Einsteinian levels of
AI and more polygons than you can shake a stick at.
Unlike most software projects, however, there is a large artistic element to game
development. These particular elements are best suited to a movie-industry style of
deployment, but the difference between movie and game production is still quite
large. Movies are mainly artistic—there is a large technical element to them, but the
production process has had a century to evolve and runs smoothly and efficiently by
outsourcing requirements when needed.
However, the requirements for the production of art, FMV, and music for a game
project do closely match those of the movie industry, and there would be a point to
mimicking the movies in this respect: The movie producers usually outsource their
requirements to external companies rather than maintain expensive in-house teams.
16 3634_CH14
11:24 AM
Page 369
Troubleshooting 369
If game development can truly be compared to the movie industry, then the movie producers would employ new engineers for every project to build new cameras, reinvent
celluloid, and redevelop all the software used from scratch.
The movie industry has reached a plateau where the technology is essentially stable,
and the “target platform” of a cinema screen is generally consistent and unvarying.
The games industry will not reach a similar plateau any time soon. The technology is
constantly evolving—games technology may not ever stabilize, even on the consoles.
Every couple of years, there is a new upheaval in the state of the technology. The advances
predicted by Moore’s famous “law” (that computing power doubles every 18 months or
so) show no signs of relenting. An insulating layer, such as DirectX, does provide a light
buffer, shielding the developer from the immediate effects of the underlying technological
changes. However, DirectX is revised every year or so to keep up with the technology.
Given the way things are moving, middleware packages will eventually evolve to
become industry-standard game-development kits containing all the components
required (such as graphics engines, AI engines, and scripting engines) to allow relatively
nontechnical game designers to direct the production of a game much in the same way
as a movie director does with a movie. The game designer will be able to outsource
graphics, music, and maybe even buy AI modules to augment the game design “script”
he is producing. Even then, there are likely to be different packages available to cover
each distinctive genre of gaming, not to mention the many varying styles of presentation. (For example, first-person 3D is not an ideal platform for a strategy game.)
For the moment, however, game development is fundamentally an engineering discipline with artistic aspects. If there were any contemporary discipline that can be effectively compared to the games industry, it would be bridge building. When building a
bridge, the plans are laid out in advance, the work is performed to a schedule, and
contingency planning is taken very seriously.
Essentially both bridge building and game development are engineering disciplines; both
produce “works of art” as output, and both require extensive planning to have any hope of
being finished on time and on budget. Can you imagine building a bridge by gathering 100
engineers, giving them several hundred tons of metal and some tools, and telling them to
get started, without any planning, and just a rough idea of what the finished bridge should
look like, whilst pointing at a vague spot just over the other side of the river?
16 3634_CH14
11:24 AM
Page 370
370 Chapter 14
You can deal with problems in generally two ways: after they occur or before they
occur. Which sounds better to you?
But how do you plan for the unknown? It is obviously not possible to predict and
make provisions for every possible problem in advance. Rather, you need a methodology that will prepare the development team for responding to the problems. This
kind of contingency planning is possible—it has been in use in military organizations
since the time of Pyrrhus of Epirus.
We should advise that managers sometimes do not understand why you would want
to spend money on things that will most probably never happen. They may call you a
jinx or a doomsayer. You should stand your ground, however. The extra effort involved
in contingency planning acts as a sort of “insurance” for the project.
There are certain risks common to nearly every project, and ignoring these lessons is
the single biggest risk. It is worth expending a small amount of effort to build guaranteed resilience to these problems.
In searching for potential problem areas, it is better to examine each aspect of the
entire project and try to identify where problems may occur. It could be an experimental architecture, a risky or unproven technique, personnel shortages, funding problems,
or other such difficulties. (This will be discussed in more detail later in the chapter.)
In the cases in which it is not possible to plan, a sensible, general approach to handling
problems is necessary.
The two types of troubleshooting are proactive and reactive. Proactive troubleshooting
is preferable but not always possible. Reactive troubleshooting is not always the best
approach but can sometimes be unavoidable.
Reactive troubleshooting is more commonly known as “fire fighting” or any other
phrase indicating that a problem is being addressed only after it has surfaced. By this
time, of course, the problem has already had an effect on the project. The effectiveness
of the troubleshooting depends on your ability to detect the problem before it has had
much effect.
16 3634_CH14
11:24 AM
Page 371
Troubleshooting 371
Unfortunately, detecting these problems is not always straightforward for a number of
reasons. The first is blindness through overfamiliarity. When you’re working day after
day on the same project doing the same set of tasks, it is common to suffer from tunnel
vision. Sometimes you just do not notice the problems mounting up around you.
Unless the problems are pointed out to you before it is too late, you may notice them
only when they are too big not to notice.
The second reason for overlooking problems depends very much upon the reputation
of the person acting as a manager. Do people want to tell him or her about problems?
Does the manager shoot the messenger? This is why being approachable is a very
important behavioral trait in a manager.
If this sounds like a manager that you know, there is a good chance that he or she is
not being told about problems at the earliest opportunity. This is obviously disadvantageous, as it does not allow the team to get cracking on solving the problem. An anonymous feedback channel can sometimes circumvent this.
There is not enough contingency planning in the game industry. The evidence is all
around. Games are cancelled or delayed and, if they are released, they are often rife
with bugs, requiring several post-release patches to bring them up to an acceptable
standard. These are signs that no contingency planning has been used. Games that are
delayed are usually the victims of reactive troubleshooting. No plans had been made to
foresee and cope with the problems, and these problems were hurriedly dealt with after
they had arisen (more commonly known as shutting the stable door after the horse has
Correctly implemented contingency planning adds very little work to a project. In fact,
the aim of contingency planning is to save money and time. By preparing an alternative
solution to a potential problem before the problem actually occurs, the impact on the
project can be assessed and prepared for. The project thus acquires a built-in resilience.
This chapter aims to set this record right. The advice and guidelines may not save your
project if the proverbial “stuff” really hits the fan, but it will at least prepare you for
ducking out of the way when it does.
16 3634_CH14
11:24 AM
Page 372
372 Chapter 14
The most important things to consider are the risks that your project will inevitably
face. Every day, you will face new ones. They will need to be carefully monitored and
checked. This section will give tips on how to implement a sensible and practical
risk-management plan in order to avoid these risks when possible, and handle them
in situations when this is not possible.
In a fair proportion of software-development projects today, risk management is not
practiced effectively, if at all. That is not to say it is not considered. In one way or the
other, risks are considered in every project under the sun. However, they are only
usually considered as part of another process, and not specifically in their own right.
Usually risks are not tackled directly, but are dealt with only as a secondary activity.
In this way, risks remain in the peripheral vision of a project manager and his team,
but are never focused on directly until they are dire enough to prevent normal work.
Because of this lack of focus, risks are managed only if they are covered by another
area of work. For example, when a project manager is working on the schedule, he
will also be examining the risk from schedule slippage, and to some extent, personnel
problems. When a developer is developing code, he is checking for risks from bugs
being introduced into the codebase, and checking for any that may already exist.
Risks to an area are covered only when that area is being worked on.
However, it is abundantly clear that, because risks are never addressed directly as an
area of work in their own right, there is the danger of hitherto unconsidered risks
falling through the cracks. This is why the area of risk management should be considered in its own right.
By concentrating at least a part of your project’s effort onto risk management, you will
make the ride a lot smoother. Projects that do not attempt to anticipate risks before
they occur appear to an outsider to careen from one crisis to another before running
out of momentum and then either failing just short of the finish line or flopping
miserably over it, never to be heard of again.
The aim of risk management is to provide as pleasant a ride as possible through rough
seas. Figure 14.1 illustrates this point.
16 3634_CH14
11:24 AM
Page 373
Troubleshooting 373
Loss of
Loss of
key employee
key employee
BEFORE Risk Management
AFTER Risk Management
Figure 14.1 The difference that risk management can make to your project.
Risk management is a field of study in its own right. How would you go about instigating a risk-management program for your project?
The first thing to consider is where—and how—to look for risks. This is a substantial
task, so you should seriously consider assigning the role of “Risk Assessor” to someone
on the team so that it becomes a fair portion of his or her duties. This position could
be rotated among employees on a weekly, biweekly, or monthly basis.
The job of the risk assessor is to keep a weather eye out for potential threats to the project. The Risk Assessor should closely monitor each “front” of the project. A good way of
doing this is to maintain a “top 10” risk list, which should be updated at least once a
week. The risks on this list should be given a priority and rated for their severity.
For a start, the mere psychological effects of this are usually positive. The idea of constantly seeking out risks and dealing with them as they occur will bolster the morale of
the team. And anything that can be seen to boost the morale of the team has to be a
good thing, even if the effects are not directly measurable.
For each of the high-priority items on the list, a method of resolution should be decided, and it should be the team’s priority where possible to tackle and eliminate the risk.
In some cases, for the more unexpected and extreme risks, it is necessary to form an
“attack team” of employees pulled from their normal work.
16 3634_CH14
11:24 AM
Page 374
374 Chapter 14
This can be a scary proposition, and it needs to be handled carefully to avoid panicking people. Sometimes, and rather unfortunately, this form of positive action can be
viewed in a negative light. Rather than being seen as an effective way of reacting to the
changing needs of the project, it can be viewed as “headless chicken” mode.
To be fair, if the risk management plan is not carefully thought out and implemented,
it can degenerate to aimless thrashing. If the risks on the list are only trivial and
unnecessary action is taken, it can detract from the real work. Before any action is
taken, the proposed risk-handling plans will have to be ratified by the relevant parties.
This could vary depending on the nature of the risk. For example, if the risk involves
a change to part of the system, it is subject to full change control.
The job of the risk assessor can be summed up by saying that he or she is the person
who sniffs around the project looking for trouble. It is his or her efficiency at rooting
out troubles that will save time, effort, and money further down the line.
The following list is an example of the sort of things that you are likely to see on your
top 10 risk list.
Project X Top 10 Risk List
1. Implementation of data packet encryption algorithm is too slow for realtime
network use. If we have to remove it, this could leave our peer-to-peer network
gameplay open to hacking.
2. Artwork for the main character is taking too long to produce. The animation
programmers are being held up by the lack of frames and information pertaining
to those frames.
3. The team is being held up by a delayed map design tool. We need to begin
designing new levels as soon as possible or we will begin to slip our schedule.
4. A new service pack has just been released for the C++ compiler. This needs to be
analyzed to see if it fixes anything that has an impact on our project.
5. There is a bug in the graphics-rendering module causing graphic corruption every
few frames. This is not a fatal error but is noticeable enough to be annoying.
16 3634_CH14
11:24 AM
Page 375
Troubleshooting 375
6. There are not enough personnel to cover the amount of work required. We are
beginning to work long hours, and this could be detrimental to morale.
7. The compression module we need to use is buggy and difficult to maintain. The
tests have shown that it has a low mean time between failure, corrupting 1 in
100 bytes.
8. The 3D rendering module has just been updated by the core team, and we need
to make a few code modifications to use it. The advantages of this will be faster
and smoother screen updates, and more support for the latest hardware features.
9. We need to bring John, the new team member, up to speed as soon as possible
so he can begin to be productive. This should help to alleviate point 6.
10. The test harness for the AI module needs to be updated to test the new features.
We would like to make use of the updated fuzzy logic functions that it provides,
because this may improve the reaction of the AI opponents to the characters’
Not all of the points are valid and worth addressing, and the order that they are presented in the list is arguable. Some of the risks may come from the personal opinion of
the risk assessor (a good reason to rotate personnel into and out of this position, or at
the very least try to find someone who can be unbiased).
The question as to which are worth dealing with and which are not is subjective,
depending very much on the status of the project, and different priorities will surface
depending on the state of play. For example, the priority might be to complete the
game as soon as possible. In this case, the project manager may not want to take the
risk of introducing a new technology such as a new 3D engine at this stage (as 3D
Realms did when switching Duke Nukem Forever from the Quake II engine to the
Unreal engine). Some managers would consider running a parallel development
stream, with one stream using the old engine and one using the new, to hedge their
bets. If the new engine is successful, they can proceed with that, assuming that the two
codebases have not diverged too much. If the codebases have diverged, the integration
then becomes a large risk in its own right.
16 3634_CH14
11:24 AM
Page 376
376 Chapter 14
However, if the project is not in the final stages, the new 3D engine may take a much
higher priority. By the time the game is likely to be released, technology will have
moved on, so it will be commercially advantageous to support the latest technology.
Releasing the game with the older engine may make the game look dated. Any problems with the 3D engine will most likely be worked out before release.
The items on the list can be grouped into a number of categories, based on the areas
that they affect. These areas are not all-inclusive, but cover the main types of problems
that affect the day-to-day running of a project. Some of these will not be detectable by
the risk assessor, because they affect a higher level of management. Someone will need
to keep on eye on things at the company level, but that is usually handled without too
much difficulty by the management. The danger areas tend to be more at the project
level, where everybody is too busy looking at the trees to be able to see the forest, or
even the chainsaw-wielding lumberjacks flitting among them!
The following sections will discuss some of these risks, how they might manifest themselves, and how they could be handled.
Design and Architecture Problems
Problems with the design and architecture are the most insidious and difficult to deal
with. These are difficulties in the very roots and foundation of a project, and have to
be dealt with as expediently as possible.
Potentially, this sort of problem can cause the maximum amount of rework. Obviously,
this could be very costly.
Changes to Baselined Requirements
Changes to requirements once they have been officially signed off can cause problems
with integration and mismatched functionality.
These changes can manifest themselves in a number of ways. Sometimes they will be
due to “good ideas” from team members. The use of change-control boards tends to
reduce this problem, but can also leave personnel disgruntled when their pet idea has
been shot down.
16 3634_CH14
11:24 AM
Page 377
Troubleshooting 377
There can be other reasons. The publisher or an external organization (such as a ratings board or a major distributor) could request changes (for example, a request to
reduce the amount of blood shown). This sort of request is usually enforced by the
imposition of financial penalties, such as the distributor refusing to carry the game, or
the publisher or ratings board preventing publication. Either way, usually the only
option is to toe the line and make the requested changes, whatever the knock-on risks.
Poorly Defined Requirements
If the requirements are poorly defined, they could be insufficient for the needs of the
project. If this is the case, it is pretty much certain that further requirements will be
defined to make up for these shortcomings.
This definition and redefinition of the inadequate requirements is most likely to
expand the project’s scope. If the scope is expanded, the work already done probably
will need to be revised to accommodate the new requests. Of course, it is never possible to define all the requirements before beginning architecture design, because some
details cannot be known before getting your hands dirty with the real work.
The point is that at least 80% of the requirements should be defined in as much detail
as possible before anybody starts work on the architecture. Likewise, at least 80% of
the architecture should also be defined in detail before construction begins.
Sometimes, the design and architecture can be well-defined, and then, out of the blue,
additional requirements are added. This can happen for any number of reasons—
publisher demand or a similar game being released with a “must have” feature that
will make your product seem out of date if it does not have it. An example would be
the addition of multiplayer features (although most games do have this now) or the
sudden spate of bullet-time features in games following the success of Max Payne.
In these circumstances, there is usually no choice except to make the modifications.
This can sometimes be impossible without a complete rewrite of most of the code and
modules. If such a rewrite is not possible due to financial, contractual, or time constraints, this sort of situation causes projects to be cancelled. Sometimes, however, the
situation can be salvaged by the promise of a patch update. This is no real solution. It
is just a quick fix based on a looming emergency; the negative publicity generated in
the interim between the forced release of the product and the release of the first patch
16 3634_CH14
11:24 AM
Page 378
378 Chapter 14
can damage the reputation of your company. However, if the choice is between releasing a patch at a later date and not releasing a game at all, in 99 out of 100 cases it is
better to published and be damned.
The only real way to prevent this sort of mess is to preempt it by trying to gather most
of your requirements in detail before even beginning architectural design. When doing
so, assign a task force to anticipate where expansion may be requested later. Each
potential expansion consideration has three possible outcomes: the potential requirement may prove essential and be included as a concrete requirement; the requirement
may not be essential but should be allowed for in the architecture; or the potential
requirement adds no real value for the effort needed to implement it.
Other problems can occur in the design and architecture phase, due to misinterpretation of the 80/20 rule (which states that 80% of the work should be complete before
starting on the next phase). The reasoning behind this is that it is impossible
to completely specify the design and/or architecture because some details will need
to be worked out as you go along. Trying to accomplish it all on paper is a fool’s
game, because there is no way of anticipating all of the possible interactions among
If the 80/20 rule is misinterpreted, it is possible that the next phase would have been
started too soon. Vaguely specified areas of the product are generally more timeconsuming to implement than would be expected.
Vague specifications have a number of knock-on effects. The first and most immediately obvious effect is that the schedule is adversely affected. If the project is on a tight
schedule with no room for maneuver, this can be a serious issue. The second negative
effect of vaguely specified requirements is the possibility that the proposed solutions
may not be compatible with the rest of the system! In this case, a major rework may
be required.
For some reason, it is the nature of developers to underestimate the time required to
implement a feature. A number of reasons contribute to this. Peer pressure is certainly
a consideration: being able to perform complex tasks quickly is part of the cool ethos.
The second consideration is that managers often do not want to hear what the
developer has to say if he tries to give a fair estimate.
16 3634_CH14
11:24 AM
Page 379
Troubleshooting 379
Case Study 14.1 illustrates “falling on deaf ears.” Although not presented here in the
most historically accurate form, it is based on actual events that are unfortunately a lot
more common than they should be.
Case Study 14.1 The Case of the Deaf Manager
“A most intriguing affair,” said Holmes, “is the Case Of The Deaf Manager.”
“That’s funny,” replied Watson, a confused look clouding his face. “I don’t seem to
be able to recall that one.”
Watson sat back in his chair as Holmes recounted his tale of Fothergill, the senior
“Fothergill was working on some modifications to a complex application for a
customer in Europe,” continued Holmes.
“Go on,” said Watson, scribbling notes furiously.
“This customer had appointed a manager, Snodgrass, who was not known for his
technical skills, but was well known for a short temper, and his famous technique
of trying to force people to agree with what he was saying by turning purple and
spluttering until they relented, fearing that he may explode.”
“One part of the application needed to be expanded, and a whole new module
needed to be researched, designed, and written from scratch, and the task fell
naturally to Fothergill, being the most senior of the developers.”
“The task in hand was assigned to Fothergill by Snodgrass. Fothergill took the
notes, glanced over them briefly, and put them to one side on his desk, by reason
of the fact that he was busy with some other modifications that were required.”
“Interesting, Holmes,” said Watson. “And what was the response of Snodgrass to
“Well naturally, Watson, Snodgrass behaved characteristically,” replied Holmes.
“He demanded to know how long it would take to complete the task.”
16 3634_CH14
11:24 AM
Page 380
380 Chapter 14
“And Fothergill’s reply?” asked Watson.
“Fothergill said that he did not know, as he had not looked at it. He invited
Snodgrass to ask him the following afternoon after he had had time to review the
documents in question.”
“A most sensible answer, Holmes,” agreed Watson. “This Fothergill is clearly a man
of principle.”
“Well, indeed,” replied Holmes, “but listen to the rest of the case. The following
afternoon, Snodgrass did indeed approach Fothergill with the same stated request.
And—to preempt your question Watson—the latter replied that, after reviewing the
document, he believed that it would be possible to complete the work in three
months, with an estimation error of one and a half months. In other words, he said
that it would take from a minimum of one and a half months to a maximum of four
and a half months, and a most likely completion time of three months.”
“Interesting, Holmes. And how did Snodgrass reply to this?” queried Watson.
“He replied thusly,” answered Holmes. “Snodgrass smiled, and said how pleased he
was that Fothergill would be able to complete the work in one and a half months.
Fothergill was naturally aghast at this, but Snodgrass had already departed from the
scene of the crime to tell his boss the ‘good news.’”
“A most unsatisfactory outcome, Holmes,” said Watson, somewhat dismayed.
“Were there any further developments after this tawdry affair?”
“Only that Snodgrass got very upset at being made to look stupid in front of his
boss when the module took three months and one week to complete,” said Holmes.
Despite the fact that it was impossible to accurately estimate the length of time due to
lack of knowledge of the detailed specifications and implementation requirements, the
developer still gave the best estimate possible. The manager chose to accept the lower
bound as the correct value mainly because it suited his aims, while ignoring the
possibility that this was the absolute best-case estimate, and would almost never
16 3634_CH14
11:24 AM
Page 381
Troubleshooting 381
be achievable without an improbable set of lucky coincidences. He heard what he
wanted to hear.
These are not the only difficulties to befall a project in the design or architectural phases. In the previous situation, it was clear that areas of the design or architecture were
vague, and that the 80/20 rule was obviously not followed.
Unfortunately, when the designer or architect produces an overly simple design, there
is no way to detect it by using the 80/20 rule. An overly simple design is one that fails
to adequately deal with the major issues, which inevitably leads to a need to redesign
and reimplement the project in a more robust form to meet the requirements. This is a
difficult problem to detect. It is not always obvious that a design is too simple, especially in the case of game design, where often oversimplification of the game tends to
show up fairly late in the process, usually during periods of extended playtesting.
The only really practical way to detect this before the best part of two years has been
spent implementing the game is to make extensive use of prototypes. It is always helpful to make a mock-up of the gameplay. It does not have to be extravagant, fast, or
detailed. For many gameplay prototypes you don’t even need a computer. We have
used whatever materials come to hand: pencil, paper, building blocks, dice, and so on.
This technique has also been used successfully in other cases. A very successful topdown racing game for the PlayStation and PC was designed using toy cars and a large
amount of floor space.
While sometimes it can be useful to implement the game prototype as a board game,
or as a paper-based role-playing game in order to get the mechanics correct, other
options are becoming available. Recently, more advanced prototyping solutions have
been released into the market. These are applications that provide a virtual testbed for
interactions and behaviors. A particularly good example is Virtools, details of which are
available at their Web site ( This package allows a complete 3D
world to be built and actors placed within it using behavioral building blocks that can
be expanded, if necessary, by a developer using a C++ interface.
And yet game developers often want to “roll their own.” They tend to not trust using
code and components that other outside developers have produced. This is more difficult to justify now that all game development (with the exception of small consoles
such as the Game Boy) goes through software interfaces, such as DirectX for the PC.
16 3634_CH14
11:24 AM
Page 382
382 Chapter 14
However, although these software interfaces are now unavoidable, a peculiar sort of
malaise has afflicted game developers. For these developers, the software interfaces
have become the new “metal,” the lowest level of the system that they can access.
None of them want anyone else getting between them and the lowest-level interfaces
of the system application programming interface (API).
This is beginning to change. Witness the success of the Quake and Quake II engines.
They have been licensed to many developers, and have been used to produce very successful games such as Half-Life. However, the Quake engines are still an exception to
the taboo of NIH (“not invented here”). The programmers at id Software are generally
considered the best in the industry for 3D engines, and it is this reputation and the
success of Quake and Quake II that caused the widespread demand for these engines.
As development techniques have matured, the games industry is increasingly willing,
like the movie industry, to outsource common components. Development of the leading titles is so complex that it is no longer financially practical to keep rewriting core
components—if indeed it ever was. The use of off-the-shelf components (“middleware”) is prefered by the publishers because it means that a stalled project can, if necessary, be recovered by moving it over to another developer. It is also desirable from
the developer’s point of view because time that is not wasted writing a new 3D engine
or a physics module is time that can be spent instead on crafting a better game.
Initial skepticism about middleware (the “Not Invented Here” syndrome) has been
largely dispelled by the obvious quality of games like Grand Theft Auto 3—built using
Criterion’s RenderWare suite. As David Lau-Kee, CEO of Criterion, says: “Middleware
changes the focus of game development. It becomes a matter of creating great games,
not technology.”
If the architecture, as opposed to the design, is overly simple, this presents an entirely
different set of problems. An overly simple architecture is fundamentally broken and
needs to be rectified as soon as possible. If the architecture cannot support the functionality required of it, there is no quick solution. It’s not usually possible to simply
patch the broken parts and carry on. An overly simple architecture shouldn’t have
even got through the review stage!
16 3634_CH14
11:24 AM
Page 383
Troubleshooting 383
The only real solution is to reevaluate the architecture and reimplement any substandard parts. It may be possible to add interfaces and continue the work. A trivial example would be a CD playing module that did not allow random access within a track.
This functionality would be quite easy to add without breaking existing uses of the
However, if the oversimplification were more fundamental, such as an interface for
obtaining information about a 2D map needed to be expanded for use in a 3D environment, the problem is more difficult. A simple addition of functionality may not be
sufficient. The whole structure of the module and its underlying data will need to
be reworked.
The converse of this situation is, of course, overcomplicating the design and/or architecture. Just what makes a game design overcomplicated is very subjective. One man’s
meat is another man’s poison. Starcraft might be far too complex for the casual gamer,
but a hard-core wargamer might consider it too simple.
Within each genre there tends to be an accepted level of complexity set by consensus.
This level of complexity tends on average to increase with each new release within the
genre, as more and more developers jump on the bandwagon. That’s because the developer has tended historically to be the kind of gamer who likes complexity; whereas,
the typical player is increasingly a person who wants simplicity. Very often, too much
complexity is added, and the game becomes a battle with the user interface. What
those designers do not seem to realize is that the interfaces were left deliberately simple because the game worked better letting the computer handle that fine level of control. It was simply a tedious experience for the player. The infamous water pipes from
SimCity 2000 spring to mind. The game required that you switch to a subterranean
map and lay water pipes to connect all the building zones to a pumping station. This
feature received lots of criticism in the press and on the Internet. It was simply not
necessary and lessened the enjoyment of the game. It was obvious that water pipes
would be needed, so why not just allow them to be laid automatically? Why would
someone leave an area unconnected? There’s no reason.
Expanding and redefining a genre is not as simple as adding complexity. If that were
the case, when the realtime strategy game Warwind was released some time after
Warcraft II, it would have been a huge hit. Warwind had layers of complexity beyond
anything Blizzard envisaged, and it allowed absolute control over almost every aspect
16 3634_CH14
11:24 AM
Page 384
384 Chapter 14
of unit management. Along with an obfuscatory user interface, this was too much for
the average gamer.
Contrast that with Starcraft. The game mechanics, the user interface, and the method
of unit control were virtually unchanged from Warcraft II. This was certainly a deliberate design decision on the part of Blizzard Software. Comparing sales figures within
almost every genre shows that while hardcore gamers and the press may prefer
complexity, for the typical purchaser—less is more.
This does not mean that you have to oversimplify your design to make the game
accessible. It just means that you should carefully consider your list of cool ideas and
determine if they are cool simply because they would be fun to implement, or whether
they would be fun to play. If you can do this objectively, you may be surprised by how
much the list is reduced. The point is that all those extra features represent potential
risks—so if a feature isn’t going to improve the game (and perhaps will even be harmful), there’s no reason to include it.
Any complexity in your game needs to be managed. You can do this in any number of
ways, such as hiding it behind a user interface or just reducing the complexity to an
acceptable level.
The best sort of complexity is emergent complexity, where simple rules combine in
order to produce complex outcomes, such as molecules or atoms stacking together to
form a crystal. The archetypal example would be John Horton Conway’s game, Life.
This probably needs no introduction, but, for those of you who have just landed, I’ll
describe it briefly. The game is played on a flat field of cells, each of which has eight
neighbors. Each cell is either occupied by an organism or is empty. Each game turn
produces one generation of cells, and each generation of cells is derived from the
previous one according to a set of rules.
1. If an occupied cell has two or three neighbors, the organism survives to the next
2. If an occupied cell has any other number of occupied neighbors, the organism dies.
3. If an unoccupied cell has three occupied neighbors, it becomes occupied.
Figure 14.2 illustrates these three rules.
16 3634_CH14
11:24 AM
Page 385
Troubleshooting 385
Rule 1
Generation 1
Generation 2
Generation 1
Generation 2
Generation 1
Generation 2
Rule 2
Rule 3
Figure 14.2 The rules of Life
If you are familiar with this game, then you will know that there are many complex
constructions that arise from these three simple rules, such as “shooters” and
“trackers.” This is emergent complexity in action.
The rules for Life appear simple; they are. However, the selection of these three rules
out of all the possible permutations was a considered process. The rules are very fragile.
Any changes and they do not work together as well. Some variations on these rules have
been attempted (such as basing organism survival on color, and extensions into three
dimensions), but these rules are neither as simple nor as successful as the originals.
This again shows that complexity in a game is not necessarily a good thing. The
process of designing a game is complex, but the resulting design need not be. If the
design seems inelegant, that’s probably because it is over-complex to the point that it is
likely to be problematic to implement. Some games try to address an inherent design
complexity as if it were just an interface issue, but you cannot hide an over-complex
system behind a simple user interface without losing something.
16 3634_CH14
11:24 AM
Page 386
386 Chapter 14
An overly complicated architecture is a different thing entirely, the direct result of
which is an increase in errors.
An overly complicated architecture produces unnecessary and unproductive implementation overhead. The architecture is the skeleton of the project, and the code is the
flesh on its bones. If the skeleton is misshapen, you’ll get one ugly baby as a result.
By insisting on an overly complex architecture, the architect is reducing the effectiveness of the entire team. More effort will be required to achieve a set level of work.
There will be more errors, and the resulting project will be more difficult to maintain
due to increased code complexity. This also means that there will be more “knowledge
content,” and it will be correspondingly more difficult for any one developer to comprehend it in its entirety.
Unfamiliar or Difficult Methodologies, Tools, and Libraries
Use of unfamiliar methodology results in extra training time and in rework to fix misuse.
Even the techniques presented in this book will need some take-up time; you cannot
expect to be able to instigate new procedures and techniques immediately and flawlessly.
When these methodologies are first introduced, there will most likely be a dip in
productivity of up to 25%. This can be very unsettling, and will most likely cause a
small outburst of panic. This can often be used as an argument by those who are
against the measures. Quite often, developers are initially opposed to the imposition of
measures designed to track their activities. The inclination towards a “free spirit” and
carefree existence seems to go hand-in-hand with intelligence and a healthy disrespect
for authority. Some will oppose the measures from the start, so any evidence that productivity is being compromised will be seized upon as a good reason to return to the
warm familiarity of the “old ways.”
It’s possible to circumvent this reaction by preparing the development team. For any
new methodology, tool, or library, there will be a learning curve. Don’t let skeptics
knock anything new until it has had a sensibly long proving period. Not everything that
is tried may work, but you’ll never know if it is shot down in flames right at the outset.
As an example, if part of the project is implemented in a low-level language (such as,
Assembler) it is very likely that productivity will be lower than expected. In general,
16 3634_CH14
11:24 AM
Page 387
Troubleshooting 387
however, for the more powerful machines, Assembler is rarely necessary. Usually the
machines are already fast enough. In the case of the Pentium processor, and particularly when running a multitasking operating system, you can never be exactly sure how
long a particular piece of code will take to run.
The only machines where Assembler is really justified are the limited-memory,
low-power machines such as the Nintendo Game Boy and other hand-held consoles.
Although Nintendo (and other companies) try to keep development information for
use only by registered developers, an underground following always seems to spring
up. There is even a freeware C compiler available for the Game Boy, and many
emulators with excellent debugging facilities.
The development times for these machines with more-limited specifications tends to
be shorter than full-blown PC and console projects, mainly because the programs are
so much smaller. The Game Boy is about the only platform left where a lone person
can be realistically expected to be able to produce a game that can match up to those
of the big boys.
Architectural Integration Problems
One of the main dangers with developing components separately (as suggested in the
software factory model) is that, if the procedure is not managed with the utmost care,
there can be difficulties integrating them.
This problem affects the whole software industry, which is one of the reasons for the
increasing acceptance of object-version technologies such as Component Object Model
(COM) and Common Object Request Broker Architecture (CORBA). These would, in
theory, be good models for game development, except if you were aiming for a multiplatform release.
In the case of PC development, COM is certainly worth considering. There is no need
to worry unduly about a performance hit, because the whole of the DirectX library is
based on the COM system. Although this will be covered more in the third part of the
book, COM allows versioning of objects by requiring that developers guarantee that
the behavior of an object interface must remain unchanged. All objects would have to
be obtained at runtime by calling a standard function with a guaranteed-unique ID
number that represented that interface. If the interface needed to be upgraded, a new
16 3634_CH14
11:24 AM
Page 388
388 Chapter 14
interface number would be assigned, and, just as importantly, the old interface would
need to be maintained unchanged. This allows upgrading of shared components on a
machine without breaking applications that use an earlier version of that interface.
COM is certainly not a panacea, but it does solve few of the most common problems
with integration of components.
Currently, COM is effectively restricted to Windows-based machines, but Microsoft has
stated that it will be promoting COM on other platforms as well. It’s possible that COM
will, in the future, become an industry standard and may be suitable for cross-platform
Schedule Threats
Schedule threats are often the most insidious type of threat and the most difficult to
detect and control.
Schedule slippages are not necessarily the fault of the developer. The chain of
responsibility stretches all the way up the hierarchy of command, and the blame may
lie anywhere. The following sections give some examples of slippage, their causes,
and how to deal with them. The main problem with slippage is that it sometimes goes
unnoticed. It’s only when all the slightly missed deadlines are totaled up that the odd
day here and there turns into weeks or months.
Too-Tight Schedules
Most schedules are tight. Giving too much time to a team doesn’t necessarily produce
the pinnacle of perfection you may expect. Even when the resources are there for a
team to be able to say, “It’ll be released when it is ready,” a tight control is needed to
ensure that time is not squandered.
Deadlines provide focus. Nothing focuses the mind of an employee better than the
knowledge that he has definite short-term aims to achieve before a certain date and
time. Unfortunately, for various reasons, this is sometimes exploited and taken to the
extreme by managers who decide a schedule needs to be impossibly tight in order to
prod their supposedly “lazy” developers into action. This always has a negative effect.
Developers are smart. They wouldn’t be developers if that weren’t the case.
11:24 AM
Page 389
Troubleshooting 389
Schedules can be too tight for a number of reasons. Sometimes, schedules can have a
fixed completion date due to outside reasons, Christmas being the most common.
We know it is supposed to be the season of miracles, but, even so, the number of
hopelessly optimistic schedules that spring up at the beginning of a year, trying to
get a new hit out in time for Christmas, never ceases to amaze us.
The only solution to a fixed-schedule problem is either to reduce the product specification, increase the resources available, or increase the time available.
These three attributes (time, resources, and requirements) can be viewed as three
points of a triangle, similar to that shown in Figure 14.3. If you imagine the triangle is
balanced about a point in its center, it is clear that you cannot alter the “weight” of one
of the corners without having to modify the others. To keep the triangle from collapsing, you have to keep the center of gravity over the center of the triangle. So, if you
shorten the schedule, you will have to increase the resources or simplify the specifications. If you decrease the resources, the specifications will need to be cut down, or the
schedule will need to be lengthened. And so on.
16 3634_CH14
Figure 14.3 The impossible triangle of resources, schedule, and specifications.
This is an immutable law. Something has to give. Getting employees to work long
hours in order to meet an impossible schedule is counterproductive. Not only will
they begin to burn out, but there will be a negative effect on morale. They may even
leave the company, an outcome that should be held against the manager responsible in
the same way as a large financial loss would be.
16 3634_CH14
11:24 AM
Page 390
390 Chapter 14
Another cause for concern is the “barely possible” schedule. This has been hinted at
earlier in this chapter (see Case Study 14.1) and is where a schedule has been produced that is fairly accurate, except that it is the optimistic “best case,” rather than
realistic, “expected case.”
These “barely possible” schedules usually happen for two reasons. The first is that the
employee who produced it is inexperienced. He or she will usually try to impress management by producing a good schedule that fits what the management would like to
see, with the assumption that there will be absolutely no problems during the development period. The second—and worse—cause of “barely possible” schedules is that
managers will sometimes ask for further cuts to be made to the schedule.
This does no favors to anybody, and it is usually the development team that gets the
blame for the schedule’s slip. There are no simple answers to how to handle this. The
only way to really deal with it effectively is for the scheduler to create a realistic schedule in the first place, and then stand his or her ground if management asks them to
reduce the length. The assumption usually is that all developers pad their schedules, so
they can just cut out some of the padding and produce the “real” schedule. This is an
alluring argument that must be resisted at all costs. You cannot just reduce a schedule
and expect that the same amount of work will just be produced faster. Something has
to give, and that something is either the scope of the project, or the amount of
resources assigned to it.
If you are asked to reduce a schedule unreasonably, try putting it this way: They can
have it quickly, but it will cost them more, and the functionality will be less—with an
accompanying increase in risk. The mention of more money, reduced functionality, and
increased risk tends to focus the mind pretty quickly, and usually another solution will
be found.
Incomplete Schedules
Another problem that can occur with inexperienced schedulers is if they omit tasks.
Maybe they forgot that all the artwork would need additional processing or that a game
needs an install and an uninstall procedure. But it is very difficult to predict a list of
tasks that will need to be performed over a two-year period. The fact is that, if a task is
not on the task list, it will not be scheduled for. This results in the nightmare situation
of an incomplete project that, according to the schedule, is actually complete!
16 3634_CH14
11:24 AM
Page 391
Troubleshooting 391
Unavailable Resources
The schedule might be based on the use of specific team members, but those team
members might not become available.
If a particular team member has skills that are required and he or she is busy elsewhere, there is nothing that can be done apart from waiting. This situation should
never be allowed to happen. It is extremely dangerous to allow one person to have a
monopoly on a particular skill set. What happens if he or she decides to leave? Never
get yourself into this situation by encouraging the spread of skills between members of
the team.
Overestimation of Schedule Savings
If certain productivity tools (such as advanced prototyping tools) are used, they are
often viewed with a sort of “new-world” style shock and awe.
Often, no doubt dazzled by the multitude of knobs and buttons available, otherwise
rational developers become convinced that they can perform superhuman feats, leaping
tall buildings in a single bound, rescuing orphans from collapsing buildings, and
single-handedly producing a fully working game prototype in a ridiculously short
period of time.
Sometimes the tool will appear to have all the features needed to implement whatever
it is that needs doing. The problem is that the more “help” a tool gives you, the more it
will railroad you into its methods of doing things. Quite often, unless you are extremely lucky, it will not be quite what is wanted, and you will spend your time working
around the perceived shortcomings of the tool. The difficulty of doing this can vary
from mildly taxing to the completely impossible. Granted, most serious tools allow a
developer to produce an add-in module, but this involves designing a module to fit in
with an unfamiliar architecture, learning a new API, and discovering how to actually
perform the task within the confines of an unfamiliar design program. It is often
(incorrectly) assumed that because a standard language such as C++ is being used to
create the add-in, a good C++ programmer should be able to rattle it off fairly quickly.
The problem is that the time needed to learn the system is often discounted.
The solution is to either reschedule to allow learning time, or to realize the limits of
the “productivity-enhancing” tool. Of course, you can also drop the idea of using it
16 3634_CH14
11:24 AM
Page 392
392 Chapter 14
Inaccurate Schedule Estimates
In some circumstances, a schedule can look exactly right. It can have every task
labeled, every risk considered, and every developer busy with out dependency conflicts. And, yet, when it comes to putting it into practice, one or more areas do not
adhere to the schedule. For example, unfamiliar areas of the product may take more
time than is expected to design and implement, or a delay in one task may cause
cascading delays in dependent tasks.
This could be because the product is larger than has been estimated, or the amount
of effort required is greater. This assumes that the difficulties are purely due to the technical complexity of the solution and have nothing to do with on-the-job research being
required. If it is the latter, then there is no solution. You just have to wait until the
research has concluded or cut the functionality. Either solution is usually intolerable, but
that is the price you pay for allowing scheduled research. You cannot schedule research.
For instance, how ridiculous would it sound to commit to a schedule of six months, with
detailed checkpoints and mi1estones, for inventing an antigravity drive? You would have
to have full knowledge of the problem and the solution before you even started to be able
to accurately estimate how long it is going to take to complete. Research is, by nature, an
investigation of the unknown. Scheduled research is an oxymoron.
If, however, the problem is due to unexpected complexity, there usually are only three
so1utions. One is to cut the functionality. This may or may not be practical depending
on your needs. The second is to reschedule the project, making allowances for the
extra time required. The third solution, which is based on the software factory methodology, is to replace the component with a compatible (but maybe less capable) one
from the library. This may or may not be suitable for release, but, if all else fails, you
will still have a working product. However, that is not the main point. The main point
is that even if the slotted-in component does not offer all the functionality needed, it
will at least act as a stopgap while the new component is being worked on. Other parts
of the project can still continue, if a little cautiously, and not as much time will be lost
due to having to wait for a critical component to be completed.
16 3634_CH14
11:24 AM
Page 393
Troubleshooting 393
Persuading people to work longer hours is another option. If the effort required for
a project has been only slightly underestimated, short bursts of overtime may be a
sensible way to get back on schedule, as long as the developers are paid for it.
Unfortunately, this system has been long open to abuse within the games industry.
We know countless horror stories of developers being made to work 80-hour weeks for
months on end for no extra monetary reward (except maybe a pizza on weekends and
a cheesy award and T-shirt at the end of the project). This is ridiculous. Developers
are the people who make the product on which this industry depends. They should
be treated with respect and be subject to reasonable working conditions. Nobody
should be forced to work for longer than eight hours a day for extended periods. It is
counterproductive, and all that results is a substandard game and a whole bunch of
disgruntled, burnt-out developers.
The main excuse used to justify this (and note that the managers do not usually come
in for the weekends) is that writing a game is about sacrifice, or some other related
falsehood. Balderdash! It’s a job, pure and simple, and anybody who tries to convince
you otherwise is deluded—or stands to make a lot of money out of your sacrifice. To
summarize, voluntary overtime is okay for short periods, as long as the developer is
paid accordingly.
Schedule Adjustment Errors
Adjusting a schedule to account for some unexpected delay is often subject to errors.
Sometimes reestimation in response to schedule slips is overly optimistic or ignores the
project history or metrics gathered from other projects.
The best way to ensure these errors do not occur is to use past information and metrics
on the nature and performance of the employee concerned.
Case Study 14.2 illustrates these procedures in action.
16 3634_CH14
11:24 AM
Page 394
394 Chapter 14
Case Study 14.2 Applied Schedule Readjustment
Andrew was engaged to analyze the schedule on an overdue project and readjust it
to estimate an accurate completion.
It was not possible to do this without first gathering some metrics from the team.
To do this, Andrew observed the tasks that each of them had been assigned and the
time that they had been allocated to do them. He began by assuring them that they
were not being blamed for the problems and that he was there to reassess the
schedule and produce a more accurate prediction. One of the reasons that the project had run into trouble was that, as the pressure mounted, procedure had been
abandoned in an attempt to meet deadlines. (This is covered later in this chapter.)
The result of this was, of course, to compound the problems and cause even more
bugs and schedule slips. At this stage, they stopped any new development and were
concentrating on bringing the project back up to acceptable levels. This work had
been going on for about two months and was nearly complete.
This was the right time to begin adjusting the schedule because new development
was about to start. If Andrew had tried to do it beforehand, the results would not
have been as useful because extra work would have been needed to work around
the problems with the old codebase.
The team members were assigned new tasks, each of which (according to the
schedule) was expected to take one or two days. The team was asked to strictly
adhere to procedures. Andrew emphasized that this was not a race and that they
were not being assessed for proficiency. Every individual has a different working
rate, and just because a coder is fast does not mean that he is necessarily good.
Andrew required the information on their rate of work solely because he needed to
adjust the schedule to match it.
Andrew observed and collected metrics from the team over a couple of weeks. He
found that, without exception, all of the team failed to hit the deadlines. The
schedule was too aggressive.
For each of the team members, he worked out the ratio between the actual time
taken and the scheduled time for the sample activities. Having done this, he took
16 3634_CH14
11:24 AM
Page 395
Troubleshooting 395
the task list for each individual, and multiplied the scheduled times by this ratio
to produce a more accurate estimate of expected times. With the guidance of the
team, he then shuffled tasks around to make sure that each individual had a
balanced task list.
The whole process was repeated for another week, and at the end of that week the
ratios were again calculated and the tasks reshuffled.
After the third week, the developers were consistently hitting the deadlines, and
strictly adhering to procedure. Morale, by this stage, had already improved considerably. The team had gone from thinking that they were slow and inefficient to
realizing that the schedule they had been working to was unachievable.
The next obstacle was explaining to eager management that the expected time of
release for the product according to the new schedule was two months later than
they hoped. Andrew was asked if he could trim the schedule somewhere to cut
those two months out.
This dismayed him because it was the sort of thinking that got the management
into this mess in the first place. He drew a diagram of a triangle on the whiteboard
(similar to that shown in Figure 14.3) and explained the delicate balancing act
between resources, specifications, and schedule. He couldn’t trim the schedule
because there was no room for maneuver. They would either have to add resources
or cut functionality and be prepared to deal with the complications of doing this at
such a late stage.
Adding resources or cutting functionality was unacceptable to them because they
had a contracted features list and budget, so in the end, they had no choice but to
go with Andrew’s recommendations.
The project was finished within one week of the newly estimated schedule.
The excessive pressure from an unachievable schedule reduces productivity, mainly
because it saps the developers’ morale, as they see themselves slipping constantly
behind despite their best efforts. The only solution is to consider actions such as those
presented in Case Study 14.2.
16 3634_CH14
11:24 AM
Page 396
396 Chapter 14
Organizational Problems
Organizational problems are the most difficult to deal with effectively. They are usually
a direct result of management error and are difficult because telling your boss that he
has made a mistake is usually the quickest route to the door.
These situations have to be handled with the utmost care. Telling your next-most senior
manager doesn’t usually work, because, if the news is bad, he or she may feel that telling
the big boss will be bad for his or her career! Everybody loves to shoot the messenger.
However, put yourself in the boss’s position. Would you rather be told about a potential problem a couple of months into initial prototyping, or would you rather be told
by a developer several months past the delivery date that he had tried to tell you about
the problems over a year ago, but couldn’t get it past his immediate manager?
The solution is to have an anonymous channel. That way, anybody can report problems without fear of repercussions on their career.
Management-Induced Difficulties
Management causes most organizational problems. This statement is not managementbashing, it’s just a natural consequence of the fact that management is responsible for
the organization of a project.
For example, management (or even the marketing department) may insist on technical
decisions that cause the schedule to be lengthened. It is unfortunate that the market is
driven by the technology, but there is not a lot we can do about that. It may be that an
OEM deal means that the product has to support a particular graphics card natively,
rather than through a more generic API such as DirectX.
In some circumstances, management insists on reviewing decisions such as purchasing,
budget approval, legal matters, etc. This situation usually occurs when a large publisher owns a stake in a smaller game-development company. One of the conditions these
publishers try to impose is the “right of veto.” This means that every decision that
affects the product has to be ratified by the parent company. If the parent company has
many subsidiaries, the chances are that this could be a significant cause of delay (a
cause that the publishers are bound to forget when they are yelling into the phone at
your manager, demanding to know why the game is three months over schedule).
16 3634_CH14
11:24 AM
Page 397
Troubleshooting 397
This is the sort of thing that can affect the morale of your team. Other management decisions can also lower the morale, such as extended forced overtime or reduced privileges
for no good reason, or an inefficient team structure that further reduces productivity.
Surprisingly, it is also the case that good management decisions can (at least temporarily) reduce morale. Such decisions include the imposition of procedure and defined
development practices. People feel uncomfortable with change. This has to be handled
with a softly-softly approach and should not be abused. There is no point in implementing procedures to allow the team to work more efficiently if you are then going
to saddle them with compulsory overtime and expect them to work efficiently for ten
hours a day. The whole point of implementing strict development procedures is to
make the team more efficient. If management is going to implement measures that
cause a dip in the morale of the team, there had better be an easily visible benefit for
the team pretty soon.
Some of the old school of management may not understand the more sedate measured
pace that procedure brings. Remember that the games industry has a decade or more of
bad habits behind it, and this is not going to change overnight. Some managers will get
nervous at the slow but steady pace and actively discourage it. They may wish to see
more cutting-edge heroics, with pretty demos and clearly visible progress rather than
steady work that may not show any immediately visible whizzy new effects on screen.
If this is the case, the team will be putting too much effort into only one area of the
project in order to satisfy that manager, and the development will become off-balance,
resulting in shortcuts and unnecessary compromises further downstream. This sort of
off-kilter development, concentrating on external appearances only, prevents accurate
status reporting, which undercuts the team’s ability to detect and correct problems
effectively. This is likely to be compounded by extra pressure as the project attempts to
draw to a conclusion and the team finds that important subsystems are either incomplete or missing. In this sort of situation, it is not uncommon that project plans are
ignored due to the time pressure, resulting in chaotic, inefficient development, as
described in Case Study 14.2.
16 3634_CH14
11:24 AM
Page 398
398 Chapter 14
Contractor Problems
Surprisingly (or maybe not, seeing as how the average games industry wage is much
lower than the equivalent outside the industry), contractors are not used very much in
the games industry. While it is common for an organization such as a bank to hire contract programmers for the duration of a project, this is virtually unheard of in games.
The typical use for contractors in the games industry is to allocate complete units of
work, such as FMV production, motion capture, playtesting, and configuration testing,
the conversion of an existing game to a different platform, or the production of 3D
Some of the larger companies enter into contracts with smaller game-development
houses, and sometimes you will find that the parent company will ask you to use the
engine developed by another subsidiary to produce a new product. A number of
obvious problems can arise from this.
Late Delivery
If the contractor does not deliver the specified components by the agreed date, a
workaround will have to be found, and this may even put the project on hold.
Penalty clauses for late delivery, although a good idea on paper, tend to be counterproductive. Unless they are sufficiently forbidding, there will be very little incentive to
continue work if a penalty is threatened. And, of course, you may not always be in the
position to insist on such measures. The best way to prevent this problem is to avoid
being in this situation in the first place. If this isn’t possible, you either need to sweat it
out and wait, cut your losses, and cancel the agreement (and most likely the project if
you cannot find a replacement component), or begin development of your own
replacement component.
Poor Quality
Usually, the contractor will have his or her own development standards and practices.
There is no way you can guarantee that the work will be delivered on time and be of
acceptable quality.
16 3634_CH14
11:24 AM
Page 399
Troubleshooting 399
In this case, time will need to be added to the schedule to allow the quality to be
improved. In other words, this is the same as if they were delivered late in the first
place. (The situation where the work is delivered late and is of poor quality is a nightmare that doesn’t even bear thinking about.)
You might be able to build explicit quality guarantees and a concrete set of specifications into the initial contract. Make sure that your requirements are definite and there
is no danger of misinterpretation. Encourage the contractor to maintain a dialogue
with the person who created the specs, so that there can be constant monitoring and
Don’t just throw the requirements out of the door and expect exactly what you wanted
to come rolling back in with a flourish some months later. You will be in for a tragic
The other important thing to make sure of is that the contractor has sufficient motivation (usually of the monetary kind) to perform the task required. The danger here is
that, if the contractor does not buy into the project due to insufficient motivation, they
are unlikely to provide the level of performance needed.
Personnel Problems
Personnel problems are inevitable. Nobody can be expected to get along with everybody all of the time. It’s like the old saying, “You can please some of the people all of
the time, you can please all of the people some of the time, but you can’t please all of
the people all of the time.”
This is a fact of development life. Get used to it.
Like giving birth, developing a game can involve a lot of pain. The pain and stress of
the development process can sorely tax relationships between people who have spent
every day of the past couple of years in each other’s company. Think about it. You are
effectively spending a minimum of eight hours a day, five days a week in the company
of a group of headstrong, intelligent, and often young co-workers in a situation of
great stress. This is not an environment conducive to peace and harmony. In the natural course of events, there are likely to have been a few arguments and bust-ups along
16 3634_CH14
11:24 AM
Page 400
400 Chapter 14
the way, and that is just between team members. This can obviously damage productivity. If the team’s members are too busy feuding, work will not get done as efficiently.
Worse, if the feuding escalates to open warfare, there will be casualties, usually in the
form of code sabotage or forced resignations. Sabotage can result in lost work or poorquality code that requires rework. This is an incredibly damaging situation.
If team members do not work together efficiently, the resulting conflicts cause poor
communication, poor designs, interface errors, and extra rework.
If this is the sort of damage that employees at the same level can inflict upon each
other, just imagine the abuse that can be heaped upon employees by an unscrupulous
and domineering manager. Poor relationships between developers and management are
some of the most detrimental to morale and productivity. They promote an “us and
them” situation, which often has the dubious benefit of uniting the team all right—but
in the wrong way.
The net result of these problems is that low employee motivation and morale reduce
productivity. Team members do not buy into the project and consequently do not provide the level of performance needed to make it a success. It can be disastrous if disgruntled employees leave before the project is complete. The team then has to learn
and shoulder the responsibilities of the missing member (or members), lengthening
the schedule and increasing stress levels. One pseudosolution to this is often to add
new development personnel. However, if they are added late in the project, the additional training and communications required adds a lot of overhead that reduces the
effectiveness of existing team members.
Skills Shortages
In the games industry, there are a number of perceived developer disciplines (such as
3D programmer, physics programmer, and AI programmer). A skills shortage can arise
in even the best-staffed project. Sometimes, a programmer will be called upon to work
in an area in which he or she may not have much experience.
This problem does not apply just to developers. In general, when people’s assignments
do not match their skill set, problems arise. Usually, the lack of specialist knowledge
increases the mean number of defects in the work (be it code or otherwise), and the
amount of rework needed.
16 3634_CH14
11:24 AM
Page 401
Troubleshooting 401
Not all companies have the resources to provide full training to everybody who needs
it, and, in fact, there do not seem to be that many training courses specifically for
game-industry disciplines, unlike other areas of the software industry.
Three general approaches can be used to solve this problem. The first is to allow extra
time for employees to familiarize themselves with the areas they will be working in.
Under some circumstances, this may be practical, but, under others (such as a limited
amount of time or money), it may not. The team members selected for this should, at
the very least, have the base set of skills to allow them to learn the new material efficiently. For example, there is no real use in asking a nonmathematical programmer to
investigate and design a new 3D engine.
The second solution is to encourage knowledge sharing (as in the software factory.) In
this way, employees with specialist skills should be encouraged to pass on their knowledge to other employees by direct teaching. An even better approach would be to have
a regular set of seminars provided by the more skilled employees with the sole aim of
passing on their experiences and skills to other employees. This is cheaper than
employing an outside lecturer, and may be more relevant to your company ethos.
Employees like to feel as if they are being taken care of, and an optional educational
program shows them that their needs are being considered.
The third solution, for those with no time but plenty of money, is to hire an outside
contractor with the required skills. This can be expensive, but if you need the skills
that badly, you may have no choice. You can always offset your losses by asking the
contractor to teach other employees while he or she is there.
We have said that contractors can be expensive. For long-term use, that is true, but in
situations where you need a small amount of highly specialized work, employing a
contractor is probably the most cost-effective solution. Contractors with the correct
skills also tend to be easier to find than permanent employees.
Development Problems
In this section, we are not going to focus on coding problems. This book is not about
coding. No, here we are going to concentrate on the problems that affect a developer’s
ability to code.
16 3634_CH14
11:24 AM
Page 402
402 Chapter 14
Developers tend to be an unfussy bunch when it comes to environment. You can sit
them down in a darkened office, with only the glow of the monitor to provide light,
and, as long as they are undisturbed, they can code quite happily, without worrying
about their surroundings. At least, that is the theory as management often sees it; in
practice, it is altogether different.
Office Facilities
Problems with offices can be a significant contribution to delays. The whole point of an
office is to provide a stable working environment for development.
If, for example, the office facilities are not available on time, where can the team do its
work? Good question, and one that does not have an easy answer. It depends on the
office facilities. If the team is part of a large company, maybe they can relocate to
another part of the building. If the team is part of a small company, maybe a startup, it
is possible that the offices they have rented may not be finished in time. Nobody tends
to move offices halfway through a project, so the probability is that the team is about
to start a new project (or maybe the first) when the move takes place. This means that
the start of the project may be put back, but, if the delay is short, it shouldn’t cause too
many problems.
If your company doesn’t have an office at all, you may have a bigger problem. You
could always take a leaf out of the Yost Group’s book. (Yost is the developer of 3D studio
max.) The first release of 3D studio max was developed with the entire team in separate
locations. The use of a virtual office such as this presented some problems, mainly
to do with transmitting large files across a standard modem, but, because of the objectoriented nature of the application, this developer partitioning worked fairly well.
If office facilities are available but are inadequate (for example, there are no phone
lines, network wiring, desks and chairs, office supplies, or computers), the only
solution is to take out the checkbook and go buy the stuff.
The office could also be noisy or crowded, or disruptive in some other way. Under
these circumstances, the imposition of various ground rules should help, such as a
clear desk policy (where desks are left clear at the end of each day), and measures to
encourage a quiet environment. Provide meeting areas so that people do not have loud
discussions where others are trying to work.
16 3634_CH14
11:24 AM
Page 403
Troubleshooting 403
Development Tools and Third-Party Libraries
One of the problems with development is that you are invariably dependent on thirdparty development tools and libraries.
If these are not as productive as expected developers will need extra time. Sometimes
the fault is in the choice of development tool—perhaps they were selected not for their
technical merits, but because they were cheap or had the coolest marketing. The only
other option, apart from implementing workarounds, is to switch to a new tool (if one
is available). Quite often, this will bring its own new set of problems, not to mention
the delay and difficulties caused by transferring the entire project to the new tool.
If code or class libraries are of poor quality, then extra testing, defect correction, and
rework will be required. Worse still, problems in an external library are usually hard to
find, especially if the source code is not provided. Sometimes, you cannot avoid using
an external library, so there is no choice but to work around the problems. For example,
can you imagine writing a commercial PC game without the use of DirectX? Any problems in DirectX that affect your project will need to be worked around. Reporting the
bug may get it fixed, but the chances of it being fixed and released in time for your
game release are low. Even if it is fixed, many machines will still be using older versions.
Even though you can distribute the runtime of DirectX on your distribution CD, you
cannot be sure that everybody has updated their hardware drivers (which are generally
released by the manufacturer of the hardware). Fortunately, DirectX seems to be working fairly well now, although there were some compatibility horror stories from the early
days. However, if a library that you are forced to use is of unacceptably low quality the
code that uses it will require more work to correct than would be expected.
Misinterpretation of Designs
Even with the best of intentions, developers will sometimes misinterpret design documents and produce output that bears no resemblance to the software that was requested.
It is possible that the developer simply made a mistake and did not understand the
design correctly. This is understandable, but it is not necessarily excusable. If the
developer was unsure of any concepts in the design document, he or she should have
asked for clarification before beginning to develop the code. A developer shouldn’t be
developing something he does not fully understand.
16 3634_CH14
11:24 AM
Page 404
404 Chapter 14
A slightly more serious situation is where the “mistake” was deliberate. He or she may
have decided to modify or rewrite the design to his or her idea of a superior implementation. It may be only a small change or it may be a large one. The end result, however,
is a divergence between documentation and code that cannot be allowed to happen.
Even if the developer is conscientious enough to update the design document, the
modifications will not have gone through the approval procedures and may not be in
line with the rest of the project. Worse still, there may be other modules that rely on
the original implementation.
Development of extra software functionality that is not required, otherwise known as
gold-plating, can extend the schedule. In the case of gold-plating, the developer implements the functionality required, but feels the urge to add extra functionality, because he
or she perceives it to be useful and it requires little or no extra effort to implement. The
developer implements it because he or she thinks that it is more functionality for free.
There is one lesson that everybody learns sooner or later, and that is nothing is free. If a developer
adds extra functionality, the cost will become apparent later during maintenance and bug fixing.
Meeting the Requirements
Requirements may be imposed on the product by the game designer, publishers, marketing, and external organizations such as ratings boards or supermarkets. These constraints
may or may not be adhered to depending on the nature of the company’s political environment, but, even so, there are bound to be some problems caused by difficult or conflicting
You may wonder why a supermarket has an influence on a game, but it is a well-known
fact that Wal-Mart has made stocking decisions based on game content. Not being
stocked by Wal-Mart is viewed seriously by most publishers. If the man from Wal-Mart
says “No,” you can wave goodbye to a large chunk of potential sales. Understandably,
many publishers will insist that their products are supermarket friendly.
There could be other more basic requirements that may be difficult to achieve. For
example, meeting the product’s size or speed requirements may need more time than
expected, including the time required for redesign and reimplementation. If a product
is fundamentally slow, optimizing it to meet the speed requirements can be a tedious
16 3634_CH14
11:24 AM
Page 405
Troubleshooting 405
If the game is developed using a beta version of a necessary library, there could be
problems when the final version is released. It is possible that not all of the features in
the beta version have made the release, or even that workarounds used to avoid problems with the beta version simply break with the release version.
If there are strict requirements for compatibility with an existing system (for example,
a related series of games relying on the same underlying data, or a multiplayer game
that needs to cooperate with a third-party online server interface), the system will
require much more testing, design, and implementation than expected. A good
example is Blizzard’s server software, a free Internet multiplayer server
for use with Blizzard’s games, Starcraft and the two Diablo games, among others. In
general, requirements for interfacing with other systems that may not be under the
team’s control will result in unforeseen design, implementation, and testing.
If your team is aiming for cross-platform compatibility, then the implementation and
testing will take longer. Many things can go wrong with cross-platform development,
including subtle differences in so-called “standard” APIs and differences in processor
Research: The Consequences
As mentioned before, research, by its nature, cannot be scheduled. The best you can
do, if you are forced to research as part of the schedule, is to set a cut-off date and
make sure you have a contingency plan.
That is the main problem with research: Pushing the game’s state of the art will
inevitably and unpredictably lengthen the schedule, and you cannot rely on any guarantees from optimistic developers that it will be finished by a certain date, as seductive
as those arguments may seem.
Other areas of the project may also be dependent on the module being researched and
developed. If no suitable contingency plan is in place, the only option will be for all
the dependent projects to wait for results.
Relying on the outcome of research in a scheduled project is just asking for trouble.
16 3634_CH14
11:24 AM
Page 406
406 Chapter 14
Process Problems
The previous chapter discussed “process” in great detail and how it could be used to
make the lives of the people on the team easier (despite their initial protestations).
We are going to briefly revisit that subject again here, and discuss how process can be
misused, so that it ends up being detrimental to the project, and additionally makes
the lives of the team a misery.
Red Tape and Bureaucracy
Some managers take process to heart…too much so, unfortunately. They will attempt
to saddle you with three pieces of paperwork to sign if you wish to sneeze. Process
(after an initial learning curve) should never amount to a chore. It is not usually too
difficult to do; the only difficulties are the initial shift from a “no process” to a “some
process” environment.
In some cases, however, there can be too much paperwork. If a task generates an inordinate amount of paperwork, it should be examined to see if it can be streamlined.
“Process” is supposed to help the employees, not hinder them. If they are filling in the
same information in multiple places, the information-gathering process should be modified. The object is to make paperwork as simple as possible, so that it is more likely to get
done and more likely to be correct. Nobody wants to have to type in the same old information several times for the same task. Too much paperwork may slow progress. Care
should also be taken to ensure that the employees do not have to spend too much of their
time reporting information to management: instead of scheduling a meeting, which is usually unnecessary for a mere progress report, just send the report by email to all concerned.
Misuse and Metrics
There is no point in having a comprehensive set of procedures if they are not used
correctly. It is in the nature of developers to try to avoid work that they assume is nonproductive, and this includes quality procedures, unless their benefits are precisely
These can be abused in a number of ways. For example, if the initial upstream qualityassurance activities are not performed properly (the idea being that they are skipping
them to “save time”), the net result is that all the apparently “saved” time (and more)
will be spent doing time-consuming rework further downstream in the development.
16 3634_CH14
11:24 AM
Page 407
Troubleshooting 407
If the statistics and metrics are not being tracked and reported correctly, early-warning
signals may go unnoticed. Quality problems that affect the schedule may be ignored
until very late in the project.
Metrics should be considered as very important sets of statistics. They are the only
consistent way of leveraging knowledge learned in one project so that it may aid the
development of the next. This information is too valuable just to be ignored, and
should be recorded accurately and consistently. If this data is not being collected
adequately, this will result in not knowing exactly how far behind schedule the
development is until late in the project.
If the risk-management program is not making good use of these metrics, there is a
good chance that the half-hearted risk management will fail to detect major project
risks before they arise, and the project may founder. There is no point in buying a
smoke alarm if you are going to leave it in its box without any batteries.
Problems with Formalities
The amount of formality required in the average software project is much more than is
usually seen in the games industry, and this is not a good thing for the games industry.
Too little formality (ignoring the development procedures) will result in a lack of communication, poor quality output, and large amounts of rework. Too little formality is
usually worse than no formality at all, as it gives the employees a false sense of security.
If there is no formality at all, at least they will know what to expect.
The converse situation can also be a problem: Too much formality (overly bureaucratic
insistence of sticking to the letter of the law for procedures and standards) will result
in unnecessary and time-consuming overhead. Employees will be spending most of
their time completing and processing paperwork.
A certain level of process is definitely a good idea. The benefits that it provides for
your project far outweigh any disadvantages from the extra work involved. The trick is
to find the balance.
16 3634_CH14
11:24 AM
Page 408
17 3634_CH15
11:27 AM
Page 409
Chapter 15
The Future of the Industry
• The hypothetical
“mass market”
• Changing management
• The film industry model
his chapter contains our best guess at what the future
holds for the games industry as a whole and what will
happen to the games market.
The games industry is still fairly young. We’re through most
of the teething troubles and the terrible twos. We’ve passed
through that golden age when everything was feasible and
anybody could design and write a hit game. We’ve even
passed through the age of back-bedroom programmers and
alleged millionaire teenage developers
Now, we’ve reached the adolescence of the industry.
Everything is still funky and cool, but we’re less willing to
experiment with the more outlandish ideas, and we’re starting to cast an eye to crass commercialism—mainly due to a
consolidation of the marketplace in a similar fashion to the
movie industry.
The State of the Industry
The games industry has evolved much as any ecosystem
does. The early proliferation of forms—games about making
pizzas, running hospitals, dating, politics, and plumbing—
has settled down to just a handful of distinct genres.
Likewise, advances in technology seem to be converging
towards a common point. On the latest generation of consoles, all games are starting to look pretty much identical in
terms of graphical quality.
• The maturation of the
• The online revolution
• The possibility of
17 3634_CH15
11:27 AM
Page 410
410 Chapter 15
The First Era
The industry was built by many talented amateurs. In the early days, an enterprising
enthusiast was able to concoct a game in his spare time and sell it for a reasonable
amount of money. These were the days of weird and wacky plot ideas. Literally, any
idea would sell—as long as it was a good game.
However, diversity didn’t always mean good games. Some attempted to cash in on
the new craze for home computer games by releasing shoddy products. As the homecomputer market grew, the competitiveness among companies increased, and the less
efficient organizations began to get squeezed out of the market.
At first, this was a good thing. The people buying the games were well informed,
computer literate, and generally knew what they were buying. Any game companies
that consistently produced low-quality merchandise soon foundered. Quite a few
companies fell this way.
This left the fittest and the strongest companies vying for discerning business in a
competitive market. These companies had to find a way to differentiate their product
from the competitors’—to make it stand out on the shelf and be more desirable.
The most infamous method was the tie-in license: the game of the film or book.
Almost without exception, all of these licensed releases were mediocre products rushed
out to catch the wave of the movie’s publicity. At the heights of the licensing feeding
frenzy, diverse products such as potato chips, breakfast cereals, and soft drinks featured
in their own games. In regard to originality, this was one of the worst periods of the
mainstream industry.
Its one benefit, however, was to further thin out the worst companies, leaving us with
only the cream of the crop, producing technically excellent (if not necessarily original)
This is how things were at the end of the first era of home computer use. With the
imminent release of more powerful desktop machines and mass-market consoles,
things were about to change.
17 3634_CH15
11:27 AM
Page 411
The Future of the Industry 411
The Second Era
The advent of the mass-market consoles, such as the Sega Master System and the
Nintendo Entertainment System (NES), provided a wider range of platforms and
potential sales. Powerful 16-bit machines, such as the Amiga and the Atari ST, boosted
gaming’s prominence in the public mind.
Apart from the endemic licenses, things at first went well for the new consoles. A new
wave of original and exciting software hit the market—before the more cynical companies discovered just what type of games these consoles were best for. Then came wave
after wave of mediocre platform games that just about killed those markets off.
The same sort of thing happened to the next generation of consoles, such as the Super
NES and the Sega Genesis (Megadrive in Europe). It’s important to remember that
unlike the more recent consoles, which are geared towards accelerated 3D, these older
consoles were less specific in their hardware focus.
However, the perennial platform game (as always) proved to be the easiest game to
develop for these systems, and the market was flooded again. Maybe in the future this
will be the omen for the death of a platform. In a similar way to the banshee of Celtic
legend, the death of a platform will be heralded by a glut of platform games (or whatever type of game is easiest to produce on that machine).
About the same time, the licensing gold mine began to dry up. The increase in the time
required to develop a game made it increasingly difficult to ensure that the game would
be released concurrently with the film. Stripped of the benefits of riding on the back of
the main publicity, these games had to survive on their own merits. Very few of these
made the grade, and most of the companies responsible have disappeared into the void.
The Third Era
The third (and current) era of the industry has seen the inexorable rise of the polygon.
3D technology has captured the mind and hearts of the general public, and the potential for true mass-market gaming appeal is almost upon us.
17 3634_CH15
11:27 AM
Page 412
412 Chapter 15
Witness the success of products such as Deer Hunter. The success of this product
caught everybody by surprise. All the pundits had quite happily slated it and condemned it to a grisly fate. When it achieved the sales it did, the gaming world had to
reevaluate its position.
It seemed that the mass-market gamer—the mythical beast that the gaming industry
had been chasing for so long—turned out not to be interested in the technical excesses
of glorified demo products and had instead been captured by a simple, averagely
executed hunting game. Finally, this was a product that appealed to a totally different
market from the dedicated gamer, and it was cheap—a great gift for mom to give dad.
It might seem that Deer Hunter belongs to a genre of niche interest, but the fact is that
it is not as recondite as Elder Gods and plasma cannons.
Increasingly some sectors of the industry have gotten the idea that controversy seems to
be a way of selling games. It is a method that has attracted the attention of regulatory
bodies across the globe. However, it’s likely that the Holy Grail of the mass market is
not actually achievable by controversy directly—most people do not want to brutally
murder simulated people or aliens—but is simply a way of getting the games to
penetrate into the general consciousness. Two object lessons in this area are the successes of Theme Park and the SimCity series. To date, Theme Park has sold over four million
copies, and SimCity 3000 alone has sold more than two million without even considering the sales of the original SimCity and SimCity 2000.
Celebrity Challenge
A recent trend in computer games is the use of celebrities to promote games. These
celebrities can be either real or virtual.
In the case of the virtual celebrity, we have a wide range to choose from: Mario, Sonic,
Lara from Tomb Raider, Link from the Legend of Zelda RPG series, Cloud from Final
Fantasy, and many more. Mario and Sonic were largely responsible for the huge
success of Nintendo and Sega, respectively.
Witness the crossover into the mainstream of Lara Croft. Would the Tomb Raider series
have been the cultural phenomenon that it has been without Lara Croft and her outstanding assets? If the character of Lara had been replaced with a more conventional
character like Cate Archer from No One Lives Forever, would we be seeing tie-in
movies, Times cover stories and advertisements for Lucozade?
17 3634_CH15
11:27 AM
Page 413
The Future of the Industry 413
The publisher of Tomb Raider, Eidos, has potential to make more money from Lara
merchandising and tie-in products than they do from the sales of the games. An analogy
is the record industry, where sales of music account for only around 15% of the revenue
earned by a star like Madonna.
But for each successful virtual celebrity are heaps of failures. Does anyone remember
Zool or Bubsy? As soon as other companies witnessed the success of Mario and Sonic,
they all wanted a piece of the action. A plethora of cute but arrogant (or cutely arrogant?) characters appeared, and—fortunately—most of these went the way of the
dodo. Nowadays, in mid-2003, it is getting very common for games to feature two
heroes of very distinct types. One is a serious, conventional hero type and usually is
characterized by big Manga-like eyes, large hands and feet, an outsize weapon, a mop
of outlandish hair, and a determined little scowl like a nine-year-old trying to look
really tough. The character’s sidekick is a small animal or robot, often given to wisecracks and banter with the hero. If the game is for older players, the Manga style will
be dropped, but then you get another formula applying, which can perhaps best be
summed up as a pair of heroes who look like the straight-to-video poor relations of
Neo and Trinity—black clothes, big guns, and stern looks.
There are two problems with this approach. First, you can’t create likeable characters
to a formula because people aren’t stupid. They can tell when a creative work is
genuine and when it’s been synthesized. (Compare Lilo and Stitch, a brilliant pair
of cartoon characters, with just about any team from a contemporary game.) Second,
the old animator’s dictum that a successful character must be instantly recognizable in
silhoutte only works if your characters’ silhouettes are distinguishable from those of a
dozen other games.
Obviously the success of any virtual celebrity depends on the success of the game featuring that character. Cute and recognizable characters such as Mario tend to capture
the hearts and minds of the young. However, the quality of the game is by no means
irrelevant. Mario and Sonic are strongly defined, likeable characters, but still would not
have achieved the success they did if they had not been featured in such excellent
17 3634_CH15
11:27 AM
Page 414
414 Chapter 15
“The idea is that if you consume every Mario artifact you can get your hands on, if you
can play the Super NES game in a Mario Brothers sweatshirt while scarfing down an
individually wrapped Mario Brothers snack, then through some mysterious process of
celebrity transubstantiation you can become Mario, or at least take on some of his
abilities. It’s kind of like Pinocchio in reverse—millions of real boys dreaming of someday turning into a digital marionette.”
—J. C. Hertz, Joystick Nation, 1997
What about real-life celebrities? Discounting the obvious appeal of sports games
endorsed by famous athletes, we are left to consider the various attempts by other
famous figures to break into the games market.
The English heavy metal band, Iron Maiden, released a CD that contained a simple
first-person shooter featuring their mascot, Eddie.
Douglas Adams, author of the Hitchhiker’s Guide to the Galaxy novels, branched into
computer games by co-founding The Digital Village. This developer’s first product was
the sprawling adventure game Starship Titanic, designed by Adams, which is set on an
eponymously named starship heading for disaster.
Tom Clancy, author of many successful political-suspense novels, has also had a hand
in the games industry with a string of branded products, most notably Rainbow Six and
Splinter Cell.
These are not the only authors who have decided to use their celebrity and creative
powers to make a splash in the games industry. Michael Crichton, famous as the creator of a host of successful properties including Prey and Jurassic Park, has partnered
with Sega to devise a new range of products that can potentially work as games,
movies, and novels.
Even old-time pop stars like David Bowie and Queen have tried their luck with The
Nomad Soul and Queen: The Eye, respectively.
Is this a valid approach? Is David Bowie the way to attract the mass market? Time will
tell. Entertainment software is a new and very different medium, and it will create its
own kind of stars—just as Valentino and Keaton were big stars in the silent era but
didn’t make the transition to talkies. Disney does recruit big names for the voices in
17 3634_CH15
11:27 AM
Page 415
The Future of the Industry 415
Toy Story and other films, but they don’t make a big selling point out of Tom Hanks
doing the voice of Woody, for example. A big-name star in an animated feature just
helps to raise the marketing profile; it doesn’t change anyone’s mind about seeing it. In
Hollywood parlance, the stars who can open a live-action picture don’t necessarily
open animated features as well.
The stars of the computer-game universe will very often be virtual stars, like Lara
Croft, rather than real-life names. Cutter Slade (from Outcast) is a character is his own
right, and the game would not have been enhanced by getting a big star like Bruce
Willis to voice him. If anything, it would have been a jarring note, reminding us that
“it’s just a game,” simply because Bruce Willis’s voice is too familiar. We know he isn’t a
game character. Whereas the actor who did voice Slade (David Gasman) did a great
job, and the very fact that he wasn’t a celebrity helped us believe in the reality of the
Violence in Games
The question of violence in games has recently become quite a big issue, because, with
the increasing power of technology, realistic violence can be graphically depicted for
the first time.
Mortal Kombat was a game made notorious for its combination moves that culminated
with the death of your opponent by various means. One of the most gruesome of these
involved ripping the skull and spine out of your opponent and holding it aloft. The
photorealistic nature of this imagery caused uproar among the moral majority.
In the U.K., the first game to receive a rating for viewing content was Frankenstein on
the ZX Spectrum computer. This was an entirely voluntary effort on the part of the
publishers in order to obtain publicity for the game. This (rather cynical) approach has
been used on a number of other releases over the years. Games such as Grand Theft
Auto played on their “18” certification to court notoriety and build sales. Max Clifford,
a U.K. public-relations expert, was called in to get publicity for Grand Theft Auto.
He succeeded in this by securing a series of scathing articles in leading broadsheet
newspapers, along the lines of “Ban this filth, now!” Predictably, sales went through
the roof.
17 3634_CH15
11:27 AM
Page 416
416 Chapter 15
GTA3 and Vice City set new heights for game violence. It is possible to hack off a
pedestrian’s head with a katana, bludgeon them to death with a baseball bat, even to
burn them alive with a flamethrower. It is also possible in these games to have sex with
a prostitute—although that, tellingly, is merely implied and not depicted.
To promote their mobster game, Gangsters, Eidos hired “Mad” Frankie Frazer (a member of the infamous Kray Gang, who was jailed for his crimes in the 1960s) to appear
at ECTS, the European Computer Trade Show. This move disgusted the press who saw
it as a cheap and distasteful publicity stunt. In response to this, Frazer’s involvement
in the publicity campaign was quickly downplayed. Whether the adverse publicity
generated by forcing Gangsters into the public eye has helped or hindered its sales
is a matter of debate.
Carmageddon appears to have been inspired by Paul Bartel’s film, Death Race 2000.
However, while the film was a satire directed against a society that derives entertainment from violence, the satirical scope of the game was not immediately evident.
Initially, the publishers had to replace human pedestrians with zombies oozing green
blood, because the censors found the red blood too realistic. They singled out the red
tire marks left by your car as being particularly offensive, and refused to allow the
release of the game unless this was changed.
The specialist press booed and jeered at this measure, baying for blood and guts in
their games. The publishers of Carmageddon appealed the censor board’s decision,
and it was eventually overturned, allowing them to release the full-blooded version.
In retrospect this may have been harmful, as it brought only negative attention to the
industry as a whole. If the industry cannot self-regulate, then it will have to submit to
external censors, which will neither please the publishers nor provide the consumer
with satisfactory products.
Interestingly enough, when Microsoft recently released a rush-hour racing game
Midtown Madness, there were complaints from some members of the specialist press
that you could not run over pedestrians. They always manage to leap out of the way at
the last moment, in the manner of ’70s car movies such as Cannonball Run. The game
was actually marked down in a number of reviews, and the inability of the player to
kill pedestrians was stated as a major disadvantage. However, the ability to kill pedestrians would not have made it a better game, and it has no bearing on the feel of the
driving within the game.
17 3634_CH15
11:27 AM
Page 417
The Future of the Industry 417
Postal, a game named after an infamous event when a sacked postal worker went into his
former workplace with a gun and shot everybody in sight, was a rather tacky attempt at
cashing in on the violence debate. Advertisements for this (remarkably average) game
involved displaying all of the condemning quotes from the media, and presenting these
as reason to buy it. However, there is no evidence that the media attention gave Postal
any mass-market appeal. After initial interest, sales soon tapered off.
The granddaddy of all “realistically” violent games, Doom, has been singled out for
many the school shootings in the U.S. This has left the moral majority howling for all
violent computer games to be banned, and several class-action lawsuits were launched
against number of publishers.
However regrettable and tragic the shootings were, this response is only a knee-jerk
reaction. It’s only natural when we are outraged and horrified by a great tragedy to cast
around for a place to put the blame.
Games do not cause violence. They may be able to reinforce violent feelings in disturbed individuals, but then so could a book, a newspaper article, a TV program, a
movie, or any of the hundreds of other stimuli that we are subjected to every day. The
fact is that if someone is disturbed, anything can set him or her off.
During July of 1999, in Japan, a young man with a knife hijacked an airplane and
killed the captain by stabbing him in the throat. Why? Because he wanted to fly under
Tokyo’s Rainbow Bridge as he had done in Microsoft’s Flight Simulator ’98, and the pilot
had refused to do so. After the terrible events of 9/11, there was even discussion in the
media that implied Microsoft’s Flight Simulator could somehow be blamed simply
because it is possible for a player to crash into a building.
Does this mean that that particular flight simulator should be banned? Should
Microsoft be sued for releasing such a dangerous product onto the market?
No. That would be ridiculous. A computer game cannot turn a reasonable human
being into a deranged killer without there being something seriously wrong with them
in the first place.
One of the cornerstones of a modern democratic society is that people are responsible
for their own actions. Saying “the Devil made me do it” is a defense that should have
gone out with the Salem witch trials. This is basically going to happen every time
17 3634_CH15
11:27 AM
Page 418
418 Chapter 15
something new comes along. In seventeenth-century England, the government of
Oliver Cromwell closed the theaters because they thought drama was a subversive art
form that would lead people astray. In the early days of cinema, gangster movies were
attacked because it was believed they glorified violence and would cause problems in
society. Howard Hawks’ film Scarface was held up for two years while extra scenes
were added to give it an anti-violent message and thus appease the moral majority.
It was finally released in 1932 as Scarface: Shame of a Nation and was still widely
condemned—yet today it is regarded as a classic. Television and then comic books
got the same treatment, being blamed for all society’s ills (a critic of the early 1950s
referred to television as “an appalling Pandora’s box”) and now it is the turn of
computer games.
Figure 15.1 shows the available markets. The hardcore market that will buy a game
on its merits alone form the peak. Then there is the larger base of floating gamers who
will buy a game if they become interested in it through other routes, such as related
merchandise. The largest area, the mass market, are those who never normally buy
games. These are the people who the games industry will have to adapt to get the
serious money—and violent games are not the route.
The Mass
Figure 15.1 The shape of the market.
In the real world there is no single cause of violence. A significant factor in casual violence, however, is the inadequacy of the perpetrator to express himself in other ways—
either by reason of emotional immaturity, lack of education, or social injustice. When
we are unable to see the body language and nuances of etiquette that normally moderate our behavior, it is all too easy for many of us to succumb to violent anger—hence
the phenomenon of road rage. Extreme violence in games occurs because of the inadequacy of the game to deliver other, more interesting forms of interactivity. Players
17 3634_CH15
11:27 AM
Page 419
The Future of the Industry 419
demand (rightly) that their actions must make a difference in the game world. In the
absence of sophisticated AI or writers capable of creating interesting characters and
scenarios, games fall back on guns, car crashes, and explosions. It is an easy answer,
but if games are to reach acceptance in the mainstream then they must find more
socially agreeable ways for players to act.
“There are two reasons why games have historically relied heavily on violence. First,
game players were typically adolescent (or at least the first generation was). Second,
the technology limited gaming environments, making more complex interactions
impractical…. It’s no accident that Sony’s next generation PlayStation [2] chip is called
the Emotion Engine.”
—Demis Hassabis quoted in Edge, September 1999
Violent action in games is a specialized taste with only cult appeal. It just happens that
most gamers at the moment belong to that cult minority. If you want people to invite
you into their homes (which is what they do when they buy your game, in effect), you
obviously need to make a commitment to entertaining them in a responsible manner.
The mass market will never take to games that go out of their way to offend and
The industry is going to have to take realistic portrayal of violence far more seriously
in the future. Computer gaming is becoming much more part of mainstream life and
consequently is becoming more family oriented. No longer is gaming the sole preserve
of the 15- to 20-something males. The fact that traditionally noncomputer-based companies have entered the market with some successful products (such as the Lego and
Barbie products) much to the disgust of the hardcore developer contingent, shows how
much the industry is changing.
The demand for family-oriented content can only increase, and this means that violent
content will have to be better regulated and more restricted in order to protect the
young and to reassure parents.
Releases such as Kingpin have shown that the degree to which games can portray
violence can be far more gritty and realistic than previously supposed. Whether this
actually adds to the gameplay is another matter. Personally, we’d rather have a game
sold on the merits of the gameplay than on the size of the body count.
17 3634_CH15
11:27 AM
Page 420
420 Chapter 15
Violence in games is not always gratuitous. For example, in Outcast the hero is fighting
a military junta that uses violent oppression to control its peaceful citizens. The hero is
a Navy SEAL with formidable weapon skills, and it is clear from the backstory that he
is no psychopath but rather a responsible soldier who uses violence only when there is
no other choice. At present, violence in computer games is unbalanced—the moral
implications of violence are rarely dealt with, making the violence merely gratuitous.
Contrast this with drama or literature, in which violence has consequences for the
perpetrator as well as the victim. Violence doesn’t come without moral implications,
and these should be part of any game that uses violence. (As a side note, presenting
the moral consequences of violence does not mean that you necessarily have to punish
any use of violence in your games. That’s playground morality. A mature art form can
present a difficult subject in many ways and explore many themes without having to
“At first when the graphics were basic, all you could do was give a character a gun and
see how many people he could shoot. As technology becomes more sophisticated,
games will have to be more imaginative. There are a lot more interesting scenarios than
gratuitous violence.”
—Demis Hassabis, creator of Theme Park and founder of Elixir Studios
Children’s Products
One area that is currently overlooked in the market is the children’s market. Only a
few companies are starting to turn onto the possibilities.
This is where we believe one of the biggest areas of growth will be. Not with the trite
“edutainment” products—which attempt to educate with such unworthy dreariness—
but with the two separate subgenres of children’s educational products and entertainment products.
“The major breakthroughs in user interface, such as icons and multiple application windows, originate in work done for children. So does the concept of the laptop. But alas,
the educators have grabbed the market of children’s software and perverted it into the
polar opposite of everything for which the personal computer stands. Rather than
empowering the children and liberating their inner creative selves, they intimidate and
17 3634_CH15
11:27 AM
Page 421
The Future of the Industry 421
enslave the child to the machine. Rather than teaching the very first lesson that every
child of the modern techno-world must learn, that Man controls the Machine, they subliminally inculcate the child with the notion that Man must obey the Machine or fail.”
—Tzvi Freeman, writing in Game Developer, September 1997
This is a huge market that is still relatively untapped, and we can confidently predict
that the children’s market is going to be one of the biggest growth areas over the next
few years. No matter how cool you think making adult-oriented games is, this is one
area you’ll ignore at your peril.
“Creatures Adventures is designed around the philosophy of free play, the aim being to
guide and teach your creatures as they experience the world around them. The product
provides the building blocks for you to create their own experiences without forcing
them to follow linear play patterns.”
—Ben Simpson, Designer, quoted in Edge magazine, September 1999
Whether the next generation of kids’ games will have the right approach or will continue to peddle the approach of “Help Roger Raccoon rescue his presents by jumping
on the number logs to complete simple sums” remains be seen. The great success of
“sandbox play” products like The Sims will hopefully trickle down to influence children’s games too, meaning that children should be free to explore and experiment at
their own pace, avoiding the regimented structure of what has passed (inadequately)
for children software in the past.
The New Model Developers
Hard times are ahead. A development company that tries to make money by its games
alone faces a tough future. The key to success is leverage. This can be leverage of the
technology itself (such as licensing the engine to other games companies or even to
nongames companies) or leverage of the Intellectual Property (IP)—Kingpin selling
Diesel clothes, Lara Croft appearing in comic books and commercials, and so forth.
17 3634_CH15
11:27 AM
Page 422
422 Chapter 15
At the moment, games based on existing licenses (like GoldenEye) outsell characterbased games originated in-house (like Grim Fandango) by a factor of three (source:
Develop magazine). However, you have to remember that those bought-in licenses cost
money, whereas ideally in-house IP will generate money. The choice is between lowrisk, short-term gain (the license) and high-risk, long-term gain (the original IP).
“I take a good deal of pride in seeing games like Valve’s Half-Life. They’ve built on our
foundation and they’ve done a spectacular job. I’m not sitting here kicking myself and
thinking, ‘We could have done that.’ Instead we’re thinking, ‘Hey, we get royalties off
—John Carmack, Edge, April 1999
Leverage can be internal also. For a small developer, reuse of technology is only marginally practical—you complete one game, by which time much of your technology
will be old hat. So consecutive reuse is of little value, but concurrent reuse (applying
the same technology, design concepts, or even artwork across several projects in
parallel) is more than a bonus, it is an economic necessity.
Obviously, a lone development company with a staff of 30 cannot easily begin to
achieve concurrent reuse internally. Therefore, over the next few years, we will begin
to see a consolidation within the industry. A few “super developers” will emerge (we
are seeing the first of these already), and there will be a move to a system closer to the
way movies are developed. This involves a centralization or pooling of resources.
An example of one kind of consolidated approach is the satellite scheme originated by
Peter Molyneux’s company, Lionhead Studios. The approach taken is to form a group
of favored developers and provide support and assistance to them within a loosely
affiliated community of development teams.
In many senses we can draw a parallel with United Artists, which was founded in 1919.
“[Film] stars Mary Pickford, Douglas Fairbanks and Charlie Chaplin, and director D W
Griffith, founded United Artists as a corporate apparatus for distributing their independent productions. United Artists never owned a production studio; rather, it distributed
features made by filmmakers on their own lots or rented facilities.”
—Extract from the “United Artists” entry in Cinemania (Microsoft, 1997)
17 3634_CH15
11:27 AM
Page 423
The Future of the Industry 423
To regard the departure of the original founders from United Artists as a failure of its
creative vision would miss the point. The advantages of such an alliance are many.
Firstly, from a commercial viewpoint, technological resources, business contacts, and
management can be shared so that all participants can benefit from economies of scale.
(How may small independent development companies could afford their own motioncapture studio, for example?) Secondly, the umbrella of an established central studio
nurtures newcomers into the fold. Even a top-notch team of developers would face
many problems at startup (securing investment, convincing a landlord to rent office
space, arranging a loan facility at the bank) that being part of a group will alleviate.
Thirdly, the central studio can maintain a talent-pool inventory, so that individuals
within the scheme are never left idle. Fourthly, a superdeveloper is better able to raise
initial finance for creative development and prototyping, enabling it to secure higher
royalties. (See Case Study 15.1.) Lastly, the studio guarantees quality control, which is
desirable to both the publisher and the consumer.
Case Study 15.1 It’s Hard for Developers
In 1998 we were involved in putting together the staff roles and design and architecture doctrine for a nascent development company that was assembling a business
plan. They were seeking most of their startup finance from a publisher, and it was
interesting to see the sales and revenue projections based on the publishing deal they
were negotiating.
Their royalty rate began at 20%, climbing to 25% after 350,000 units. This placed
them in profit at around 375,000 units. Contrast that with the publisher, who was
set to recoup their investment and go into profit at around 200,000 units.
There is no reason why the publisher should not arrange things this way. The risk
was to be theirs, after all—it was their two million dollars! In effect, they were simply building in self-protection. If the startup capital had been obtained by bank
investment, the risk would be represented by a rate of return on the investment as a
loan. In the case of venture capital, it would have been reflected by equity.
The founders of the company had some creative business ideas of their own. Armed
with evidence of firm interest from a publisher, they were able to arrange private
investment (from business angels) in return for a moderate equity share. This would
17 3634_CH15
11:27 AM
Page 424
424 Chapter 15
allow them eventually to take a finished product to auction between all the top
publishers, thereby securing a royalty closer to 45%. The business angels, incidentally, acquired equity in a specific company formed just for the production of that
one game, in direct analogy to the Hollywood model, where a new company is
formed to create each film. This meant that the development team could also grow
to undertake other projects without having to have “sold their souls” right at the
In order to take full advantage of the link-ups within a satellite scheme, the independent teams will have to make use of the full possibilities of shared resources, as suggested by the software factory methodology. Getting a number of independent teams to
explicitly cooperate in an expertise-sharing scheme will not be achieved by wishing it
so. Formal process is the only way.
This scheme is similar to the production model used in animation for cinema and
television, in that there is a central creative core that kicks off designs and architecture
but hires individuals or teams of contractors for the code spadework. However, you
wouldn’t want to get the modules of your product back from outsourcing and discover
that they are way off what you intended. This means:
The creative core must oversee the project. Somebody from the creative core
therefore retains ownership throughout. Traditionally, we might expect this to
be the producer of the product, but more probably there will have to be a shift
towards design-led control, so we might predict that the project lead will be
closer to the director’s role on a movie. We’d expect this to work best (people
being what they are) if you’re either working with an independent team that
you’ve used in the past and can trust, or if you bring the contractors to your own
location where you have the resources.
There must be a standardization of design methods, libraries, and such. If you’re
making a movie, you can show your second unit director the storyboard and discuss the lenses and lighting, and then you’re fairly sure he will shoot it the way
you want. We need the equivalent of universal standards in development. Yet
(remarkably), some developers are deliberately choosing to define their own
unique standards—the equivalent of insisting on speaking only Navajo and then
wondering why the Fox network can’t find a job for you!
17 3634_CH15
11:27 AM
Page 425
The Future of the Industry 425
The Lionhead model seems destined to create the kind of “superdeveloper” we’re discussing. Lionhead’s satellites will adopt a standard way of doing things—or, at any rate,
the potential is there for them to do so. They share resources, and they have mutual
trust as to standards of performance. Moreover, the top game designers like Peter
Molyneux could use the scheme to kick off a couple of projects a year as opposed to
the current average of maybe one every two years. On the assumption that star quality
(the ability to deliver a AAA product) is what publishers will pay for, it makes sense to
streamline development so that the guiding genius is not tied up in finalizing every little detail. (This does, of course, require development companies to gain a better understanding of what full design should entail. In particular, they need to recognize that
testbeds belong to the design phase, not the development phase.)
The trend towards the superdeveloper has come about principally because publishers
(or rather their shareholders) have been asking, “Why is so much of our money tied
up in development? We make money from publishing. We’re not in the banking business!” A few years ago, most publishers were directly investing in internal development. That mean large development staff payrolls and at any one time a possible sunk
cost, for a medium-scale publisher, in the millions of dollars, Small wonder that the
trend has been toward partial funding of independent development studios, or (ideally,
from the publishers’ view point) just buying a completed game and putting it straight
onto the market at no risk.
So, we will see a trend towards the film industry model of finance, with the publisher
or main studio providing confidence in product for which they are the eventual purchaser, but which may be bankrolled through production by completion bonds. The
president of Sony was quoted recently (1999) as saying there would be only five developers in the world. The staff of these superdevelopers will constitute the creative core
of talent, like the software factory design and architecture model on a considerably
larger scale. Production staff (as in the film and animation industries) might not be
part of the company—may not even be in the same country. Not all of the creative core
need be permanent staff, come to that, just as authors can drift from one book publisher to another. If you happen to be a venture capitalist, the lesson is clear: Sink $25 million into a superdeveloper by all means, but don’t waste $2 million on a small development team because you will never get the return that high risk expects.
17 3634_CH15
11:27 AM
Page 426
426 Chapter 15
Some projects will be originated in-house by the superdeveloper “studio.” The advantage of this, as we saw in Case Study 15.1, is that you don’t get investors (angels and
others) to directly buy a share of your studio. They invest specifically in Shudder, Inc.,
a specially formed subsidiary vehicle that is developing the game Shudder, on the basis
that this doesn’t dilute your core equity. (If Shudder does very well, they’ll reap the
rewards of sequels and merchandising, which is the ideal from the investor’s viewpoint.)
Other games will be brought to the studio by small nascent teams. For instance, a
designer, planner, and lead programmer might present a concept-and-technology
demo to the studio bosses just like a director, scriptwriter, and star pitch a package in
Hollywood. Armed with letter of intent from the studio (“We are very interested in
producing this game…”), they may then get finance from a bank to assemble a full
team (equivalent to a film crew) a perfect example of synergy, or what in economics
is called the principle of cooperative benefit. The investors have money but no expert
knowledge to tell them if it’s a good game. The studio has expert knowledge but
doesn’t want to invest more money than is strictly needed to secure a first refusal.
The developers just want the money to make their game.
But how will even superdevelopers make money? Even at a 50% royalty, with the gross
wholesale price dropping, you will need to sell some 300,000 units before the deal
starts looking attractive to serious investors. Well, firstly the income from sales of
games is not the only asset the superdeveloper is building; they will own a lot of IP.
Additionally, development is cheap for them (the economies of scale again), and they
can afford R & D across multiple projects, which, for a small developer, would be
impossible. Lastly, as have seen, the creative superstars will be able to kick off more
projects by this method. Suppose that Will Wright could personally oversee one game
every two to three years, or he could mastermind one every six months or so by delegating part of the work to his trusted inner circle. You may think the quality of the
games would suffer. We don’t happen to think so, if rational development procedures
are used; raw creativity is very rare in all industries, but the craftsmanship required to
finesse a great idea to completion is relatively common.
And what about the contractors? (These contractors hired by the superdeveloper will
be not only individual coders but, more often, small development companies of some
12 to 20 people.) How will they make money? The short answer is they won’t (not the
big money to be made on back-end participation, anyway). They’ll just get salaries, the
17 3634_CH15
11:27 AM
Page 427
The Future of the Industry 427
upside being that those salaries will be on a par with the mainstream IT industry. We
can draw the analogy of session musicians in the music trade or, again, of film crews.
The director and producer and stars are the ones who get rich on royalties, but, if
you’re the camera operator, you must make do with a wage.
And publishers? The advantage to them is that they are brought a product bearing the
name of a known creative force. It could be an individual, like Sid Meier, or a wellknown brand name guaranteeing quality, like Blizzard. They know it will conform to
quality standards and be delivered on time, and the extra trust and reduction of risk
inherent in the new model are worth the doubling of the royalty.
The Online Revolution
The kind of content we see in games is the result of games having evolved as a retailbased, hobby market. Suppose that television wasn’t broadcast but instead worked solely off retail. You’d be buying an entire 26-episode television series on DVD for $40.
That would have a dramatic effect on content. Shows like Star Trek would still exist in
such a world and maybe Buffy. But you can bet there’d be no Six Feet Under or Scrubs
or Curb Your Enthusiasm—much less shows like Big Brother. Television would be a narrow, deep market in the same way that games have been.
Except now games are going to be broadcast—or, at any rate, available on demand as
part of a massive range of broadband content. And more often you’ll be accessing that
content on your TV screen via your Xbox2 or Playstation3 than via the PC in your den.
How much is that going to change the nature of games? A lot!
Delivering Games Online
Let’s say that as a rule of thumb, the cost of building the first level of a game is about
half of the total production budget. Clearly, as the publisher, you would prefer to get
revenue flowing in off the game as soon as that first level is complete. For one thing,
money from broadband subscriptions or pay-to-play will finance the remaining levels
so you don’t need to stump up the cash yourself.
17 3634_CH15
11:27 AM
Page 428
428 Chapter 15
Even better, releasing the game episodically maintains its marketing profile over a
longer period. You can roll material out through the year, keeping your sales presence
high. A new level every couple of weeks adds up to a special add-on release of the
retail product every three months, say, and a complete new game in a box every year.
If you already have Penultimate Phantasy 1 in a box on the shelves, it helps if you can
be releasing Penultimate Phantasy 2 throughout the year online and then packaging all
those levels into a new retail release, and so on.
And there is yet another advantage, again analagous to television. If the game isn’t
doing well, you can see that straightaway and you can pull it. The pilot might cost you
50% of the full production budget, but if the game is not selling then at least you only
wasted that 50%.
Playing Games Online
We were drawing a parallel with television. Television also uses the model of episodic
“online” delivery followed (in the case of many shows) by a boxed product at retail.
However, in TV, broadcast isn’t just a way of eking out that retail revenue. It has a
defining effect on the content. It’s going to be the same with games. Games are what
they are because of the retail revenue model. As games go online, the nature of the
beast will change.
First off, most people would rather pay $5 for two hours of entertainment than $35
for 40 hours. That’s pretty much the defining difference between mass market and
hobby market. The online publisher is effectively a broadband channel and doesn’t
have to worry about manufacturing overheads and the retailers’ shares, so selling that
40-hour game as 20 episodes at $5 each makes much better economic sense. (In practice, it wouldn’t be 20 episodes; it would be more like 10. That’s because the majority
of game purchasers don’t persevere with a game beyond the mid-point. But by delivering the game online, the publisher still gets considerably more than at retail and doesn’t need to fund those levels that most people are never going to get to anyway.)
Most importantly, this is the point at which games become part of television. Television
is a place in the home, not a technology. With online delivery of episodic games, we
will be looking at a new market of casual gaming and the way for developers and publishers to cater to this larger and (as we saw above) potentially much more lucrative
17 3634_CH15
11:27 AM
Page 429
The Future of the Industry 429
market is with entertainment software that can be enjoyed by everybody in the living
room. It needs to be family entertainment that Mom won’t banish to the den. It needs
to be something that works if you want to sit forward and interact or if you’d rather sit
back and just watch. Ico is an excellent example of one such product, although Ico for
all its brilliance and beauty might still be too much of a game for the casual tastes of
tomorrow’s video entertainment audience.
What about the online multiplayer audience? People are predicting games for thousands
of concurrent players in persistent worlds. Isn’t that the future of online gaming too?
The main disadvantage of massively multiplayer online games is that they are difficult
to dip into and out of. The mass market is not interested in spending every spare hour
online. They want to be able to choose when to play and not be penalized if they don’t
have time to put the hours in.
“The hardcore Ultima Online player is putting in six to eight hours a day. You might have
eight hours sleep, then some hours at work or school, and the rest is spent online.”
—Richard Garriott (“Lord British”), interviewed in PC Gamer, September 1999
In fact, we would liken massively multiplayer gaming to ham radio. It’s the preserve of
highly committed enthusiasts. This is not to say that a large number of people won’t
want online multiplayer environments, merely that those online products will be places
rather than games. Part of what people do in those places will be to play games. But
they will also socialize, enjoy art, and go shopping. And the games they do play for the
most part will be casual, like activities at a beach resort, not hardcore gameplay.
Gaming in the strictest sense of pitting yourself against rules and challenges is and
always will be for the minority.
17 3634_CH15
11:27 AM
Page 430
18 3634_Part 3
11:28 AM
Page 431
Part III
Game Architecture
Chapter 16
Current Development Methods
Chapter 17
Initial Design
Chapter 18
Use of Technology
Chapter 19
Building Blocks
Chapter 20
Initial Architecture Design
Chapter 21
Chapter 22
The Run-Up to Release
Chapter 23
Chapter 24
The Future of Game Development
18 3634_Part 3
11:28 AM
Page 432
19 3634_CH16
11:29 AM
Page 433
Chapter 16
Current Development
• The history of game
• Language wars
• Platform Wars
• Optimization and the
his chapter discusses the progression in development
languages and methods that have been used in producing
home computer games since the early 1980s.
The evolutionary path of game development has taken a
slightly different route than the mainstream development
While mainstream development generally started on large
mainframes and worked its way down to the PC, game
development has always started on small computers, with
the developers working around the limitations of the system.
These two different approaches have now, for the most part,
converged, and little Johnny can play the latest games on
the same system that daddy uses to file his tax returns. This
platform convergence (ignoring both consoles and antitrust
suits) has to be a good thing. It now means that, ostensibly,
both business and game software are written in similar
styles. Both are now either coded in C or C++ and generally
use the application programming interface of the native
operating system and a sprinkling of third-party libraries.
This is good news for companies seeking employees—and
for prospective employees seeking work—because it means
that the spread of skills required for programming is diminishing. If the skills required for programming a game are
• Monolithic versus
19 3634_CH16
11:29 AM
Page 434
434 Chapter 16
essentially the same as those for programming business software, this means that there
are more skilled workers available. Some people may say that writing a game is more
technically exacting than writing business software, which is considered easier and less
innovative. If you put aside any prejudices for a second and think about it objectively,
you will probably realize that this cannot be true. Usually software written for business
is just as technically exacting (if not more so) than game software of a similar size.
The main difference between the two is that game programming usually requires a
correspondingly huge amount of supporting material (in the form of audio and
graphical work).
Programming, however, is not the only skill needed to develop game software. Anyone
expecting a career in the games industry (or anywhere in the software industry) without secondary areas of knowledge such as mathematics or more game-specific areas
such as AI and 3D calculations will find it difficult to progress to anything above the
level of code monkey.
This convergence, or homogenization, of skills has in some ways simplified game
development. More people with the right skills are readily available for work, and,
although the market is tight, there is a wider pool of people to choose from. At least
there would be if recruiters could modernize their attitudes. Looking through the job
ads in the specialist press, recruiters seem to want only programmers with exactly the
right skill set (3D math skills, and so on) or fresh young graduates; they just want
either an off-the-shelf package or the cheapest available.
We can understand this viewpoint, but we are curious to know for how long it can be
maintained. We don’t know of many game software companies that provide a structured training program for new employees—with the possible exception of larger companies, but even then, they expect the new hire to already have solid skills. The reason
for this may well be that they are afraid of training programmers only to lose them
when, after a couple of successful releases, they leave create their own “super-team”
studio, as has become common over the past few years. The money offered to teams
with a few successful hits behind them is certainly enticing enough.
19 3634_CH16
11:29 AM
Page 435
Current Development Methods 435
Although we honestly don’t know what the solution is, the software factory has builtin measures to thwart this phenomenon by organizing the teams in a different way.
This ensures that no one particular team is responsible for an entire product. With
the software factory, no one can say that “their” team developed the game, because
they would have developed only a part of the game. The entire game itself would be a
product of the effort of the entire company. The simpler option is to make sure that
the team that developed the game didn’t even feel the need to leave. Maybe if the distribution of money for effort expended was more this sort of situation wouldn’t occur
so often. The distribution of royalties has long been thought to be unfair, and we know
of companies who use small print and legal clauses to reduce the amount of royalties
they pay to the absolute minimum, which in most cases can be zero.
On one occasion we chanced to go to the European Computer Trade Show and sat
for a while in the “Suits Only” room listening to the various conversations. The topics
varied wildly, but one theme was common: The employees of the software companies
were all discussed in terms of how they could be exploited. They were viewed as cattle
to be milked. This was something we had cynically suspected for some time, but to
actually hear it openly discussed was a bit of an eye-opener. It gave us some new
insight on the industry. For your average Joe Programmer, the industry is not a fair
place, and the distribution of wealth is weighted heavily against them. Revolutions
have been started for less. The industry has generated its own myths and expectations,
and this appears to be one of them.
This is not to say that we are advocating all programmers to overturn their desks,
smash their monitors, and storm the management offices. We are merely making an
observation that the games industry cannot continue in this fashion if it is to become
a mature and stable industry. Programming teams are not rock groups, and as soon as
people can look through all the artificial glitz and glamour to examine the true situation, the industry will take a big step towards maturity. Over the past few years, with
the world economy heading towards a recession, natural selection has taken care of
this to some extent. However, there is still a long way to go.
19 3634_CH16
11:29 AM
Page 436
436 Chapter 16
The History of Development Techniques
Back in the early days of game development, the machines were very limited, both in
variety and in capabilities. The three home computers that most people think of when
asked about this period are the Commodore 64, the Sinclair ZX Spectrum (very similar
to the American Timex Sinclair 2048), and the Amstrad 464. These were the three
main contenders in Europe during the 1980s.
Of course, these computers were antiquated by today’s standards. They had slow, 8-bit
processors running at a few megahertz, limited graphics, and between 48 kilobytes and
64 kilobytes of memory. 3D graphics were virtually unheard of, and those that did exist
were wireframe or primitive solid. Texture mapping didn’t exist as the graphics capabilities and the processor speeds were just not up to the task. Programming for these platforms involved the constant need to work around their limitations in order to produce
the best results. All games generally had to be coded in Assembler in order to achieve
speed and space requirements. There were no good C compilers then, and the few that
did spring up did not produce code small and tight enough to be usable for games.
This meant that porting a game to another platform required a complete rewrite. There
was no easy conversion method, no compiler switch to change the target from, say, a
ZX Spectrum to a Commodore 64 so that you could recompile and have your game
targeting multiple platforms.
There were some advantages, such as the fact that there was no variance within the
bounds of a platform. You knew that, except in bizarre (and commercially discountable) cases, the machine that you were developing on was exactly the same as the
machine that little Johnny had at home. It was a situation similar to that experienced
with console programming today: The target machine has exactly the same capabilities
as the development machine.
Developing on exactly the same machine as the target platform means that you didn’t
repeat a common disaster that has plagued some software developers—producing a
game that runs too slowly on the player’s machine—because the machines used to
develop them were considerably more powerful. Back then, the choice of development
language was a no-brainer: You chose Assembler. Anything else just wasn’t fast enough.
19 3634_CH16
11:29 AM
Page 437
Current Development Methods 437
Because of the restrictions in hardware, the coding priorities then were quite different
to those of today. While speed and optimization are still issues today, they are not
emphasized as much as they used to be. You had no second chances then; there were
no new faster processors coming out to aid your efforts. You were stuck with the limitations of the machine that you were developing for.
In the eyes of the nostalgic, these limitations were not necessarily a bad thing. Because
of the restricted memory, graphical capabilities, and sound hardware, more emphasis
was placed on the gameplay, and the games simply played better.
This is not what we believe has happened. Apart from the obvious nostalgia trips,
gameplay does not seem to have developed on the same scale as have graphics and
sound for other reasons, which are covered in the next section.
The Rise and Fall of the Original Game Idea?
It is true that games are less original than they used to be. Witness some of the titles
released back in the early 1980s: Head over Heels, Manic Miner, Cliff Hanger, Trashman,
and They Stole a Million (shown in Figures 16.1 to 16.5).
Figure 16.1 Head over Heels.
19 3634_CH16
11:29 AM
Page 438
438 Chapter 16
Figure 16.2 Manic Miner.
Figure 16.3 Cliff Hanger.
19 3634_CH16
11:29 AM
Page 439
Current Development Methods 439
Figure 16.4 Trashman.
Figure 16.5 They Stole a Million.
Many games had odd subject matters and peculiar objectives. For example, in Cliff Hanger,
you played a cartoon hero who had to kill a gun-toting bandit who was coming over the
horizon to shoot up the town, using a Wile E. Coyote-style set of tricks and traps.
19 3634_CH16
11:29 AM
Page 440
440 Chapter 16
In Trashman, the aim of the game was to collect the trashcan from each house on the
street and empty it in the back of your truck before placing it back where you found it.
If you did your job well, the house owners would offer you tips. If you performed
badly, dogs would chase you. Admittedly, this doesn’t sound that exciting. However,
taken in context, this was one of the best and most innovative games of the day. Try it
out on an emulator if you get the chance.
Similar in style to one of today’s role-playing games, They Stole A Million put you in
charge of planning a robbery: You would select your team, buy plans of a likely target
building, pore over the blueprints marking out elaborate plans, and then send your
team off to perform the robbery, stepping in if necessary in order to smooth minor
flaws. This game is ripe for a remake.
Would these concepts be successful today? They may be because of the burgeoning
retro-remake scene. However, it is more likely that they would not be deemed commercial by the marketing department. Because of the pressures of the “mass market” (that
may or may not exist), today’s games have become very formulaic in their approach.
The other aspect these games is that they were limited by the hardware they were ran
on, which would have impacted the gameplay in a lot of cases and forced a deliberate
simplicity. However, it this simplicity that made the gameplay successful. The theory
was that if the game looked bad and sounded worse, it had better play well!
The fact is that, while graphical capabilities and processing power have increased
drastically since the early days, gameplay has not. If anything, gameplay has slightly
regressed. We are still playing the same old games, except we are being fooled by the
shiny new packaging (a clear case of the emperor’s new clothes).
Worse still, we are not even experiencing the best of the old gameplay; we are replaying the average run-of-the-mill games. The games with the best gameplay, such as
those mentioned above, are deemed not commercial enough and too “way out.” The
early days of the industry were very experimental, and there would be games on any
subject or combination of subjects you could imagine. They were very original and, in
some cases, very strange. There was, of course, a price to be paid for this flexibility:
the loss of the hypothetical mass market. One of the reasons given for the increased
19 3634_CH16
11:29 AM
Page 441
Current Development Methods 441
stagnation and lack of originality of most games today is that the more “way-out” stuff
lacks mass-market appeal. Unfortunately, this argument is true, because it is market
pressures that have forced the games industry into this state. But how much of it is
the tail wagging the dog?
The Development Environment
Back in the 1980s, the development environment consisted of one thing: a glowing
screen filled with incomprehensible (to anyone else) hexadecimal digits, and bizarre
codes such as LD A, 2Ah.
Nothing much has changed today either, except the incomprehensible hexadecimal
digits have been hidden behind several layers of user interface. The code is still there,
but it is well hidden behind a bewildering selection of menu options. Just in case you
might hurt yourself. See Figures 16.6 and 16.7 to see what we mean.
Figure 16.6 A game developer’s screen, circa 1983.
19 3634_CH16
11:29 AM
Page 442
442 Chapter 16
Figure 16.7 A game developer’s screen, circa 2003.
Developers have become spoiled in terms of the variety of tools available to them. The
life of a game developer has been made easier by a conspiracy of excellent optimizing
compilers, game libraries that automatically take advantage of hardware acceleration,
remote debugging, multiple-monitor debugging, and, in some cases, the ability to
develop the game using an emulator on more-powerful hardware, rather than the
target hardware.
In the 1980s, code had to be developed on the target machine. When the first game
programmers started out, they were assembling their own programs in their heads.
By this we mean that they were writing down the op-codes (machine language
commands) on paper performing the conversion to hexadecimal, and then typing
them into the computer using a simple program (usually self-written) called a hex
loader, which, as the name suggests, allowed the developer to load hexadecimal codes
into memory.
As time progressed, programs called assemblers became available. These programs
allowed the developer to skip some of the paper stage, and enter the op-codes into
the computer program, which would then perform the conversion phase and assemble
the program memory, resolving any references to other parts of the code as it did so.
19 3634_CH16
11:29 AM
Page 443
Current Development Methods 443
Although this made some things easier, much was still difficult. A fair amount of juggling was required, because the development was always done on the target machine.
The assembler itself took up some of the memory, which meant that for a game that
used all available memory, it had to be assembled in parts and pasted together.
Debugging was also a restricted and difficult activity, because the debugger itself
required some of the system memory, so consequently it was difficult to fit it all in
with the game in memory as well. This difficulty was slightly offset by the fact that
the developer usually knew all the code inside out, because the only way to get the
required speed was to completely dump the operating system (such that it was) and
write to hardware directly. Because all the machines were standard, there was virtually
no danger of intra-platform incompatibility, as there were with (for example) the latter
days of DOS PC game programming, where the developer had to take many possible
hardware configurations into account.
The main driving force shaping the development of computer games used to be speed
and size. This resulted in every game being written in Assembler and addressing the
hardware directly. Writing in Assembler allowed the developer to have direct one-toone control over the size of his or her code. In theory, you could take an Assembler
listing and calculate the exact size of the output that would be produced when it
was assembled. This is in contrast with a C or C++ compiler, in which the size of the
output object file is entirely at the whim of the compiler writer.
This imperative of writing directly to the hardware, and using bit-twiddling optimization gave rise to the expression “writing directly to the metal”—writing code that
accesses the hardware of the machine in order to obtain the maximum speed.
The concept of writing to the metal is considered part of the whole game programmer
mythology and has become less prevalent. The only way for a programmer to realistically write to the metal today is to become a hardware driver programmer, writing
the drivers for audio and sound cards and other computer peripherals. The myth still
persists, however, that unless you are writing to the metal you are writing slow and
bloated code. This has abated greatly since the first edition of this book. In fact, in all
but a few straggling areas, such as AI, third-party libraries are uniformly accepted.
19 3634_CH16
11:29 AM
Page 444
444 Chapter 16
The Development Languages
Assembler is not really used very much nowadays; most products are written in a
higher-level language, with only small critical parts being written in Assembler. Doom,
arguably the most famous computer game of all, was written almost entirely in C.
The only parts that were written in Assembler were a couple of vertical and horizontal
line-drawing routines.
This shocked many people at the time, and the Internet newsgroups were full of discussions on this new development. People had a hard time believing that it was possible
to write a game like Doom unless it was completely written in Assembler. You have to
remember that, although Doom looks pretty clunky by the standards of today, at the
time it blew everything else out of the water. It was a turning point in the development
of games for the PC; some would even argue that it kick-started the industry. It certainly
made people realize what the ugly business machine was actually becoming capable of.
Doom heralded the beginning of the C era of game programming. Watcom, the makers
of the C/C++ compiler used to write Doom, must have experienced a huge boost in sales
as everybody shifted from TASM (Borland Turbo Assembler) and MASM (Microsoft
Assembler) to Watcom C/C++ 10.5. The main strengths of the Watcom compiler were
that it could compile to a 32-bit DOS application and that it produced fast and welloptimized code, whereas the only real competitor, Microsoft Visual C++ 1.5x, could
produce only 16-bit DOS (or Windows) applications that were not as well optimized.
The fact that the Watcom compiler produced 32-bit DOS applications was a great boon
to game developers. The difference between a 16-bit and a 32-bit application is the memory model. In a 16-bit application, the amount of memory available to the application
is limited to 640K (actually 1 megabyte, but the system used the rest), unless you go
through a slow and clunky interface to get to the expanded memory (EMS) of the
machine. This memory was addressed by using a 20-bit segment:offset system that
limited you to addressing only 64-kilobytes of memory at a time using the 16-bit offset
pointer, unless you modified the segment pointer (which was slow). On the whole, it
was a nasty system and wasn’t really considered fast enough for the needs of game
programmers (let alone the fact that it was notoriously difficult to program for).
The move to the 32-bit model solved this at a stroke, as 32-bit pointers allowed the
application to directly address four gigabytes of memory, and the segment:offset model
became thing of the past. If the developer wished to allocate 2 megabytes of memory
19 3634_CH16
11:29 AM
Page 445
Current Development Methods 445
using the C function malloc(), then he could do so without problem. Moreover, the
returned pointer was a 32-bit pointer that could address the whole of the 2-megabyte
range simply by incrementing it (no messing around with segment pointers to access
a different 64-kilobyte block).
With the widespread use of Watcom C/C++, Dos4gw—the DOS extender used to provide 32-bit DPMI (DOS Protected Mode Interface) functionality to DOS executables—
became familiar sight when loading a game. The use of the DOS extender to allow
fully 32-bit executables was a watershed for game developers. Now they got the best
of both worlds: not only could they write directly to the metal, but they could also use
a simple and flexible memory model to do so.
The dominance of Watcom C/C++ in the games industry continued until a few years
later, Microsoft turned its attention to the games market and its reliance on DOS.
Microsoft had been trying to quietly kill off DOS for several years by this stage, had
succeeded in promoting Windows (3.1 and NT 3.51) as the platform of the future. All
new applications being released for the PC at this time were Windows executables,
except for games mainly because Windows abstracted the hardware to such an extent
that it was impossible to get the speed needed for a computer game. On top of this, the
user interface of Windows was, to say the least, visually bland and designed for homogeneity of application user interfaces. You were strictly limited in your access to the
hardware: for example, if you wanted to modify the contents of video memory, you
would use the Windows API to request the address of the area in video memory that
you wanted to change. Windows would then copy the area from video memory to
system memory, and when you had finished performing your modifications and release
surface lock, the memory would be copied back to the video memory. This is obviously
slow process, mainly because the Windows API calls used to do this were written with
flexibility in mind: they were designed to copy all video memory configurations, and
were not particularly optimized. They were fast enough for the average application or
word processor, but for a game—where the graphics usually change extensively every
frame—the speed just was not there.
Microsoft did not want to continue supporting DOS. It saw the future as a 32-bit
graphical operating system that was fast enough to support games. In what could charitably be considered to be practice run, they released WinG, a form of the Game SDK
(and a precursor to DirectX), designed to allow faster access to hardware resources on
19 3634_CH16
11:29 AM
Page 446
446 Chapter 16
Windows-based machines, we can remember only one game, SimTower, that was commercially released using this (although we’re sure that there were probably more), and
the game suffered from appallingly slow graphic updates (at least on the 486 DX2
66MHz that we saw it on, which at the time was a top-of-the-line system).
WinG died a quiet death, and, meanwhile, Windows 95 was released with a fair amount of
pomp in the closing months of 1995, as we’re sure some of you will remember, Windows
95 sounded the official Microsoft death knell for 16-bit applications; DOS was dead.
Except, that is, if you were writing a game. The games industry practically ignored this
new operating system, except to check it for compatibility with their Dos4gw executables. It was not considered then as a serious target for game development. Microsoft,
on the other hand, did not want to continue supporting DOS, seeing the games industry as the last bastion of the DOS programmer. If Microsoft could persuade the game
developers to begin producing games for Windows 95, it will have achieved its aim of
phasing out DOS, and shown the world that Windows 95 was the operating system
of the future. To this end, Microsoft developed DirectX, a library to allow the game
developer direct access to the hardware if required, and to take advantage of any
acceleration provided in the hardware. The ideology behind DirectX was to provide a
standard interface to the hardware that was easier to access than had been traditionally
possible using DOS. Initially DirectX was touted by Microsoft as being potentially
faster than DOS for graphics, because it would automatically take advantage of any
hardware acceleration that was available.
DirectX spawned mild interest (and derision), but it was not until the release of
DirectX 2 that people took notice. (Maybe it was the lurid yellow CD with a black
radiation warning symbol that caught their eye.) With DirectX 2, the API had just
about become usable, and was taking advantage of the newly developed Microsoft
Component Object Model technology (COM). COM is an object-oriented technique
that allows objects to be easily versioned and (at least in theory) accessible from
languages other than that in which the object was originally written.
During this period, Microsoft had not been lazy in updating its C/C++ compilers. It
was painfully aware that Visual C++ 2.x (which compiled for 32-bit applications) was
showing its age, and consequently an update was released in the form of Visual C++ 4.
This was a bit of a shock to the industry: a whole “generation” of game programmer
19 3634_CH16
11:29 AM
Page 447
Current Development Methods 447
had grown up with the concept of “Microsoft = slow, Watcom = fast,” and they were
surprised to discover that the Microsoft compiler produced faster, tighter code than the
Watcom compiler.
The result of this was the gradual transferal of game development from targeting DOS
to targeting Windows 95. The first game released as a Windows 95 native executable,
taking advantage of DirectX functionality, was a version of the classic Activision game,
Pitfall. The initial release of DirectX was not fully compatible with the Watcom
compiler, forcing anyone who wanted to do serious game development to switch to
the Microsoft compiler. We’ll leave it for the reader to decide whether this was by
accident or design.
With the increasing momentum of Windows 95 game releases, Apple, manufacturers of
the Macintosh line of computers couldn’t sit back and let its (already tenuous) position
be eroded further. More and more games were being released for Windows 95, which
further strengthened Windows’ position as the operating system for the masses. To
fight this onslaught, Apple released the Game Sprockets, a set of components analogous to (but not compatible with) DirectX for the Macintosh operating system. How
successful this was still remains to be seen, but, to date, only a handful of companies
(Blizzard and Bungie being two examples that spring to mind) are releasing games for
both platforms. We’ll plead the fifth on whether this is due to the difficulty of writing
cross-platform code, or whether it is due to lack of sales on the Macintosh. Recently,
however, Apple has reversed its fortune with the relative success of the “designer”
computer, the iMac, but whether this will impact the games market in any significant
fashion remains to be seen. Since the first imprint of this book, the Game Sprockets
have come and gone with little attention, and Apple is now focusing on their OS X
platform with integrated OpenGL and multimedia support. Although this hasn’t
pushed them to the forefront of gaming, a steady stream of titles are making their
way to the Mac, including such notables as Warcraft III and Neverwinter Nights.
This just about brings us up to the present day, at least for home computer development DirectX has matured through nine versions, and there is no serious PC game
development done without it, except in the case of certain OpenGL based exceptions—
which are becoming more plentiful nowadays. The Microsoft compiler has been
substantially enhanced and now is considered to produce the fastest and most
compact code and have the friendliest user interface. Advantages such as network and
19 3634_CH16
11:29 AM
Page 448
448 Chapter 16
multimonitor debugging and “edit and continue” (which lets the developer modify
code on the fly without fully recompiling the program) have served to make the life of
the game developer easier—at least in some ways.
The growing power and diversity of computer systems have also made the developer’s
life more difficult. The variety of hardware, and the subsequent complexity of programming the API—compounded by the amount of new information to absorb—is definitely
more than before. It’s like the development of scientific knowledge: In the eighteenth
century, it was possible for one person to know all the areas of science in depth. Now,
even after years of study, the best that could be hoped for would be to specialize in a
tiny area.
A Note on the Future
The last great taboo of game programming is C++. Up until the first imprint of this
book, C++ was considered too slow to be useful for game programming. The only
reason that C was accepted by game programmers was the assurances that you could
take a listing from a C program and (with some effort) predict what Assembler language output you would get. With C++, there were all sorts of strange goings-on in the
background, and it was not possible to directly relate C++ code to assembly.
Fundamentally, game programmers did not trust the compiler. They assumed that anything the compiler produced was inferior to their own efforts. In the early days, this
may have been true (assuming that they were a very skilled Assembler programmer),
but now, with the advances in compiler technology and the nondeterministic nature of
multitasking operating systems, this argument is invalid. This is the “Not Built Here”
(NBH) argument, and it is a particular favorite of the game programmer. This argument is obsolete: it is impossible to write a game nowadays for all but the smallest
systems without using the operating system (somebody else’s code)!
Of course, it may well be possible to produce better output than even the best compiler,
but the amount of time and money that would have to be expended would be out of
reach of all but the richest companies. The last famous example of “It’ll be finished
when we say so!” was Quake, for which Michael Abrash was hired (as he is widely
acknowledged as the king of Assembler optimization) to tune and retune their inner
loops, although, interestingly enough, the majority of the game was written in C.
19 3634_CH16
11:29 AM
Page 449
Current Development Methods 449
Although C++ has become much more prevalent in game programming, there is still
one compelling argument for using C: cross-platform compatibility. The implementation of C++ is varied slightly from platform to platform, particularly on the newer
features such as templates and namespaces, and this was the reason for Quake being
written in C. C has been standardized for years, and it is a stable language. You pretty
much know that a program written on one platform will recompile and function pretty
much as intended on another platform, unless you have made any processor-specific
However, the ANSI standard for C++ has been ratified. Hopefully, this will mean that
C++ will benefit from the source-level cross-platform compatibility currently enjoyed
by C. When this happens, there will be no further obstacles to using C++ for the
majority of cross-platform game development. Even so, Microsoft has been successfully
dodging the standard for five years and only recently have begun to claim 99.x%
compliance with the ANSI C++ standard.
This is the present, but what of the future? There will always be a hard-core contingent
that wants to exploit the machine to the max, but there have always been “game
maker” packages that attempt to allow nonprogrammers to produce their own games.
Nearly without exception, these were universally terrible. The advent of STOS (Atari
ST) and AMOS (Amiga)—programming languages based on Basic with game-specific
extensions, and their much more recent PC second-generation counterparts, Blitz Basic
and Dark Basic—were exceptions to this rule. While you couldn’t do anything spectacular with these languages—and although they were easier than learning C, they were
still quite technical and had quite an initial learning curve for a nonprogrammer)—
they could be giving an indication of what is to come. In fact, the first PC incarnation
of these, Klik-and-Play, took one step forward and two steps back: it allowed the use of
dragging and dropping of backdrops and actors to create a game, but this brought
inflexibility, and any games created with it tended to look as if they were about five
years out of date.
One of the advantages of the consoles over the PCs is the level of standardization. On the
consoles, everything tends to work the same way; on the PC, there is a plethora of different ways to do things. Standardization can make developers’ lives easier, and this was the
idea behind DirectX and other third-party libraries. The use of third-party components
such as this can not only save development time, they also ensure that the feature is
implemented to a higher standard than would usually be possible in the available time.
19 3634_CH16
11:29 AM
Page 450
450 Chapter 16
The Anachronism of Console Programming
Console development is, at least on the surface, very similar to the way games were
developed in the old days. Consoles have been traditionally programmed in Assembler,
much in the same way as the old eight-bit computers.
Only the most recent consoles (such as the Sony PlayStation 2) have broken this
pattern and used C as the main programming language. Before this, the lineage of
consoles traced through from the heady days of the Atari 2600 through the Nintendo
Entertainment System (NES), Super NES (SNES), Nintendo 64, and the Nintendo
Game Boy and Color Game Boy.
The market for console development tools has always effectively been a closed one.
The console manufacturer strictly controlled the availability of development kits, and
(particularly in the case of Nintendo), there were strict guidelines to follow before you
would be granted a development kit. Chris and Tim Stamper, founders of the U.K. software house, Rare Ltd. (producers of classic Nintendo 64 games such as GoldenEye 007
and Gamecube games such as Starfox Adventures), revealed in an interview that they
had reverse-engineered the NES in order to produce a game to show Nintendo what
they were capable of. Nintendo, being duly impressed by the work of the brothers,
signed them up as preferred developers and promptly gave them all the information
they had figured out by reverse-engineering the console in the first place.
Originally, the choice of development kit was limited to that provided by the console
manufacturer. These were usually quite expensive, and were not always the easiest
things to use. Pretty soon, however, third-party development kits came onto the market. One of the first was GLAM, which was developed in England and could target a
number of different platforms, including the Game Boy, NES, Sega Master System, and
a number of arcade boards. (This was the kit used by Rare for many of their products.)
These development kits were never quite as polished as their computer equivalents,
probably due to the inherently short lifespan and limited upgrade capacity of most
consoles. The development tools usually got only one major development cycle, unlike
Microsoft Visual C++, which is currently in its eighth incarnation. Obviously, this has
allowed for a lot of polishing and development of new features.
19 3634_CH16
11:29 AM
Page 451
Current Development Methods 451
This situation seems to be changing with the increasing prominence of consoles:
whereas originally they were seen as toys for small children, the massive success of the
Sony marketing machine made the PlayStation and Playstation 2 into the accessory of
choice for 20-somethings. What Sony started, Lara Croft certainly helped improve
(even though she seems to be sagging a little as of late).
This mass attention has caused a vast increase in the number of developers wanting to
produce console software, which has consequently created a demand for better tools.
Metrowerks, producers of CodeWarrior for the Macintosh, produces versions of its
product for major consoles.
Microsoft is also (somewhat surprisingly) keen to get in on the act: The Sega
Dreamcast used a version of Windows CE as an operating system, with a specially customized version of DirectX. To facilitate development, Microsoft released an extension
package for Visual Studio, allowing software to be written on PC and compiled for the
Dreamcast. The use of DirectX and Visual Studio in this way is meant to allow easy
development of cross-platform PC/Dreamcast games. If this works (there is some doubt
about the viability of the Dreamcast in the wake of the PlayStation 2), then the future
looks bright for the production of cross-platform games.
We left the previous paragraph intact from the first edition of this book. Even though
the Dreamcast was a good machine, it was (as we suspected) crushed by the mighty
force that is the Playstation 2. It will be interesting to see if history is to be repeated
with Microsoft’s latest attempt, the X-Box, which similarly to the Dreamcast, runs a
Microsoft operating system, allowing for easy porting to and from the PC.
Over the past few years, there has been a burgeoning interest in writing homebrewed
software for the consoles. Usually however, the development tools and hardware manuals
are restricted to official developers, and the fee to become an official developer is usually
beyond the budget of average individuals. This turned out not to be a problem, as there
is a large and active community of people on the Internet who hack the consoles and
share the information that they have found. This, coupled with the rise of the emulator
scene (you can now emulate practically any machine in existence on your PC, with the
Spectrum, the Commodore 64, the PlayStation, the Nintendo 64, and the Game Boy
Advance being some examples) meant that pretty soon a lot of free or shareware development kits were written by enthusiastic amateurs. In many cases, these kits are so good
that official developers use them in preference to those provided by Sony or Nintendo.
19 3634_CH16
11:29 AM
Page 452
452 Chapter 16
For example, for the Game Boy Advance, you can find full documentation of the hardware, programming tutorials, a fully featured GBA C++ Compiler and an emulator for
the PC that has a built in debugger. Using these resources, combined with a cheap
cartridge writer (that, due to fears of piracy, are now extremely difficult to obtain), you
can set yourself up with a full-blown unofficial Game Boy Advance development kit for
less than $200. Similar software and hardware exists for the other major consoles, too.
Game Boy Advance programming is an anachronism: due to the limited power of the
machine, programming for the GBA is the closest experience you can find to the original flavor programming with the old machines. It is the last computer left on the market
for which one person can be expected to be able to write a commercial quality game as
a single programmer. In the case of the GBA, it seems we have come full circle.
This book is not about producing games for the GBA, as the techniques presented here
would swamp the GBA, even if a C++ compiler were available for it. However, the GBA
does form a valuable part of the still-growing history of game development, and this
discussion would be incomplete if it were not mentioned.
By all means, go ahead and produce GBA games. It is a good way to learn about producing games with limited hardware, and some game developers would swear that—unless
you knew what it meant to have to scrape the bottom of the barrel for that last few
bytes of memory, or to hand optimize a critical loop in order to achieve a reasonable
frame rate—then you won’t be able to appreciate just how easy development is nowadays. Parts I, “Game Design,” and II, “Team Building and Management,” of this book are
still applicable, of course. Game design and project management the same, no matter
what the platform, but tread carefully when using the techniques in Part III, “Game
Architecture,” which may well be a bit too meaty for the GBA. The basic structure of a
game is the same on any platform, but the amount of work you can delegate to the
compiler in order to make your life as a developer easier most certainly is not.
The Present Day
So, where are we now? What is development actually like on the computers and consoles of today? In a word—easier. The amount of work that is automatically performed
for us by the compiler and the operating systems has increased dramatically.
19 3634_CH16
11:29 AM
Page 453
Current Development Methods 453
By producing a few libraries and frameworks to access the hardware interfaces, it has
become possible to concentrate more on the game logic, which is the fun part. All the
grunt work—machine initialization, joystick detection, hardware setup—has been
done for (or very nearly), and you can concentrate on the task at hand, rather than
battling with hundreds of possible hardware configurations.
Reusability—the concept of producing software components that are useful for more
than one project—appears to be a very difficult concept for the average game developer.
The average game developer believes that using shared components ensures that all the
output from the company will be the same. (“We can’t keep using the same engine or
all our games will look the same.”) That may well be true, and the world certainly
appears to have enough Warcraft and Quake clones to support this theory.
This is the difference between monolithic and component-based development and the
mindsets that go with them. Monolithic development is the practice of developing a
program in one large “lump.” This lump may consist of separate parts, but the parts
are so interwoven that they are inseparable. Each part of the system is interdependent
on the other parts, and the parts rely on internal knowledge of how they function,
rather than using the published interface, This reliance on internal knowledge means
that these parts are effectively immovable; it’s as if they have extended tendrils deep
into the hearts of the other parts, and have tangled each other up irretrievably.
The contrast of a monolithic system is (obviously) a component-based system, the sort
of thing that Microsoft has been touting for quite a while with its COM-based system.
Just because Microsoft has suggested it doesn’t mean that it is automatically a bad idea,
although it is a little bit heavyweight and is not yet supported fully on all platforms. You
can’t rely on it if you want automatic cross-platform capabilities. This may change in the
future, but for the time being, COM is for all intents and purposes Windows only.
Most game developers consider COM too slow. In most cases, this is simply a gut feeling, based on the (incorrect) assumption that anything C++—and especially everything
C++ and Microsoft—must be slow. This may or may not be true, but in any case, it is
not safe to make such assumptions without using a profiler to test the speeds. Bear in
mind, DirectX is COM-based.
19 3634_CH16
11:29 AM
Page 454
454 Chapter 16
We have not found any particular problems with COM, and neither have most PC
game developers; whether they know it or not, DirectX is a collection of COM objects,
and they have been using those for a while now!
The arguments for and against using COM are not as simple as “Is it fast enough?” A
much more fundamental concern needs to be considered first, and that is whether you
actually need to use it. COM is very powerful, and, for a lot of game development, it is
using a sledgehammer to crack an egg. COM is useful for designing operating system
components, because all details to the location of the object are stored in the system
registry (a sort of database that contains all the settings for the operating system), but
how often will you need to run your game out of a bunch of different directories?
The major advantage of COM is easy versioning. A COM object can export one or
more interfaces to a client. COM objects always export at least one interface,
IUnknown. This is the daddy of all COM interfaces, and supports just three main
operations: incrementing reference count, decrementing a reference count, and querying for an interface. This not very useful on its own, so a nontrivial COM object will
export more than just the one standard interface. These extra interfaces are obtained by
first obtaining a pointer to IUnknown and then asking it to provide a pointer to the
interface that you want to use, using the QueryInterface method. The golden rule of
COM objects is that, once an interface is published (in the wild), it cannot be changed.
If you need to add extra functionality to an interface that is already being used by
other programs, then you do it by adding a new interface to the object. This interface
can support all of the old functionality, and can also support the new features that you
wish to add. That way, both old programs and new programs are served adequately. For
example, if you had an older game that used the interface IAIEngine. And then you
released an updated and/or expanded version, you would probably name the new interface IAIEngine2. Any change to the functionality of an already published COM object
requires a new interface. It is best to do this, even if you have changed only the internal functioning and have not modified the external interface (unless you can absolutely guarantee that the way the object appears to behave externally has not been changed
at all).
Another advantage of COM—at least in theory—is that the objects are easily accessible
from other languages. For example, we use a COM object that we have written that
exports the interface, IBHCodecCOM. This is a compression/decompression object that
can compress or decompress to and from areas of memory, resources built into the
19 3634_CH16
11:29 AM
Page 455
Current Development Methods 455
executable file, or files on the hard disk.
We also have written a number of programs that use this object to decompress files,
and they all use it from C++. The COM object can be treated just like a normal C++
object, and there is no difference in syntax, except for a call to CoInitialize to initialize
the COM system.
Sometime after writing this object, it occurred to us that it would be useful to have a
graphical browser of archive files, so we could check and update the contents.
What did we do? Did we spend a couple of weeks writing an MFC graphical browser?
No, we fired up Visual Basic and wrote a user interface similar to WinZip in a couple
of hours. While it wasn’t the prettiest thing in the world, it did the trick, and was
usable enough to release to other people as an internal tool. Accessing a COM object in
Visual Basic is as simple as referencing the type library (a library of interface information built at compile time), declaring a new object, and using the methods provided
just like you would with any other Visual Basic object.
A word on the speed of COM objects: the typical response of game developers to suggestions of COM is that it will be too slow for “real” use. Of course, it depends how
you define “real.” While we wouldn’t suggest that it be used in the most critical inner
loop of your game, everywhere else would be fine. With a good compiler, there is not
much more overhead to using a COM object compared to using a standard C++ object.
19 3634_CH16
11:29 AM
Page 456
20 3634_CH17
11:30 AM
Page 457
Chapter 17
Initial Design
• The importance of a
good architecture design
• Platform considerations
• The problem domain
• Tokenization
art I of this book covered the game design process, from
taking the germ of an idea through to a fully fledged
game design, with some interesting detours into what a game
actually is. This chapter picks up where Part I left off. You
have your cool game design nicely specified, and now you’ll
turn it into the killer game you’ve been aiming for. Okay,
guys; get coding!
Well, that’s what usually happens.
If you’ve been paying careful attention so far, you will have
noticed that most game development goes directly from a
game design to the programming phase. You also will have
noticed that the games industry is the only place where this
occurs. Why? Because the games industry hasn’t yet caught
up with modern development techniques. Current thinking
is that the games industry is some five or so years behind the
rest of the industry. Okay, so the programming itself is up to
date, but the techniques are a bit behind the times, much
like building a car using period techniques (but modern
tools). Five years in this industry is a long time.
Developers do not write technical design documents because
it helps them kill time. Rather, they write technical designs
as roadmaps of the way ahead. The technical designs
improve project visibility and allow someone who has a view
of the bigger picture to see how the work fits in with the
overall company plans. There’s no point in having a developer beaver away for weeks on a CD-playing module if the
• Token interaction
• Finite-state machines
• Events, states,
transitions, and
20 3634_CH17
11:30 AM
Page 458
458 Chapter 17
team next door produced one a couple of months ago, or if the company bought one
and it’s on the shelf gathering dust behind the office copy of Quake III. There is also
not much use in building a module that nobody else knows about.
The technical design is an advertisement. It tells people what you are doing, so they
can fit their work around yours. It helps to prevent unnecessary duplication of work,
and it improves the overall visibility of the development process. No production-level
coding should start until the technical design is complete (subject to the 80/20 rule).
Of course, a number of intermediary phases must be gone through first. Before the
coding can begin, the design has to progress from the biggest picture (the initial game
design) right down to the smallest picture (the technical design for an individual
module). So put away your coding hat—which you’re not going to need for a while—
and pull out your architect’s hard-hat, and read on.
This chapter covers the initial phases of the technical design process, and its aim is to
take a well-specified game design and produce from it a well—specified overall architectural design.
As stated before, the ubiquitous 80/20 rule again comes into play. It is not possible to
fully complete an architectural design before beginning coding. You should, however
be able to complete approximately 80%; the remaining 20% will be modified and
rewritten as the project progresses. An important point to realize is that the architectural design is never “finished,” but will be constantly evolving as development proceeds. There is nothing wrong with this; in fact, it’s the sign of a healthy project. It’s
impossible to be able to predict with 100% certainty what will be needed before you’ve
actually got your hands dirty with the code, but, with a good design, the amount of
rewriting will be minimized. It will then be a simple process of refinement and clarification of the document. At this point, we must stress the importance of keeping the
documentation in line with the code. (This has been covered before: the documentation work should be considered an integral part of the development work. The code
itself is just a small—but important—detail tacked onto the end.)
The classic argument against having to do all this “boring” documentation is that the
code is the best documentation possible. That can be true, of course, but it shouldn’t
be, because if the code is the “best” documentation, then the documentation is for all
intents and purposes, nonexistent. This is a chicken-and-egg scenario. The code is the
20 3634_CH17
11:30 AM
Page 459
Initial Design 459
best documentation only because nobody bothers keeping the real documentation up
to date. What you must consider is that you will not be the only person reading this
documentation. In theory, anyone who wants or needs to know exactly how the project
is going—anyone from another member of the technical team to the game designer to
the company president.
Another classic response is that the code is the only documentation required, and anybody who can’t read it probably doesn’t need to. If you really believe that, we cannot
help you. Maybe you should choose a new career—a more solitary one, in which you
won’t have to interact with people, such as a lighthouse keeper.
Then again, if you tell your company president that she should be able to read the
code for herself when she asks for a technical design document to check the progress,
the decision to choose another career may be already made for you!
The Beginning
A lot of hard work has gone into your game thus far. The game design has had to jump
a fair number of hurdles before reaching the stage it is at now. It has satisfied the company management, the marketing department, and the game designer’s peers that it is a
game worth developing. It’s been through a number of iterations and refinements, and
will continue to do so throughout the development process. The sign of a good design
is that these refinements will become increasingly minor as the technical phases begin.
Radically redesigning a game halfway through development does not tend to make for
happy developers.
In this chapter, we will consider three game designs, and how we would take them
from the design phase through to the architectural phase.
Everybody should be familiar with two of these games, and the other one, Balls!, is
the game that was designed (and partially written until a burglar made off with the
computer containing the source code—let that be a lesson to those of you who think
that off-site backups are a waste of time) as a test-bed for some of the concepts in
this book.
20 3634_CH17
11:30 AM
Page 460
460 Chapter 17
Before we even discuss these, we must consider something else that is even more
fundamental—the platform(s) on which the software is running. The architecture of a
game has two main aspects: the hardware abstraction interface and the game abstraction itself. The hardware abstraction interface is conceptually easier to deal with,
because the architect is providing an abstraction for something that already exists.
There is no real point in providing concrete abstractions for phantom hardware,
although it’s not a bad idea to try to consider future developments—at least in the
interface design.
The second aspect is the game itself. All games share a few standard aspects—which
almost count as hardware interfaces themselves—such as the menu and options system, and on a lower level, the main game loop. When we venture out of this familiar
territory, we are on our own (or nearly so, as there are very few problems in modem
computing that someone else hasn’t already solved). Game developers tend to like the
feeling that they are pioneering uncharted territory, or that they are boldly going where
only a few have gone before (and that they are doing it better). Of course, syndrome
is endemic within the games industry, and even the developers who licensed the
Quake II/III engines proudly boast about how many modifications they have made.
It’s almost as if they cannot admit that they used an engine “as is,” because that would
make them seem like lesser developers—or worse still, little more than glorified level
designers. The trouble is, that whatever these guys do with the Quake II/III engine, the
games are going to look inherently similar anyway. The problem is in the hardware:
any given 3D card tends to produce similar looking output, no matter what graphics
engine is forcing the polygons through it.
The game itself—the way the game world is defined, and how the objects contained
within that world act and interact—is effectively an uncharted area of design. Sure,
there are a few standard guidelines and ideas, but only you, as the architect, will
have an intimate knowledge of the game world, its inhabitants, and their interactions.
This knowledge can be put to good use simplifying the design, as we’ll see later.
The balancing act comes in making sure that the simplification does not turn into a
restriction if new elements that had not previously been considered are inserted into
the game. Some of the ugliest hacks in history have been directly due to this. (We
should know, as we’re responsible for a couple of them.)
20 3634_CH17
11:30 AM
Page 461
Initial Design 461
The software factory methodology (described in Part II of this book) recommends producing components that perform common functionality so that they could be reused in
future projects. One of the things about game development that amazes us the most is
the almost total abhorrence of reuse. In fact, up until the first printing of this book, the
only reuse we had ever seen was of the less-than-useless cut-and-paste variety.
With careful design, you can produce an abstraction that maps nicely onto multiple
platforms, making the task of porting easier.
We can already hear the seasoned game developers scoffing that if you abstract a
system to be cross-platform, you’ll lose too much in speed. It’ll be unworkable, they
are thinking. These people might find it helpful to read Case Study 17.1 in order to
gain a fresh perspective on just what is possible with a sensible design.
Case Study 17.1 Abstraction in Quake II
Quake II, by Id Software, caused a huge stir when it was released a few years ago.
Pretty soon after the original release, the game was also released on a number of
other platforms, including Linux.
This was made possible due the advanced design of the software. Quake II, fundamentally designed as a multiplayer game, is based around the client-server architecture model, even in single-player mode.
“Client-server” is an architecture whereby the majority of the functionality (AI calculations, game mechanics, etc.) takes place in a server process, and the results of
the processing are made available to the clients.
In Quake II, the responsibility of the client was to send user input to the server and
to render the results dispatched from the server.
As you can imagine, this made the multiplayer capability easier to develop than, for
example, a peer-to-peer system, in which each of the machines involved in the game
accepts input from all of the others and runs its own AI calculations. With a clientserver setup, all of this is centralized, and so all the client has to do is to connect to
the server to upload and download data.
20 3634_CH17
11:30 AM
Page 462
462 Chapter 17
All very impressive, but what does this have to do with cross-platform capabilities?
The server was written in mainly portable C, in a machine-independent fashion.
This allowed it to be easily ported to another platform. Any platform-specific code
was partitioned out into a separate library. The client itself was written in a similar
way, interfacing with the server through a platform-independent interface.
This is how the single player game worked; as far as the server is concerned, it does
not know whether the client is on the same machine, or half a world away across
the Internet. Any platform-specific code was also hidden behind an interface.
For example, the graphics renderer just receives a pointer to a block of memory to
draw to. There is no reliance on platform-specific architecture. This method is
repeated throughout the code wherever an interface is needed with the hardware.
The example of Quake II shows that platform independence through abstraction is
possible and doesn’t necessarily have to impact performance. Nobody seems to complain about the performance of Quake II.
Hardware Abstraction
Hardware abstraction is very useful in game development. Apart from the obvious, the
main benefits are the homogenization of the development environment and the development of reusable components that allow easy setup of the hardware without all the
grunt work usually associated with it. One of the reasons for all the grunt work—even
with libraries such as SDL, DirectX, or OpenGL—is that APIs are designed to be as
flexible as possible. Millions of people use SDL, DirectX, and OpenGL, and they all
probably have different uses in mind. The main result of this in-built flexibility is
increased complexity.
A company that knows the plans for its next few releases can take advantage of this
knowledge by producing a hardware abstraction library that takes into account the
needs of the future releases.
20 3634_CH17
11:30 AM
Page 463
Initial Design 463
For example, you may develop a screen-handling library that supports a common
range of resolutions, double or triple buffering, bitmapped and TrueType fonts, and
sprites. We know that most games now use 3D, but let’s keep this example simple for
the time being, and anyway, 3D is just a presentation method. Many games that are
presented in 3D nowadays would work just as well (if not better) in 2D, but marketing
pressures say otherwise. At least some developers are realizing that 3D isn’t everything,
and are not trying to shoehorn 3D into places where it doesn’t belong.
Graphics Hardware Abstraction
So let’s take a closer look at our example. For Balls!, we developed a simple graphics
library that performs all of this setup for us. We needed only basic functionality: a
screen setup, bitmapped font support, sprite support, transparency and translucency,
and some temporary working buffers for compositing some of the more complex
graphics. Bear in mind that when we wrote this, 3D acceleration was only just coming
into play. We aimed to provide only the level of functionality required for Balls!. For
more information, read the game design documents included in Appendix A.
The graphics abstraction layer shown in Figure 17.1 is fairly simple, and has lots of
room for improvement. However, it is a good demonstration of the sort of thinking
that goes into abstraction.
Figure 17.1 The graphics abstraction layer.
20 3634_CH17
11:30 AM
Page 464
464 Chapter 17
When we were designing our abstraction, we thought in terms of a console or handheld. We did not want to do anything overly flashy with the system, so that it would
be easier to port onto less capable hardware.
The actual hardware of the graphics card itself is wrapped by a class called
CDisplayCardObject. This class is private and is used internally by the rest of
the graphics classes. The only access to this class that is externally provided is an
initialization function that creates a system-wide singleton that is globally accessible.
(A singleton is an object that is instantiated only once per process.)
The initialization function is needed because the system needs to know to which
window all output will be directed.
We took as the root the concept of a drawable surface, encapsulated by the class
CSurface, a surface that can receive graphical output. (In reality, it’s just an area of
memory that DirectX knows about.) A graphical surface can be created in two types of
memory: system memory or video memory (local or nonlocal). The class CSurface
provides functionality that is common to all surfaces, such as initialization. The surface-initialization member takes parameters that indicate the height, width, and bit
depth of the surface, whether it is 3D capable, and where it is to be located in memory.
Other private members provide error reporting, and placeholders for restore capabilities and surface-release capabilities. The class CSurface is never instantiated directly,
because not all functionality has been implemented. That must be done in derived
classes. (CSurface is technically known as an abstract base class.) All other surface
classes derive from it and provide or override the needed functionality as necessary.
The only publicly available members of the CSurface class are those for querying the
height, width, and bit depth of the surface.
The CScreen class derives from the CSurface class. CScreen is a singleton. (As most
games still tend to run on one screen, this is not a particular problem.) If we ever
decide that we need two or more screens—or more than one virtual screen, as in the
case of picture-in-picture—then a number of small modifications will allow this.
20 3634_CH17
11:30 AM
Page 465
Initial Design 465
The CScreen class provides a wrapper for a double or triple-buffered surface that is
displayed on the physical screen. It can operate in windowed or full-screen mode completely transparently to the programmer. Even though there are some subtle differences
(when using DirectX) between windowed mode and full screen mode, these differences
are taken care of in the class, and both modes appear programmatically identical.
The rest of the functionality of the screen class is fairly simple. Debug text output is
provided, as are pixel-setting capabilities, although they are slow. The usual range of
surface-querying functions are available, such as querying for width, height, bit depth,
or transparency capabilities. Support is also given for screen or area clearing, buffer
flipping, and copying areas of the screen to another surface, so that they can be modified elsewhere, and then redisplayed.
The CFrame class encapsulates sprites and working surfaces that are to be displayed on
the screen. (The name CFrame was taken from the animation frame concept, because
we felt that the word sprite didn’t describe the functionality too well.) One of the
important things in this sort of design is naming. Make sure that the name of an object
aptly describes its function.
The first important thing to note about the CFrame class is that there are three methods of initialization. No initialization can be performed until the CScreen object has
been created. This is because all CFrames are dependent on the screen, and an internal
pointer to the CScreen object is statically maintained within the CFrame class.
The first—and simplest—initialization member lets you create a frame with read/write
functionality, initialized to a given color. You can specify whether you would like the
frame to be created in system memory or video memory. Usually, although video
memory is faster to display, it is slower to modify than system memory and is often
a limited resource. In the intervening time since this class was created, AGP memory
on PCs has effectively removed that limitation. It might still be a problem on other
platforms though, so the interface to the class still allows the choice.
The second initialization function allows you to create a read-only frame initialized
with given image. This image can be located in a file or resource and can be retrieved
from compressed archive. The decompression and loading is encapsulated in another
class and is transparent to the user. Because the CFrame class keeps track of what
20 3634_CH17
11:30 AM
Page 466
466 Chapter 17
images have been loaded, it does not unnecessarily duplicate images in memory—it
uses a copy-on-write system to minimize the number of duplicate images in memory
simultaneously. If an image has already been loaded, then any attempts to reload it will
cause the encapsulating CFrame object to reference the original surface rather than create a new one. This conserves resources and provides useful tracking and debugging
functionality. The developer does not have to worry about managing surfaces because
they are automatically taken care of. All of this grunt work occurs during initialization,
so there is no worry about the game being slowed by extra overhead, even if it means
that surfaces initialized from bitmaps have to be read only, because writing to the surface of one CFrame object would otherwise affect all of the other CFrame objects that
referenced it. This is where the copy-on-write system comes into play. The code detects
that you are modifying a shared image and creates a new unique copy of that image to
The third initialization function lets you specify a source CFrame instance, an (x, y)
offset into that surface, and the width and height of the new surface. This creates a
read-only surface that is merely a window onto the larger surface. The most common
application of this is displaying individual sprite frames that have been loaded onto a
larger surface, because this is faster and more efficient. Figure 17.2 illustrates how this
works. The sprite frame is a window that shows a small portion of the larger surface.
Figure 17.2 Sprites packed onto a large surface.
20 3634_CH17
11:30 AM
Page 467
Initial Design 467
Other functions provided by the frame class include a draw member that draws the
sprite to the screen (with transparent areas if required) and an alpha drawing member
that allows the sprite to be drawn to the screen and be translucent so that anything
behind it can be partially seen.
The CBitmapFont class is simply a collection of CFrames that have been initialized
with a set of ASCII character bitmaps. Because fonts are not usually drawn to the
screen as frequently as most other types of graphics, the underlying surfaces are created in system memory by default to conserve video memory. The CBitmapFont class
has a print member that will take a text string as a parameter and output it to the
screen at the specified coordinates, allowing you to center the text vertically and/or
horizontally, and query a string to get the length in pixels.
The above example, although simplistic, demonstrates the reasoning behind the
abstraction. For the Balls! project, we wanted a simple interface to the screen and
drawable surfaces. We did not want to have to worry about whether we had drawn a
picture to the wrong buffer, and we did not want to have to specify which buffer to
draw to every time it was required. We also wanted a consistent interface, whether we
were drawing into a window or were using full-screen mode.
This class library solved those problems. Setting up the hardware is as simple as creating a screen object, and then initializing it with width, height, bit depth, and whether
it should be double or triple buffered and windowed or not. Then the programmer can
create some frames, call their draw members to draw them onto the back buffer of the
screen object, and call the buffer-flipping member of the screen object to flip the back
buffer to the front position, making it visible.
The DirectX setup code behind this is rather complicated, but that is not a problem. All
of the complex (and thus slow) code is away from the critical execution path. What do
we mean by critical execution path? Well, if you ran the code through a code-profiling
utility, such as TrueTime (Compuware), the critical execution path would be the functions that are called most frequently and take up the lion’s share of the execution time.
For example, the initialization code for each instance of an object is always going to be
called only once, so nobody really cares whether it is lightning fast or not (unless you
are calling it thousands of times, in which case some optimization work may be
required). On the other hand, the screen-flipping and frame-drawing members are
going to be called very frequently, and so these need to be as fast as possible. In this
20 3634_CH17
11:30 AM
Page 468
468 Chapter 17
code, we strive to ensure that the object is fully set up and checked for errors during
the initialization phase, so that there is a minimum of error checking during the execution of the most frequently called members.
A common complaint by game developers is that hardware interfaces such as those
described above restrict flexibility and provide too much overhead. This complaint may
well be valid, but on the whole, the time spent calling the member functions of the
class interface will be minimal compared to the amount of time actually doing the real
work within them. Your game will be spending most of its time in system calls.
The perceived lack of flexibility is no real problem either. In the case of generic thirdparty libraries, the underlying library is most likely too flexible for your needs; after
all, it has been designed to be applicable to every possible use a developer could want.
Haven’t you ever heard of the expression “too much rope…”? What is the problem with
ignoring parts of the functionality that you are never going to use, and making your life
simpler by producing an easy-to-use interface module? Even more importantly, a simple
interface can be more easily implemented on other platforms, so that porting a game
to a new platform involves only rewriting your hardware interfaces. Compare this to
having liberal sprinklings of DirectX throughout your code. Are you going to rewrite
DirectX for the Macintosh and PlayStation 2? (If you are, be sure to send us your source
code!) By writing your own cross-platform interface, you may have to provide simple
capability querying functions (similar to those in DirectX except much simplified) for
some of your more advanced functionality, but most of the current platforms have
very impressive and (fortunately) similar core capabilities, even though the underlying
hardware and the programming interfaces are radically different.
Sound Hardware Abstraction
The sound hardware abstraction is based on a similar model to the graphics system. A
good thing to consider when designing components is to aim for a basic similarity of
structure across different modules where possible. This means that new modules can
be learned easily from experience with current modules.
The CSoundCardObject directly wraps the sound card object, a private singleton
accessed through an initialization function. There is no sound-hardware analogy to
the screen object of the graphics abstraction, as it is pretty much included in the
CSoundCardObject object.
20 3634_CH17
11:30 AM
Page 469
Initial Design 469
The CSound object wraps the sound. The usual range of member functions are
provided: playing, pausing, stopping, changing volume and frequency, and specifying
the position of the sound in 3D space. The initialization function provides parameters
for dictating which of this functionality you want enabled in the object.
A similar system to that used with the graphics object prevents loading of duplicate
sounds. Instead, a new object is created that refers to the original data.
If a sound file is too large (and the size limit can be set using a CSoundCardObject
member function), a small buffer is created that loads the sound, piece by piece, on
demand when it needs to be played. The small overhead for doing this has so far not
been a problem.
Figure 17.3 shows the similarity between the abstraction for the graphics hardware and
that for the sound hardware. This theme can be extended for most of the hardware that is
present on a machine. The use of abstraction in this fashion almost creates a virtual
machine, providing a consistent interface that can be written to, whatever the platform.
The windows implementation of the libraries used for Balls! performs a lot of W1N32 and
DirectX jiggery-pokery behind the scenes, but none of this leaks out beyond the watertight interfaces. In theory, if the libraries were implemented on other platforms such as the
Macintosh or the GBA, then the main body of the source code could be recompiled without too many changes, Unless your company wants to fund two or more separate teams to
produce versions for different platforms, then the abstraction library is probably the best
way to go, especially if the underlying platforms have basic similarities—for example, 3D
accelerated consoles with DVD-type storage and limited RAM/EEPROM storage.
Figure 17.3 The sound hardware abstraction layer.
20 3634_CH17
11:30 AM
Page 470
470 Chapter 17
The important thing about hardware interfacing is that it can be viewed as building
upon an essentially stable platform. Even though new features are being added to
DirectX and other libraries with every release, they are usually backward compatible.
By building an interface, you are effectively consolidating and simplifying the original
interface, providing functionality that is suitable for your projects and ignoring the
rest. As we have already stated, most of these APIs are very general, designed to be useful
to the widest possible developer base. By wrapping this stuff up in an easy-to-use interface, it becomes more specific and more suitable to the requirements of your project.
Other Hardware Considerations
With a little thought, we’re sure that you can design simpler interfaces for the other
aspects of the hardware needed for a game.
For example, on top of the sound object, we have a CJukeBox object that plays the ingame music and operates in a similar fashion to a normal jukebox (except we don’t
have to insert money). It can play selected tracks, use auto-repeat over a number of
tracks, or just randomly select tracks to play. In order to set the tracks to be played, a
member function is provided that allows the insertion and removal of tracks (analogous to which records are placed in the jukebox). The tracks that can be played are
MIDI files, CD tracks, Ogg files, MP3 files (don’t tell the RIAA!), and MOD files (a
music format that originated on the Commodore Amiga).
To support these different types of music are a number of separate classes, all derived
from the base class CTrack. This base class provides an interface that the jukebox uses
(for querying the track name and controlling play, pause, and fading the track’s volume
in and out). All the CJukeBox object needs to do is call the correct member of the
CTrack object, and the tune will play, irrespective of the underlying format. The jukebox doesn’t care what it is, as long as it provides the CTrack interface. The jukebox
itself is just a fancy collection object that allows the selection of tracks and ensures
that only one is played at a time.
A similar library has been built to take care of user input from diverse sources. These
resources include the obvious (keyboard, mouse, joystick, and joypad) and the not-so
obvious (such as network and scripting).
20 3634_CH17
11:30 AM
Page 471
Initial Design 471
These last two options are not suitable for all situations. The network interface merely
allows user input across a network. The scripting interface (bizarrely based on a form
of extended SQL combined with LISP) is very useful for a number of diverse situations, such as AI interfacing, attract modes, and automated testing. At the core of this
system, each object that needs to be moved (such as the player) and all the mobile
game objects implement a control interface. This control interface does not know what
is driving it, be it network, player, AI script, or whatever. As such, it would be easily
possible to allow the player to control any object that provides this interface. This is a
very important point, and you should be beginning to get a realization of just how
powerful these architectural techniques can be.
All games, no matter how trivial, require a timer of some sort. In the early days, the
timer would be the machine itself. The main loop of the game would simply run as fast
as possible, and tie itself in to the clock speed of the processor. This was because the
early machines were lacking in power, and there generally was never time to spare.
Another consideration was that all machines were identical, and so there was no need
to normalize timing, unlike today’s PC, where your game could be expected to run on
any system across the vast range of speeds available. It would be no good if your game
ran at different speeds on different computers (unless it’s a turn-based game such as Sid
Meier’s Alpha Centauri), although amazingly, a lot of the earlier PC games did so.
When game developers were initially faced with this new problem, they took a fairly
simplistic approach. What has surprised us the most is that, even up until comparatively recently, this approach is still occasionally presented as a sensible solution—and
not always by amateurs. Every now and then, we still see frame rate based timing suggested as a solution to the timing problem. This is fine for a uniform hardware platform,
such as the GBA or another console, but absolutely horrendous for use on the PC.
These game developers attempted to use the frame rate as a timer. They would choose
a frame rate (possibly dictated by the hardware)—say, thirty frames per second (fps)—
and construct all their game logic around that. There would be one AI tick every
frame, so doing the math, you can see that each game object would be updated to
move the distance it could normally move in each thirtieth of a second. Note that this
technique is still used with platforms that are limited in power and universally identical, such as the GBA—it has the least overhead. No special effort is required to dynamically adjust the timing parameters. This is shown in Figure 17.4.
20 3634_CH17
11:30 AM
Page 472
472 Chapter 17
Check Controls
Fixed Frequency
Check Controls
Check Controls
Figure 17.4 Ye olde game loope.
This was all fine and good, but there were two main problems. The first was that owners of faster machines weren’t too pleased that their brand-new game didn’t take advantage of their machine’s speed, and that the game looked exactly the same on their
friend’s machine that had a processor half as fast.
The second (and rather more serious) problem is that, below a certain threshold level
of processing power, the game would suddenly halve in speed. If the AI tick took just
longer than one frame to calculate, then every other frame would be skipped, resulting
in a drop in speed. The same sort of thing could happen if a frame took too long to
draw. Even if it only happens once in a while, the effect on the game is a noticeable
jerkiness in frame rate.
Fortunately, some thinking was done on the subject, and a couple of suitable solutions
were developed. We’ll call them semi-decoupling and full decoupling.
Semi-decoupling is the simpler of the two solutions. The trick is to decouple the frame
rate update from the AI ticks. This technique is suitable for games that don’t particularly stretch the hardware of the average midrange machine, but also want to run
acceptably well on the lower-end machines. The concept is shown in Figure 17.5.
11:31 AM
Page 473
Initial Design 473
Check Controls
Fixed Frequency
20 3634_CH17
If not calculating AI
Calculate AI
Figure 17.5 A semi-decoupling game loop.
The AI runs at a fixed rate as before, but the frame rate just runs as fast as it possibly
can. Assuming the AI is running at 30 ticks per second as before. A separate loop
draws each frame, and then waits for the next AI tick to be complete before drawing
the next one. This means that, if the frame takes too long to draw, it would not be
such a disaster as before. The AI loop would continue independently, because it
wouldn’t be stalled waiting for the frame to complete, and the next frame to be drawn
would be correct. The player may notice a drop in frame rate, and some of the animation may be slightly jerkier, but the important thing is that the internal Al would still
be running correctly. The main disadvantage of this technique is that your maximum
frame rate is limited to the fixed tick rate of the AI loop. There is no point in updating
the screen if nothing has changed in the AI, so the drawing loop will be idle while this
is going on. This technique is fine, if you are sure that your AI is going to run at a constant rate and you don’t mind being limited to that rate. But what if you want your
frame rate to be the maximum possible, irrespective of the speed of the AI loop? This
is where full decoupling comes in. Most games today use this technique.
Full decoupling, as shown in Figure 17.6, attempts to run the AI loop as fast as possible.
A reference timer is used to adjust the changes in game object states per AI tick, so that
the game does not appear to speed up and slow down. This means that, on faster
machines, there will be more AI ticks per second, so, for each tick, the delta value (which
is the reciprocal of the length of the tick) will be smaller. On slower machines, ticks will
take longer, and so the delta value will be larger. Imagine that we are animating a clock
face that has one hand. This hand rotates once every second. Internally, the AI tick
updates this clock according to the delta value. The clock is drawn at the end of each
20 3634_CH17
11:31 AM
Page 474
474 Chapter 17
AI tick. Obviously, the smoothness of the animation depends on the frequency of the AI
ticks. The animation will still show the hand performing a full rotation every second, but
the smoothness of the animation will vary. This is illustrated in Figure 17.7.
Fixed Frequency
Check Controls
If not calculating AI
Calculate AI
Use 1/frequency
as delta value
for calculation
Figure 17.6 The full decoupling game loop.
Real-life motion of hand
4 Al ticks per second
tick 1
tick 2
tick 3
tick 4
3 Al ticks per second
tick 1
tick 2
tick 3
Figure 17.7 The effect of tick length on screen updates.
This is a pretty smart technique that allows a program to max out frame rates, but it
is not true full decoupling. The frame rate is still limited by the rate of AI ticks. It is
impossible to have more than that with this technique.
11:31 AM
Page 475
Initial Design 475
To achieve full decoupling would be difficult, but possible. To do this, the AI loop
and the frame-update loop run completely separately, most likely in separate threads.
Then the frame loop is given the ability to take a “snapshot” of the state of the AI loop
whenever it likes (assuming that it is not so often that it chokes the AI loop), as long
as at least one visible object has been processed. Care has to be taken to make sure
that an object is not halfway through being processed when the snapshot is taken (it
may be in an internally inconsistent state), and the easiest way to achieve this is to set
the snapshot granularity to the object level.
To explain this concept in more detail, let’s assume that we are processing 50 AI
objects per tick. Let’s also assume that the graphics hardware is powerful enough to
take a snapshot three times per tick. This means that we will get a request for a snapshot every 16.67 objects. Because we are only two-thirds of the way through processing a particular object at this stage, it may not be a good idea to allow the snapshot to
be taken. In this case, we delay the snapshot until we have finished the current object,
allow the snapshot to be taken, and then continue with the next object. This is shown
in Figure 17.8.
As fast as possible
Check Controls
As fast as possible
20 3634_CH17
Calculate AI
If at a suitable place to
pause calculation of AI
to allow drawing then
Draw Frame
Use 1/frequency
as delta value
for calculation
Figure 17.8 The real McCoy!
When used correctly, this technique allows for some truly phenomenal frame rates
to be achieved on powerful hardware. For most practical purposes though, semidecoupling is often good enough.
The timer object in Balls! attempts to use the best timing facilities available on the
given machine. If a particular facility is not available, then the next best is used, and
so on until all possible options are exhausted (an unlikely occurrence). Despite all of
20 3634_CH17
11:31 AM
Page 476
476 Chapter 17
this complexity, the class that provides this functionality, CTimer, has a simple design
based on handshaking. This class provides an initialization function that allows you
to specify the timing interval, and where you would like the timer to notify when triggered. Pause, resume, and stop functionality is provided, as would be expected. When
a message is received from the timer, there is a requirement to call the timer message
acknowledge function. The class then times the interval between when it sent the
message and when the receipt was acknowledged.
This information is then packaged and sent with the next timer message. This allows
for some pretty accurate delta timing, and is suitable for all three of the techniques
mentioned above.
“Not Built Here” Can Be Better
Although the proverbial “seasoned” game developer may scoff at these ideas, he or she
may not be able to in the future. The new generation of machines from the likes of
Sony, Nintendo, and Microsoft are becoming incredibly powerful. Today, many game
development organizations typically work in teams of 10 to 15 for a period of 24-36
months at a time.
There are fears that the new generation of powerful machines will mean that only the
largest and most cash-rich development houses will be able to afford development and
that the smaller development houses will either disband or be swallowed up by these
corporate giants. The fear is that the destruction of these smaller houses will drown us all
in a wave of corporate mediocrity. Since the first imprint of this book, we have witnessed a
large consolidation within the games industry, partly brought on by the recession, combined with increasing development costs. It seems that our fears have been realized.
The development community is proposing a solution, driven by a mixture of market
pressure and consensus. As an example, Sony actively acquired third-party technologies in order to augment the libraries provided as an interface to the hardware of the
PlayStation 2. A number of smaller companies are producing third-party modules that
can be used in your development. Modules such as physics engines have recently
become the vogue.
20 3634_CH17
11:31 AM
Page 477
Initial Design 477
These sorts of modules are going to become essential as the hardware increases in
power. It is pure fantasy to expect a small team of developers to be able to compete
against the big boys in terms of technology. Those days are (sadly) gone. The main
problem, we think, with the increasing technology is that the gameplay is becoming
less important. When a development team has to spend most of the time developing
ways to harness the technology, it doesn’t leave much time for developing the gameplay. This is why bringing in third-party modules is going to become more and more
essential if a development house wants to be able to bring out a technically excellent—
and, more importantly, playable—product. Half-Life, by Valve, is a perfect example
of this. For Half-Life, the developers brought in the Quake II engine, made some
modifications to suit their purposes, and then spent the rest of the time developing
the sophisticated AI and storyline. In doing so, they probably saved themselves 12
to 18 months of development.
Even the artwork required for these engines is becoming more detailed and time consuming to produce. As the power of the technology grows exponentially, so does the
effort required to produce cutting-edge games for it. What can be done? If an increasing amount of time is spent on harnessing the technology, obviously a decreasing
amount of time is being spent on the gameplay. The provision and use of third-party
components such as 3D engines and physics engines is really the only way of staving
off this disaster waiting to happen.
But it doesn’t stop there. Music, sound, and artwork can also be bought off the shelf.
Libraries of 3D objects are available from companies such as Viewpoint Datalabs and
have been used in games such as Activision’s Interstate ‘82 and Microsoft’s Combat Flight
Simulator. Viewpoint supplies ready-made and animated models, and this can cut months
off a tight schedule. Why pay artists to produce a perfect reproduction of a town hall,
when you can buy a ready-made version that can slot directly into your game?
You can hire a musician to produce suitable music for your game, and you can also
buy CDs full of top-quality, royalty-free samples. These products all go some way
towards reducing the development time—and, consequently, the development costs—
of a game.
As soon as developers begin to realize the benefits, they will make more use of these
third-party components. All of the basic functionality—the graphics engine, sound
system, music, save game system, AI system, artwork, and music—will be brought in
20 3634_CH17
11:31 AM
Page 478
478 Chapter 17
off the shelf. This will allow the game developers to concentrate on what is really
important—the gameplay.
We feel only sorry for those developers who don’t realize this in time. Many developers
may have their heads in the sand, and they are going to miss the technological revolution that is coming to the game development world. Unless they are in there right from
the start, they will have a difficult time catching up.
The previous two paragraphs, unchanged from the first printing of this book have
proved to be rather prophetic. In 2003, middleware is the rule, not the exception for
most game development teams. Maybe this is because those that did not adapt, died.
The Twilight Zone
But what of things that are pretty much common to all games, but not part of the
games itself and not part of the hardware either? We are talking about setup programs
(where applicable), loading and saving of games and data, and menu systems that
allow you to configure the game.
These intermediate-level components can be reusably implemented in many ways.
They can be frameworks that support the internal structures required to implement
the features required. For example, most games employ some sort of menu system
that allows starting new games, configuring settings such as sound or music volume,
redefining controls and screen resolutions, and the like. This is one of the most obvious areas to target for reuse. In virtually all projects we have seen, the time required
to implement the menus and other supporting areas that aren’t part of the actual game
engine itself has always been severely underestimated. Consequently, it is usually
rushed through at the last minute.
With a little more thought, it would be possible to create a generic framework that
allowed menu screens, option screens, volume screens, and any other “out of game”
screens to be constructed quickly and easily. I’m sure that some companies must have
done this, but whoever they are, they have either done it so well and their framework
is so configurable that each new product looks like there is completely different code
driving the user interface. Or maybe they just hack out new code each time. Given the
current state of game development techniques, we’ll leave the jury out on that one.
20 3634_CH17
11:31 AM
Page 479
Initial Design 479
Of course, you don’t necessarily have to write your own framework: you can license
one that another company has developed. And, let’s face it, if another company is
marketing a framework or component for a particular purpose, they are likely to have
expended much more effort on making it applicable to that purpose. In short, it’s likely
to be a lot better than whatever you or your team are going to be able to produce in a
reasonable amount of time, given that you are at the same time focused on producing
the game itself.
Another point to consider is that the developer of the component or framework may
also have a view of the bigger picture, and you can benefit from this. The classic example would be DirectX. If you’re developing a game for the PC, do you write your own
interface to the hardware, or do you just use those provided by DirectX? This one’s a
The Problem Domain
With the hardware interfacing taken care of, it is time to look at the interesting part.
This is the only part that is unique to your game (or at least should be). All of the rest
of it is not unique. Everybody is developing for it. Some can do it better than others, but
that is not so important any more. The power that is now available in hardware (where
previously it would have been written in software by the developer) is now available to
all. The “indefinable” quality that makes your game different is the game itself.
The game design has been taken care of in the first part of the book. Here is where
we start taking this game design and converting it into something workable that a
developer can understand, and that will be easy to expand, debug, and maintain over
the coming months.
A good design can turn technical “issues” into complete non-issues. With a good architecture, questions such as “How could we do this?” can become less and less frequent.
The object of a good architecture design (accompanied with the good dissemination of
information) is to make it obvious how to do something: the only way possible—the
best way.
20 3634_CH17
11:31 AM
Page 480
480 Chapter 17
Earlier in the book, we introduced the concept of “hard” architecture and “soft” architecture. Up until now, we have been discussing the hard architecture: the interface
with the hardware, and the various housekeeping tasks, such as the game loop, that
virtually all games perform.
Now we are into a much more fuzzy area—the soft architecture, the part of the game
that is (or should be) unique to the game. Just because something is unique does not
mean that the framework that it is based on need be disposable. We should strive to
achieve reusability here, too.
How can we do this? Well, for a start, we can exploit certain factors that are common
to the soft architecture of virtually all games. Okay, so what are these factors? How can
we be sure that they are common to all games? And how can we be sure that building
within such a framework won’t impede development?
What Is a Game? (Revisited)
Let’s take a random walk through a list of games and see if we can spot any commonality:
Pong, Frogger, Pac-Man, Elite, The Sentinel, Missile Command, Tetris, The Legend of
Zelda, Virtua Fighter, Carmageddon, Defender, Chess, Warcraft, Zork, Doom, FreeCell,
Scrabble, Tennis, Balls!.
What do all these games have in common? The first thing is that they all have a player.
A game without a player is not a game: it’s a movie.
All games also have discrete elements that are directly or indirectly manipulated by the
player. For the time being, we shall call these tokens. These tokens are the elements
of the game that are supervised and managed by the computer. For example, let us
consider Pong, as shown in Figure 17.9. What are the tokens in this game?
20 3634_CH17
11:31 AM
Page 481
Initial Design 481
Figure 17.9 Pong.
Pong has two players (represented by bats and a score), a ball, and some walls. Which
of these are tokens? Conceptually, they all are. Each player controls a bat, and the players’ prowess with this bat will improve their score. Therefore, it could be said that the
players’ bat and score are subtokens of the token representing the player. The ball is
another token, as it is indirectly influenced by the players and, aside from small interactions, is essentially managed by the computer. The walls of the play area themselves
are tokens. They interact with both the bats and the ball, in order to prevent them
escaping from the top and bottom of the screen.
What about the goal areas? These are the areas that the ball travels to when the player
has failed to deflect the ball in time. When the ball reaches here, a point is awarded to
the opposing player, and the game begins again. These are tokens too—although they
are defined by area and have no visible representation.
All the games mentioned above (in fact, empirically speaking, all games) can be
described in terms of players and tokens. We don’t have the space or the time to go
into why all the games can be decomposed into players and tokens, but we have yet to
find a game that cannot. We’re going to cover two more in this chapter (Pac-Man and
Balls!) and leave the rest as a proverbial “exercise for the reader.”
20 3634_CH17
11:31 AM
Page 482
482 Chapter 17
Thinking in Tokens
We can also consider these tokens to be arranged in a form of hierarchical structure.
The playing area, or game world, in itself is at the top of the hierarchy. From then on
in, it is an essentially flat hierarchy. This is shown in Figure 17.10.
Figure 17.10 The Pong token hierarchy.
The game world token contains all the other tokens. Obviously every token has to
operate within the game world in order to form a part