Chess engine

Run a visual match between two AI configurations:

cargo run --release --bin engine_match

Run a batch of 100 games without opening a window:

cargo run --release --bin engine_match --features benchmark

The benchmark alternates colours each game, so each engine plays 50 games as White and 50 as Black. It prints each engine's running win count and shared draws after every game, then each engine's total wins, draws, and losses.

Edit ENGINE_A and ENGINE_B in src/bin/engine_match.rs to compare other searchers, evaluators, or depths. Update each configuration's name along with its factory. Both modes use these factories; the defaults are negamax with BoardEvaluation at depths 2 and 5. Each benchmark game starts from the standard position with fresh engines and empty search caches. Random tie-breaking is preserved, so results can vary between runs.

Each run creates a uniquely named CSV under benchmark-results/ in the current directory. It records colour assignments, the winning engine's name (or draw), ending reasons, plies, elapsed milliseconds, and cumulative engine_a_wins, draws, and engine_b_wins. Engine A and B retain their identities when colours swap. The final row contains the overall engine score. Results are flushed after every completed game, and previous files are preserved.

Checkmate awards the win to the opposite colour from the side to move. Stalemate, threefold repetition, and the fifty-move rule are draws. A game still running after 400 plies (200 full moves) is recorded as a draw with the distinct reason move_limit; natural endings on the final ply take precedence.

The benchmark feature skips UI startup and retains the existing SDL build dependencies.

Run the tests for both modes:

cargo test --all-targets
cargo test --all-targets --features benchmark

Ordinary tests do not run the full 100-game benchmark.