Anti-computer tactics

Anti-computer tactics are strategies that human players use to try to defeat a computer opponent in board games such as chess and Arimaa, particularly during the period when chess engines were strong but still beatable by a well-prepared human. The tactics exploit a structural weakness of traditional game-tree search: a computer opponent can only search a position to a limited depth before it must evaluate the board and choose a move, so an advantage that only pays off beyond that search depth, sometimes called playing over the horizon, is one the machine cannot detect. A common anti-computer approach is therefore to play conservatively, steering the game toward a long-term positional advantage that develops too slowly for the computer's search to notice, rather than seeking a sharp tactical advantage a computer would spot immediately. Garry Kasparov attempted tactics along these lines against IBM's Deep Blue in 1997, without success. As chess engines moved from traditional search toward machine learning based evaluation, these newer programs tend to play the long game as well as, or better than, human opponents, which has made classic anti-computer tactics considerably less effective than they once were.