Generalized game trees

摘要: We consider two generalizations of the standard two-player game model: different evaluation functions for players, and more than players. Relaxing assumption that players share same function produces a hierarchy levels knowledge as deep search tree. Alpha-beta pruning is only possible when behave identically. In extending model to minimax algorithm generalized maxn applied vectors N-tuples representing evaluations each If we assume an upper bound on sum components player, lower individual component, then shallow alpha-beta possible, but not pruning. best case, asymptotic branching factor reduced (1 + √46-3)/2. average however, does reduce factor. Thus, found be effective in special case with common function.