Texas Hold'em Poker Bot: A CFR/DLS Implementation

This repository contains a implementation of a No-Limit Hold'em (NLHE) poker bot designed for 6-max games. (Inspired by Pluribus)

Key Features

• Counterfactual Regret Minimization (CFR): Employs CFR variants for computing strategies
• Blueprint Strategy: MCCFR training generates a comprehensive strategy profile covering numerous game states
• Depth-Limited Search : Integrates search (inspired by Pluribus) to refine blueprint strategy decisions based on the exact game state encountered during play
• Buckets: Utilizes abstraction techniques to group strategically similar hands, managing state and action space complexity
• Comprehensive Game Engine: Includes a detailed game engine accurately modeling rules

Installation

1. Clone the Repository:

   git clone https://github.com/ivng8/PokerAI

2. Install Dependencies:

   pip install -r requirements.txt

Usage (Command-Line Interface)

Train a blueprint strategy using CFR.

python main.py --iterations 5000 --checkpoint_interval 500