The developer integrated the Maia chess model into the ChessNavio browser extension to provide human-like move suggestions. Unlike the Stockfish engine, which calculates the objectively best move, Maia predicts moves based on human play patterns across different rating ranges.
The system uses a hybrid approach: Maia identifies candidate moves, while Stockfish acts as a filter to prevent catastrophic blunders. If Maia's top suggestion leads to a significant loss, the system evaluates subsequent candidates or reverts to Stockfish's analysis if no human-like move is safe.
The extension runs locally using ONNX models, ensuring privacy and offline functionality. It provides users with both the engine's objective evaluation and a label indicating how typical a move is for a specific skill level, helping players understand their decision-making process.
The developer notes that while the tool helps users identify potential errors before they occur, it is intended for training and post-game analysis rather than live play. The project utilizes the open-source Maia platform under the GPLv3 license.
