Solving winning probability problems in Go

Positional analysis needs to be made in the Go game, and it is true for professional players or Go program. The present Go program regards the territory scores as a major standard in its positional analysis, which is a flawed approach. In this article, an approach is proposed to calculate the winning probability, and the model parameter is further optimized by genetic algorithm. Through this approach, various winning probability models based on different levels can be obtained. After the test, it is concluded that the precision of this model can be applied in the module of middle game and end game in the computer Go.