Search in Co-Wiki

Late move reductions

game-theory 538 tokens 10 outbound links

Late move reductions

Most of the chess programs (or engines) typically search the first one or two moves in full depth. If the score of the first few moves are lower than alpha, the move is assumed bad. However, if the score of the moves are larger than alpha, the reduced search tells us nothing so we will have to do a full search (called as a fail-low).

This search reduction can lead to a different search space than the pure alpha–beta method which can give different results. Care must be taken to select the reduction criteria or the search will miss some deep threats.

Description Late move reductions work on the idea that the higher a move is on a sorted list, the better it likely is. When an engine evaluates a node, it uses heuristics to order moves so that the most promising lines, such as captures or those suggested by the killer-heuristic, are searched first.

See also * [[efficiently-updatable-neural-network|NNUE]] - A neural network which is used to evaluate moves (or calculate the approximate "score" of a move). * Stockfish (chess) - A popular strong chess engine that uses NNUE as it's [[evaluation-function]]. * YaneuraOu - A shogi chess engine that first implemented NNUE.

References ## External links *