Iterative deepening

Iterative deepening makes a chess engine search first with a small limit and then repeat the analysis with progressively larger limits. In chess, depth is commonly measured in plies, where one ply is one move by one side.

This initially looks wasteful because positions near the start are visited repeatedly. Earlier iterations, however, produce valuable information: a provisional best move, the principal variation, meaning the line currently considered best, stored results, and evidence for move ordering. Searching promising choices first in the next iteration often lets more branches stop early and offsets much of the repeated work.

Its clearest practical benefit is time control. If the clock forces the engine to stop during a deeper iteration, it still has the complete result from the previous one. It does not have to risk everything on a search that may remain unfinished.

Iterative deepening does not mean extending only one favored line. Each iteration revisits the starting position with a larger limit, although selective techniques may assign different local depths inside the tree. It also moves the search horizon outward without eliminating it.