Unused GDC 1999 Roadtrip Presentation

1. Context

Tim Cain discovered an old PowerPoint titled "GDC 99 Road Trip" in the archives of his first machine at Troika Games. He has no memory of whether he submitted it and was rejected, or simply never submitted it. At the time (early summer 1999), Troika was only a year old and about a third of the way through Arcanum's development — he was also teaching a class at UCI. The presentation is titled "RPG 2000: Design Issues for the Next Millennia".

2. Defining an RPG

Cain offers a strict definition with five requirements:

• Character creation — The player must make their own character, not play a pre-defined one
• Character-expressed choices — Player choices must be expressed through the character they built
• Character-based success — Action success should depend on the character's level, not the player's skill (distinguishing RPGs from shooters)
• World reactivity — The world must react to character actions with visible, understandable consequences
• Player-driven story — The character should be immersed in a story driven by their actions

2.1. What Doesn't Qualify

Under this definition, Cain explicitly excludes:

• Diablo — An action game, not a real RPG
• MMOs (Ultima Online, EverQuest) — No player-driven story, limited world reactivity with thousands of players
• Adventure games — Usually no character creation, limited player agency, linear stories
• Games with "RPG elements" — If a game has to tell you it has RPG elements, it's not an RPG

3. Character Creation: Class vs. Skill

3.1. Class-Based Systems

• Easier to design (skills grouped logically, balance is more manageable)
• Easier for players to learn and play (clear roles)
• Favor group play — classes designed with rock-paper-scissors dynamics (tank, damage dealer, crowd controller)

3.2. Skill-Based Systems

• More complex for developers and players
• More rewarding when done well
• Favor solo play — the player builds a character they believe can handle anything
• Tabletop RPGs were already trending toward skill-based systems in 1999

4. Story Architecture

4.1. Stories vs. Quests

Cain draws a clear distinction:

• Stories provide the framework — the reason to enter the world, the bread of the experience
• Quests provide the depth — diving into world lore, NPC relationships, and player rewards

4.2. Story Arcs

Stories should be arranged as arcs with mandatory nodes the player must reach, but the path between nodes is entirely player-determined based on character abilities:

• A stealth character might sneak past guards, pick locks, or climb walls
• A fighter might kill everyone in the way
• A speech character might charm, bribe, or talk their way through

The same story node feels different because the context of arrival differs — a fighter faces hostility inside the castle, a talker may have been given a badge granting access.

4.3. Quest Design Principles

• Quests should have preconditions (story act, NPC relationships, items found)
• Quests should track states (unknown, mentioned, accepted, in progress, failed, succeeded)
• Quests should be modular — easy to add or remove during development
• Story-critical quests are harder to swap but can usually be replaced with alternatives

5. Linear vs. Nonlinear Play

• Pure linear — Easy to pace, but feels confining; nothing the player does matters
• Pure nonlinear — Freeing but directionless
• Cain's preference — Stories provide linearity (fixed nodes), quests provide nonlinearity (variable paths between nodes), and world simulation adds reactive nonlinearity

6. World Simulation

NPCs should react to multiple character elements:

• Appearance, race, equipped items and armor
• Actions taken, dialogue choices made
• Quest solutions (stealing a key vs. picking a lock vs. killing guards vs. bribing someone)

Violence should be just one option among many. Too many games treat it as the only option.

7. Combat: Real-Time vs. Turn-Based

• Real-time — Exciting and immersive but limits tactical complexity
• Turn-based — Allows deeper tactical decisions; better separation between player skill and character ability (a 70% hit chance feels character-driven, not player-driven)
• Player preference for combat style significantly determines a game's demographics

8. AI Requirements

Cain identifies five AI domains RPGs must handle:

1. Combat AI — How NPCs fight, use abilities, choose targets, take cover
2. Conversational AI — NPC dialogue reactions, guard awareness of player actions
3. Merchant AI — Pricing, inventory, interaction with player skills
4. Skill-integrated AI — How player abilities interact with all systems (healing, crafting, repairs)
5. Companion AI — Fundamentally different from world AI; companions fight for the player and may leave if the player acts against their values

8.1. Player Expectations for AI

• Smart (or at minimum, not stupid)
• Reactive to player actions
• Predictable enough for players to strategize around

Cain notes a pet theory: pets as followers work well because their smaller action sets make them seem smarter and more predictable.

9. Multiplayer Considerations

• Single-player favors story-driven design
• Multiplayer favors quest-driven design (groups prefer lore-diving over advancing one player's storyline)

9.1. Multiplayer Challenges

• Quest synchronization between players at different progress points
• Monster respawning rules (time-based, player-count-based, level-scaling)
• AI struggling against coordinated player groups
• ISP/connection issues beyond developer control
• Pausing problems — one player stopping the game for everyone
• Time commitment coordination and lobby systems
• Griefing — a topic Cain says could be its own entire talk

10. Camera and Viewpoint

• First-person — Great for immersion
• Isometric/third-person — Great for group play and tactical decisions
• Some players have a strong first-person preference and will refuse to play isometric games

11. Licenses and Genres

11.1. Licensed Games

• Pros: Reduced learning curve, pre-existing scenarios and villains, built-in player familiarity
• Cons: Restrictive licensor demands, player expectations can work against creative freedom, licensing costs

11.2. Genre Expansion

RPGs were "finally leaving the D&D nest" in 1999. Tabletop RPGs had been doing non-fantasy genres for years; computer RPGs were just catching up. Some players resist non-fantasy settings as outside their comfort zone.

12. Localization

A large RPG can have over 300,000 words — equivalent to about six books. Localization challenges include:

• Voice acting locks text in place; changes require re-recording at additional cost
• The "Holy Trinity" of content restrictions: violence, sex, and nudity vary dramatically by country — what can be shown, described, or implied differs across markets and must be planned before development begins

13. Cain's 1999 Prediction

RPGs would become far more complicated — not just in features, but in environment simulation, AI sophistication (combat, dialogue, NPC reactions), and storytelling depth. He predicted the rise of what he calls "professional storytelling" with much richer characters.

14. References

• Tim Cain. YouTube video. https://www.youtube.com/watch?v=q8otGbKSOo0