An offensive and defensive expect minimax algorithm in EinStein Würfelt Nicht!

EinStein Würfelt Nicht/(EWN) is an imperfect information game that has randomness. The traditional methods reduce the randomness of dicing by the Expect Minimax Algorithm (EMA). However, they only involve partial chess knowledge, which brings inaccurate evaluation and low-precision game tree search. In this paper, from the offensive and defensive views, the evaluation function is optimized on the two forms of win (taking the corner or capturing all the enemy's pieces) in EWN. And then, an Offensive and Defensive Expect Minimax Algorithm (ODEMA) is proposed to find the optimal move. The experiments show that ODEMA has gotten 75% winning ratio with traditional EMA.