clarkerubber/irwin

Analyzes player games against Stockfish evaluations using TensorFlow neural networks trained on known cheating patterns, then reports suspected cheaters to Lichess moderators. The modular architecture separates core data handling (MongoDB serialization, database interfaces) from the assessment engine, allowing alternative detection approaches. Integrates with the Lichess API for player data retrieval and uses Stockfish engine analysis configured via JSON config for position evaluation.

524 stars. No commits in the last 6 months.