In this paper the same (in my view misguided) view is taken:syzygy wrote:No, regular minimax does not assume that your opponent evalutes positions the same way as you. That really is a misconception. Regular minimax simply optimises the evaluation against all possible counterplay.
Putting your money on the assumption that your opponent is inferior is extremely dangerous under most circumstances. Only when swindling is the only option left could it be of help.
Wrong. That computer programs don't adjust their play to the opponent is for the simple reason that they use the same search engine against all opponents... Same search, same evaluation. Nothing to do with any assumption that the opponent uses the "same strategy".While human players adjust their playing strategy according to their opponent, computer programs, which are based on the minimax algorithm, use the same playing strategy against a novice as against an expert. This is due to the assumption of minimax that the opponent uses the same strategy as the player.
That minimax is symmetric merely reflects the fact that it deals with zero-sum games. Chess is a zero-sum game. Minimax is therefore the correct approach for chess. In my maybe not so humble opinion, the approach of the paper's authors is simply flawed.
The paper contains a rather painful statement:
The most interesting part is the adaptation of alpha-beta pruning. Regular alpha-beta relies on the evaluation for white being the opposite of the evaluation for black. This does not exclude asymmetric evaluations; it just means that if v is the position's value from white's point of view and w is the position's value from black's point of view, then v + w = 0. If v +w <> 0 in general, then some pruning can still be done provided that a bound on |v + w| is known.Tic-tac-toe on a 3 x 3 board. There is a simple winning strategy for this game. (...)