Designing a hybrid position evaluation function for Chinese-chess computer game

The position evaluation function plays an important role for building an intelligent Chinese-chess computer game (CCCG) player. The mostly used evaluation functions include standard heuristic evaluation function (SHEF) and self learning evaluation function (SLEF). SHEF depends on the board position features to large extent, but it hardly includes all the features due to the limit of knowledge of the designer. And, SLEF can explore the knowledge hidden in the current position which is difficult to find in SHEF. In this paper, a hybrid position evaluation function (HPEF) is designed by fusing SHEF and SLEF. The temporal difference learning (TDL) is used to train the proposed HPEF on professional game records. Based on HPEF, a CCCG player called CCCGp is developed. We experimentally validate that HPEF is quite effective by competing with different kinds of testing players. With the help of HPEF, the intelligent level of CCCGp can be improved incrementally with the increase of number of professional game records HPEF learned.