Estimating the Probability of Winning for Texas Hold’em Poker Agents

The development of an autonomous agent that plays Poker at human level is a very difficult task since the agent has to deal with problems like the existence of hidden information, deception and risk management. To solve these problems, Poker agents use opponent modeling to predict the opponents next move and thereby determine its next action. In this paper are described several methods to measure the risk of playing a certain hand in a given round of the game. First, we discuss the game of poker and the expectation in its events. Next, several hand evaluation and classification techniques are described and compared in order to determine the advantages of each one. The current methods to deal with risk management can help the agent’s decision. However, in future work, the integration of these methods with opponent modeling techniques should be improved.