Research On Texas Poker Game System Based On Risk Model

Computer game has always been one of the most challenging research directions of artificial intelligence and it consists of perfect information game and imperfect information game. With rapid development of computer science and technology in recent years, a lot of technology and difficult problems have been resolved in perfect information game. And, the perfect game has achieved very good results in game matches hold at home and abroad. However, the research of imperfect information game started relatively late. Due to restricted by various conditions, there are still many unsolved problems. That players in imperfect information game are unable to get all information and unable to acc urately know what strategy will be taken by the opponent is similar to some situation in reality. So, the research in imperfect information game can provide reference for the similar situation in reality. So, it is important and necessary to research imperfect information game.Texas poker is a typical research object in the field of imperfect information game. The key research problems of it are multi-agent game, risk analysis and risk aversion, opponent modeling and behavioral decision and processing. The main content of this paper is that according to the above problems, we achieve an imperfect information Texas poker game system on the basis of forefathers’ research. It uses risk analysis to get the reason of the loss in the processing of choosing action, and then classified the loss, and gives the method to estimate the loss. We used artificial neural network algorithm in the processing of opponent modeling to train opponent models, then, we use these models to predict the moves of the opponents and get the action probability table. This table will be used in the decision model. And then, we cluster the opponent models with group clustering algorithm and get group models that perform between general model and special model. At last, we analysis risk and decided our action with risk dominance model. Risk dominance model can find a balance on the use of buff in poker game and it neither deliberately ignoring bluff and it will not lead to irrational abuse.At last, the above modules form an imperfect information Texas poker game system that base on risk dominance model. And the system is verified that it can do better decisions on bluff with annual computer poker competition platform.