gambit is a from-scratch alpha-beta chess engine. It uses a 0x88 mailbox board representation, Zobrist hashing for transposition detection, and a principal-variation search with a full set of modern pruning and move-ordering heuristics. It speaks the UCI protocol and can also read and write SAN, PGN and EPD.

Features

Board — 0x88 mailbox representation with incremental Zobrist hashing.
Move generation — pseudo-legal and legal generation for all piece types, including castling, en passant and promotions, validated by a perft node counter.
Evaluation — tapered midgame/endgame evaluation with piece-square tables, pawn structure (passed, doubled, isolated), mobility, the bishop pair, rook file bonuses and king safety, plus a static exchange evaluation (SEE).
Search — iterative-deepening PVS with quiescence search, a transposition table, null-move pruning, reverse futility and razoring, late-move reductions, late-move pruning, killer moves and a history heuristic, with aspiration windows at the root.
Time management — UCI go limit parsing (clock, increment, movestogo, movetime, depth, infinite) with a per-move budget.
Notation — SAN formatting/parsing, PGN reading/writing, and EPD parsing with a test-suite runner (bundled Win at Chess positions).
Self-play — a game loop with draw adjudication (fifty-move, insufficient material, threefold repetition).

Usage

# UCI mode
scala-cli run . -- uci

# self-play a game (maxPlies, depth)
scala-cli run . -- selfplay 200 6

# perft / search on a position
scala-cli run . -- <fen|startpos> <depth>

# run the EPD (Win at Chess) suite, N ms per position
scala-cli run . -- epd 500

Running with no arguments prints usage and a short perft/search demo.

Tests

scala-cli test .

Tests use munit and cover move generation and perft, make/unmake and hashing, evaluation and SEE, search heuristics, the transposition table, draw detection, SAN/PGN notation, time control and the UCI handler.

Requirements

scala-cli (Scala 3.8.4, as pinned in project.scala)