An Approach to Estimating Decision Complexity for Better Understanding Playing Patterns of Masters

This paper proposes a method for estimating decision complexity of positions by using correlation coefficient between two evaluation values of the root node and leaf level of a game tree. Search is performed to determine the evaluation value at the root node with focus on the ratio between positive and negative values of position scoring at leaf nodes. Moreover, Kalman filter is employed to minimize the variance of scoring errors. Several games, played by masters and computers in the domain of shogi, are analyzed while applying the proposed method. The results show that the proposed idea is a promising way to better understand playing patterns of masters, for example, identifying the right moment for possibility of changing strategies such as speculative play and early resignation.