Estimating Winning Probability for Texas Hold'em Poker

Among all the technologies in creating a good poker agent, estimating winning probability is a key issue. In this paper, we propose an approach to estimating winning probability for Texas Hold’em poker. We design a data structure using both the observable data from the current board and the history. A Support Vector Machine classifier is trained and 5-fold cross-validation is employed. We create a poker agent with some decision making strategies to compete. Experimental results show that our method has outperformed three other agents in precision of estimating winning probability.