Texas Hold'em Poker AI

Overview

A research prototype exploring decision‑making for No‑Limit Texas Hold'em using a mix of counterfactual regret minimization (CFR) style iteration for strategy refinement and lightweight opponent modeling. The goal is to play balanced, adaptive poker that generalizes across table sizes and blind structures.

Features

• Hand strength evaluation with equity simulation across possible boards.
• Abstracted action buckets for raise sizing and mixed strategies.
• Exploit/defend toggles via opponent frequency tracking.
• Training loop for iterative regret updates and policy convergence.