Optimal Independent Chip Model (ICM) Using Reinforcement Learning

Current poker tournament strategies often rely on the Independent Chip Model (ICM), which estimates players’ tournament equity based on chip stacks. However, traditional ICM models provide only a simplified approximation of tournament dynamics and are not directly integrated into practical decision-making agents. At the same time, recent advances in deep reinforcement learning (RL), particularly Deep Q-Networks (DQN), have demonstrated strong performance in sequential decision-making problems but have seen limited application in tournament poker settings.

 

To address this, this project proposes combining a learnable neural-network-based ICM module with a Deep Q-Network framework to develop more effective tournament poker agents. By training these networks using rewards that reflect tournament outcomes and prize distribution, the goal is to learn decision-making strategies that better capture tournament equity considerations and improve upon the limitations of classical ICM approaches. This hybrid architecture aims to produce stronger and more practical tournament-playing agents through deep reinforcement learning.