boardviz

Analyze your chess.com history for recurring mistakes — especially the "big think → blunder" pattern that bites when you're winning — and drill those positions in an interactive, timed trainer that scores each move on a +2..−2 scale.

What it does

• Import your games from chess.com (last N games, walking monthly archives).
• Filter by color, time control, and outcome.
• Review engine-confirmed mistakes and cluster the recurring ones (by pawn structure, move type, phase, opening).
• Big-think analytic — tests whether long thinks lead to more mistakes, split by game state (winning / equal / losing).
• Trainer — replays your mistake positions; you make a move and get a +2..−2 grade, with a time penalty (e.g. blitz −1 after 10s, −2 after 20s).
• Scout any opponent's history, and grade both sides of your own games.

Setup

uv sync

Stockfish engine (required)

The analysis needs a Stockfish binary. It is not committed (GPL-3.0; the app itself is MIT and drives the engine arm's-length over UCI). Provide one of:

• Place stockfish.exe in engines/ (this repo already vendors it locally), or
• Set STOCKFISH_PATH to your own binary.

Download official builds from https://stockfishchess.org/download/ (Windows AVX2). The vendored engine's license is in engines/Stockfish-LICENSE.txt.

Usage

uv run boardviz fetch --user hucker233 --n 100     # download last 100 games
uv run boardviz analyze --user hucker233           # engine analysis (cached)
uv run streamlit run src/boardviz/app.py           # launch the app

Data (raw archives + the SQLite DB) lives in data/ and is gitignored.