Code for the paper "Enhanced Equilibria-Solving via Private Information Pre-Branch Structure in Adversarial Team Games" (published on ariXiv and UAI 2025.
-
Choose the algorithm: CFR or CFR+. In
runner.py, set the--algorithmargument to eithercfrorcfr+. -
Set the number of players, ranks, and suits. In
runner.py, configure the following arguments:--players: Number of players.--rank: Number of ranks (e.g., 3 for 3-player poker).--suits: Number of suits (e.g., 4 for hearts, diamonds, clubs, spades).
-
Choose stopping condition: number of iterations or time limit. In
cfr.py, modify thecheck_funfunction to set the stopping criterion. You can stop the CFR iteration based on:- A fixed number of iterations.
- A predefined time limit.
stop_time: the total runtime limit (in seconds);checkEveryTime: how frequently to output logs.
The entry point of the code is the python script runner.py.
It currently supports {CFR, CFR+} as regret-minimization algorithms, which can be run on instances of the game varients of {Kuhn, Leduc, Goofspiel}.
If you find this project helpful in your research, please cite our paper:
Chen Qiu et al. “Enhanced Equilibria-Solving via Private Information Pre-Branch Structure in Adversarial Team Games.” UAI 2025.
Our code is built upon the open-source platform contributed by Tommaso Bianchi.