Arka Play benchmarks

Use symbolic-rule games to benchmark local model outputs.

Use symbolic-rule games to benchmark local model outputs:

pip install python-chess
arka play chess --moves e2e4 e7e5 g1f3 b8c6 --json

Arka validates every move against the chess rules, reports illegal moves and accuracy, and returns the final FEN. The same harness can be extended with other deterministic games; an illegal model action never changes the board.