MindInChess — Chess Analysis App

Problem Statement

Analyzing chess games after playing them is essential for improvement, but existing tools either require paid subscriptions (Chess.com analysis) or have complex interfaces (Lichess analysis board). I wanted a simple, free tool that gives immediate insights into game quality.

Demo

Upload a PGN file or import directly from Chess.com/Lichess. The app analyzes every move against Stockfish and provides:

Accuracy score — percentage of moves that match or are close to the engine's top choice
Blunder detection — moves that lose significant evaluation (>1.5 pawns)
Color-coded moves — green (good), yellow (inaccuracy), orange (mistake), red (blunder)
Critical moments — positions where the game's outcome shifted

Architecture

Loading diagram...

Each move is evaluated by running Stockfish at depth 18 on the position before and after the move. The evaluation difference determines the classification:

< 0.3 pawns loss → good move (green)
0.3 - 0.8 → inaccuracy (yellow)
0.8 - 1.5 → mistake (orange)
> 1.5 → blunder (red)

Tech Stack

Python / Flask — Backend API for PGN parsing and Stockfish integration
python-chess — PGN parsing, move validation, FEN generation
Stockfish — Chess engine for position evaluation (runs as subprocess)
React — Frontend with interactive chessboard
Chart.js — Evaluation graph showing advantage over time

Lessons Learned

Stockfish depth matters for accuracy — Depth 12 is fast but misses tactical nuances. Depth 18 takes ~0.5s per move but gives reliable evaluations. Depth 20+ is diminishing returns for amateur game analysis.
Chess.com and Lichess have different PGN formats — Chess.com includes clock times, Lichess includes evaluation annotations. The parser needs to handle both gracefully.
Mobile-responsive chessboards are tricky — The board needs to be square and resize with the viewport. CSS aspect-ratio: 1 plus max-width: min(100vw, 560px) solved it.

Demo

Upload a PGN file or import directly from Chess.com/Lichess. The app analyzes every move against Stockfish and provides:

Accuracy score — percentage of moves that match or are close to the engine's top choice

Blunder detection — moves that lose significant evaluation (>1.5 pawns)

Color-coded moves — green (good), yellow (inaccuracy), orange (mistake), red (blunder)

Critical moments — positions where the game's outcome shifted

Architecture

Loading diagram...

Each move is evaluated by running Stockfish at depth 18 on the position before and after the move. The evaluation difference determines the classification:

< 0.3 pawns loss → good move (green)

0.3 - 0.8 → inaccuracy (yellow)

0.8 - 1.5 → mistake (orange)

> 1.5 → blunder (red)

Tech Stack

Python / Flask — Backend API for PGN parsing and Stockfish integration

python-chess — PGN parsing, move validation, FEN generation

Stockfish — Chess engine for position evaluation (runs as subprocess)

React — Frontend with interactive chessboard

Chart.js — Evaluation graph showing advantage over time

Lessons Learned

Stockfish depth matters for accuracy — Depth 12 is fast but misses tactical nuances. Depth 18 takes ~0.5s per move but gives reliable evaluations. Depth 20+ is diminishing returns for amateur game analysis.

Chess.com and Lichess have different PGN formats — Chess.com includes clock times, Lichess includes evaluation annotations. The parser needs to handle both gracefully.

Mobile-responsive chessboards are tricky — The board needs to be square and resize with the viewport. CSS aspect-ratio: 1 plus max-width: min(100vw, 560px) solved it.