README
Research on Counterfactual Regret Minimization (CFR) algorithms for poker, Game Theory Optimal (GTO) solvers, and AI poker tools.
Documents
| Document | Description |
|---|---|
| Development Journey | Complete timeline of building the CFR poker solver (Phases 0-4) |
| Apps and Services | GTO solvers and AI trainers you can play against |
| Hardware Requirements | CPU, RAM, and server specs for professional solvers |
Related Research
- packages/cfr-poker - Personal MCCFR implementation for Limit Hold’em
- Game Theory & CFR - Foundational research on CFR algorithms and Nash equilibrium
- GPU Optimization - GPU acceleration findings and optimization strategies
Background
CFR (Counterfactual Regret Minimization) was first published in 2007 by Martin Zinkevich at the University of Alberta. It’s the core algorithm behind most modern poker solvers, proven to converge to Nash Equilibrium given enough iterations.
Notable milestones:
- 2017: Libratus defeats top human players in heads-up NLHE
- 2019: Pluribus defeats pros in 6-player NLHE
- 2023+: Cloud-based neural network solvers (GTO Wizard AI, Deepsolver) achieve near-instant solves