Crazyhouse Trainer

A personal crazyhouse training app, built from an analysis of my own Lichess crazyhouse games. It runs entirely in the browser and works offline once installed to the home screen.

What is in it

• Puzzles — positions taken from my own games, each verified by Fairy-Stockfish at depth 16 with a MultiPV search so exactly one move is clearly best. They are the moves I actually got wrong, plus the opponent mistakes I failed to punish. Nothing from the opening: every position is from move 8 onwards. Untimed, with spaced repetition so positions come back days later.
• Openings — guided lessons built from recent games against strong human and AI opponents. Choose a player to start from openings seen in their games. The board asks for each course move, offers hints and explains the engine line and the reply to watch. An explorer lets me compare the prepared engine alternatives from each position.
• Collapses — the hard one. A mate you allowed one move earlier teaches nothing, so for each game lost by checkmate this finds the earliest moment where the position was still holdable and my move made it lost, and asks for the move that would have held. Those moments sit 3 to 12 moves before the mate.
• Plan — the analysis and the training plan it produced.

Install on iPhone

Open the site in Safari, tap the Share button, then Add to Home Screen. It then launches full screen and works without a network connection.

Development

npm install
npm run dev      # local dev server
npm run build    # production build into dist/

Regenerating the training data

The data files in src/data/ are produced by the scripts in the sibling zh-analysis/ directory from a Lichess PGN export:

cd ../zh-analysis
./refresh.sh ~/Downloads/lichess_<user>_<date>.pgn
cd ../zh-trainer && npm run build

That pipeline parses the PGN into SQLite, scans the games with Fairy-Stockfish to find recurring mistakes, verifies candidate puzzles with a MultiPV search, builds the opening explorer, and regenerates the report.

The opening course can be refreshed independently from the latest PGN:

cd ../zh-trainer
../zh-analysis/.venv/bin/python scripts/make_opening_course.py <PGN>
npm run build