Bigram Realization Probability for Game Tree Search

In chess-like games in particular Shogi (Japanese Chess), Realization Probability[1](RP) is an indispensable tool for calculating move ordering and game tree search expansions/reductions. In this paper, we propose a brand new way of calculating Big ram Realization Probability (BRP) based on Tsuruoka´s Realization Probability and show the efficiency of move ordering and node expansions/reductions. It is a kind of Realization Probability that whole features are based on the information of a current move and its previous move. In our experimental results, a Shogi playing program implementation by BRP wins 62 -- 38 with the program based on original Realization Probability.