Python code to do multi-valued states with OpenSpiel.
Multi-valued states is from "Depth-Limited Solving for Imperfect-Information Games" by Noam Brown, Tuomas Sandholm, Brandon Amos (2018) and was used (with multi-valued states for both players, which we refer to as "matrix-valued states" in the code in this repo) in superhuman NLHE agent Pluribus, described in "Superhuman AI for multiplayer poker" by Noam Brown, Tuomas Sandholm (2019).
This code is from my private repository with all of my game AI code. This code was part of a project with collaborators to do continuous resolving on a large game with collaborators, that we eventually tabled. I also used part of the code for Recon Blind Chess @Neurips 2021. I'm open-sourcing it now so anyone can look at it and use it.
I put the code up here without cleaning it up. It is definitely not "release-ready".