How AI Meaningfully Improves Prototyping
Generating Prototype Assets
Generating card decks, tabletop mats, game boards and rules drafts is a perfect task for AI. These are things that are often iterated quickly, sometimes few times in an afternoon. Manually updating them and keeping them consistent is simply a bad use of time. We store raw game data in spreadsheets and text files, and use SVG or open-sourced art for our prototypes. AI is then used to generate assets from those sources. We do not use AI to generate raster artwork.
Testing Rule Changes
AI can grind through scenarios quickly to find weaknesses or confirm a solid design. In one early prototype of Shepherding Cats, we had two very carefully crafted movement decks for both the Pope and the cats. It was discovered that the game skewed in favor of the Pope early and in favor of the cats as the game progressed. We used AI to iterate our card decks in 10000 move increments to find the weaknesses in them and identify the scale of the changes that would be needed to balance them. Interestingly, after playtesting, that version of the game was abandoned in favor of the much more straightforward and fun version we have now.
Finding Gameplay and Thematic Overlap
The simple fact is, many game designers have similar ideas and many games tread on similar territory. Identifying the overlap in mechanics or themes is a huge task. But not for AI. We have used AI to help identify games where our ideas have significant overlap with others to ensure that we are not unknowingly reskinning an idea that someone else already had. In the case of Robot Prospectors and the Rings of Saturn, it resulted in changing the game from a Cold War-inspired sci-fi espionage game, with the working title "Space Spies" and some frustratingly standard mechanics, into the game we have now. It has made our game designs better.
Lessons Learned
Overall, we have found AI to be a great tool for shortening prototype development and testing timeframes. In the process we have also learned a few lessons:
- If you need AI to tell you what’s wrong, you may have more serious problems.
- AI can act as another playtester in your development chain and keep you from putting half-baked ideas in front of real people.
- AI requires a firm hand. Simple questions can produce pages of output. Stick to your plan. Make sure AI understands what you want and then use it as needed to help you get there.
Subjective Things AI Has Surfaced:
- For tabletop board games, fun can be inversely proportional to the number of cards in the game. Cards can manuipulate the game experience by covering up fundamental design flaws, adding needless complexity or stifling the creative joy of playing the game. Additionally, players who know the card decks can be at an advantage over those who don’t. In our case, the archived prototype of "Wine Country Dynasty" suffered from card rash. After carefully reconsidering it, we came up with "For the Glory of Mars", which borrows some of the best ideas from Wine Country Dynasty but removes most of the cards.
- Complex mechanics should not require micromanagement. The best games have a few core mechanics and hide complexity from players. Again, "Wine Country Dynasty" provides a great example here. After many iterations of the game, various mechanics and feature orphans were all competing for space. Using AI, we were able to quickly identify the deadwood and either remove it or consolidate it. While the changes didn't save the game, AI did help with the process of weeding out needless complexity and focus us on what really mattered. Tabletop games are intended to be played with a group of people in less than 2 hours. They should be fun!
- AI is a great tool, but nothing replaces the joy of seeing real humans loving your game. ❤️
Happy gaming, everyone!