Patzer

Patzer is a self hosted chess platform for households that want private game analysis, training, and local play. It can import public games from Chess.com and Lichess or accept pasted PGN files, then analyze each move with bundled Stockfish. Reviews include move classifications, accuracy scores, evaluation graphs, engine lines, opening detection, and explanations of tactical mistakes.

You can play against seven Stockfish levels or challenge another profile on the same server. Multiple users, household ratings, kid mode, interactive lessons, an opening trainer, and five interface languages are included. Everything runs in one Docker container with game data stored in a local SQLite database.

The optional AI coach connects to an Ollama or vLLM server on your network. Stockfish and chess.js calculate the position, candidate moves, and tactical facts before the language model writes an explanation. This keeps chess analysis grounded and means the main review features still work without an LLM.

Patzer assumes a trusted home network. It lacks rate limiting for analysis requests and a normal password reset workflow, so remote access should use a VPN or properly secured reverse proxy. Chess.com and Lichess imports, opening statistics, and update checks contact external services. Optional DeepSeek and hosted engine integrations send relevant data to those providers.

Patzer replaces much of the paid game review and private training workflow, but it does not provide Chess.com style public matchmaking, tournaments, social features, or video lessons. Its own code is MIT licensed, while bundled Stockfish and Chessground components use GPL licenses.