Home

GamesmanNova

GamesmanNova (or Nova) is a research project developed at the University of California, Berkeley, dedicated to the efficient computation of optimal strategies in large extensive-form games. It does this via a standalone binary with many game interfaces and implementations within an extensible software system, made accessible through multiple user interfaces.

What is Nova for?

Nova is centered around efficient search. In the context of games, this means looking into many possibilities of where the game might go in the future based on the players' actions and the underlying ruleset, and generating the data necessary for optimal decision making.

Not all games are the same; this project is specifically for large extensive-form games, also known as extensive or sequential games. (For smaller game models and more equilibrium concepts, see the wonderful Gambit project.) Since Nova is focused on efficient search, it lends itself towards computing strong and weak solution sets. The data that conforms these "solutions" precisely answer the prompt, "if I give you any game state, tell me the best decision to make" (in the strong case) and, "if I give you a single game state, tell me the best decision to make" (in the weak case).

While complete-information deterministic games are the primary focus of the project (for which these "solutions" can give definitive answers to what the outcome of a game will look like under perfect play), there is also significant room for the analysis of incomplete-information or non-deterministic games. While some might argue that it is less satisfying or fruitful to research probabilistic games, it would be a disservice to practicality as they are arguably the most "realistic" cases.

History

This project's inspiration is due to GamesmanClassic, a project created by Professor Dan Garcia and contributed to by numerous undergraduate and graduate students. Nova was created in the Spring of 2023 in an independent effort with the intention of making a reformulation of GamesmanClassic with richer and more general interfaces which would be more ergonomic to use.

Note from author

I am elated that the project has received as much attention as it has, even if it is a comparatively small amount. I think that game theory as seen from a computational perspective encapsulates the essence of rational decision-making, highlights our shortcomings as human actors, and gives us a chance to digest the beautiful structure of mathematical games and strategic situations.

What makes me most excited about this project is being able to share the love I have for this subject with others in a constructive way. There are also many vectors of research I am excited to see develop; in particular, I am looking forward to seeing the intersection of collaborative game theory and applied computer science grow into a useful policy-making toolset for the distributed optimization of interactions within both human and machine systems.

-- Max Fierro, 03/06/2024