A simulation system to support computer poker research

Abstract

The development of Poker agents is a meaningful domain for AI research because it addresses issues such as opponent modeling, risk management and decision-making under uncertain information. The competitiveness of Poker agents is typically measured through simulation systems that run a series of games. However, current systems do not provide an adequate toolset for as-sessing the agents capabilities since they were built to play and not specifically for the creation or validation of new agents. Hence a new simulation system intended for research was created, featuring a modular and easily expansible architecture. This system accommodates bankroll management component, allowing the evaluation of the agents survivability between games, with limited initial recourses. An evolutionary simulation module was also included with the purpose of improving agents that have evolving strategies. The simulator will support further research into Computer Poker, thus fomenting the creation of an autonomous agent that considers all game components.