Doubt regarding the minimax algorithm

0

The minimax, by definition, is great if the two players perform optimally, right? However, can we say something when one of them plays great and the other does not? In the case of the old game, as the state space is small, the player that acts with minimax never loses to a random player for being able to explore the entire search tree. And in the case of a game with more space, like Checkers? What can we say when a player makes bad plays against another player with minimax? The minimax will not be more great, is it?

By definition, the minimax algorithm is optimal only if the two players are playing against each other, optimally, is not it? However, we can say something when one of them plays optimal and the other does not? More precisely, what can we say when a player makes bad plays against another player with minimax with the aim that the minimax player plays worse (make some bad moves)? Thanks !!

    
asked by anonymous 19.04.2018 / 00:42

1 answer

0

The Minimax algorithm is great in the sense that it will always make the play that minimizes the maximum loss (that is why it is called MiniMax), this fact is independent of how the other player plays, it is clear that depending on the play of the adverśario, which minimizes the maximum loss may be different, but the algorithm always chooses the best move for it in that situation, it is important to note that great means the best possible and not the perfect move, it is possible to play optimally and lose (or tie) the game, simply because the lowest possible maximum loss is greater than 0. So, the answer to your question is: independent of the opponent, Minimax will always make the best move, with the metric being the lowest possible maximum loss, for games with the largest search space, the tree becomes very large, in those cases use other algorithms (such as alpha-beta prunning ) which is a variation of the Minimax.

    
02.05.2018 / 14:42