Selective search

Selective search answers the enormous number of possibilities in chess. Every legal move opens several replies, and every reply opens further continuations. A branch is one such possible sequence. Because a person cannot examine every branch to equal depth, they choose a small number and spend most of their time there.

Selection begins with candidate moves, the options chosen for serious examination. The player uses known patterns, features of the position, and forcing moves, such as checks, captures, and direct threats that restrict replies, to decide what deserves attention. Promising branches are explored more deeply, branches with a refutation, a reply proving that the candidate fails, are dropped, and calculation should reach positions stable enough to evaluate.

Efficiency and risk

The benefit is efficiency: calculation is directed toward decisions likely to matter. The danger is missing an unexpected candidate, fixing attention on one idea too early, or rejecting a defense before understanding it. Research using protocol analysis, which records what a person verbalizes while deciding, can reveal which branches were visited, but it cannot make the selection correct.

In computer chess, the term also describes techniques by which a chess engine, a program that calculates positions, reduces or extends parts of its search. The broad principle is similar, unequal allocation of resources, but programmed criteria are not identical to human attention and pattern recognition.