How personal chess-history analysis works
This method summarizes observed decisions and outcomes from one player’s imported games. It is descriptive, not an engine evaluation: it can identify repeated correlations and practice positions, but it cannot prove that a move caused a result.
Use real games
Check whether your history has enough signal
Enter a public username. We read up to 80 recent standard games and show a factual sample before you create an account.
Data and scoring
The importer reads public standard-chess histories from Chess.com or Lichess. A game is recorded with colour, result, time class, opening when supplied, date and PGN move text. Wins score 1, draws 0.5 and losses 0.
Reports compare subsets with the player’s own baseline. This avoids comparing a beginner with a titled player, but changing rating, opponent strength and small samples can still affect the result.
- No private platform credentials
- No live-game assistance
- No engine evaluation
- No claim that correlation is causation
Thresholds and reproducibility
Broad DNA findings require at least five games in a split. Repeated-position training uses higher per-position and per-move floors so a single game does not define a habit. The public preview is capped at 80 games; a saved import can read up to 200.
The same imported corpus and settings should produce the same result. New games can change the baseline, counts and ranking of findings, which is why the product refreshes linked accounts periodically.
Frequently asked questions
Why not call the result a blunder?
Blunder is an objective engine label. This method observes that one repeated choice scored worse than another in the player’s own games.
Can opponent strength affect the result?
Yes. Opponent strength, changing player skill and opening selection can confound observed scores, so findings are prompts for review rather than causal conclusions.
Are variants included?
No. The import is limited to standard chess games and supported common time controls.