This is my attempt to understand CFR as a part of my final project for CS3243 - Intro to AI. I would like to thank https://int8.io/counterfactual-regret-minimization-for-poker-ai/ for the comprehensive explanation. We also submitted a paper which can be viewed here.
In this limited version, the user only allowed to raise for four times in one round street. For more information about the project, one can head over to https://github.com/changhongyan123/mypoker.
- Make AI better and extensive by changing branching scheme of chance nodes.
- Find a way to generate all possible scenarios for the algorithm.
- Compute nash equilibrium incrementally (will not make a huge time difference). The main time difference was obtained by playing / sampling games and updating the strategy together.