Taboun 0.1
Chooses a random legal move.
The intentionally simple baseline. It scores no point at real time controls, so it is listed unrated instead of anchoring the scale.
View this version →Chooses a random legal move.
The intentionally simple baseline. It scores no point at real time controls, so it is listed unrated instead of anchoring the scale.
View this version →Searches with minimax and evaluates material.
Introduces forward search and material-aware decisions. Fixed at 1000 as the origin of the relative rating scale.
View this version →Uses minimax with a simplified positional evaluation.
Adds piece-square knowledge instead of counting material alone.
View this version →Searches three plies with alpha-beta pruning.
Prunes irrelevant branches while preserving the 0.3 evaluation.
View this version →Orders captures and checks before quiet moves.
Makes alpha-beta pruning more effective through move ordering.
View this version →Extends tactical capture sequences with quiescence search.
Reduces horizon mistakes in volatile positions.
View this version →Caches repeated positions in a transposition table.
Avoids recalculating positions already visited by 0.6.
View this version →Combines iterative deepening with stronger move ordering.
Prioritizes promotions, captures, checks and the previous best move.
View this version →Evaluates mobility, pawn structure, king safety and development.
Adds a broader positional model to the 0.8 search.
View this version →Manages a move deadline and bounds quiescence search.
Returns the last completed depth instead of overrunning its clock.
View this version →Uses a Polyglot opening book before the 0.10 search.
Adds varied, established opening play; the book is disabled in the arena.
View this version →Scores mate distance and stores best moves in its search cache.
Prefers faster mates and improves transposition-table move ordering.
View this version →Runs 0.12's search on PeSTO's tapered piece-square tables.
Replaces the hand-written evaluation by two tuned tables per piece, blended by game phase. Cheaper leaves, deeper search.
View this version →Searches with the classic pruning and ordering techniques, each behind a switch.
Killer moves, history heuristic, null move pruning, principal variation search with aspiration windows, late move reductions, futility pruning and optional Syzygy tablebases on the 0.13 evaluation. Every technique can be switched off to measure it.
View this version →