Find your five biggest opening leaks

How this works

Your games are fetched by your own browser, directly from Lichess's public API. Nothing is ever sent to Repertoire Builder's servers, because this site has no servers that could receive it. Your report is stored only in your browser's local storage, and only if you choose to save it. A "leak" here means: at a position you've reached at least three times, the move you usually play scores measurably worse, across hundreds of thousands of games at your rating band, than the move your band's own data recommends. Your own games only tell us how often you reach a position and which move you pick. The band data, not your own small sample, tells us what that move actually scores. Example: a player who reaches the position after 1.e4 e5 2.Nf3 Nc6 and answers 3.Bc4 twenty-five times might be shown that 3.Bb5 scores several points per 100 games better among their rating band, worth roughly a full point of rating over a season of play at that frequency.

We fetch up to 300 of your most recent rated blitz and rapid games (about 15 seconds of streaming, per Lichess's own published rate limit for this endpoint) and look only at the first 24 half-moves of each, since we're identifying openings, not analysing whole games. Coverage is bounded: our band dataset only reaches as deep as real games commonly go before branching out, so some of your games will have no comparison available. That's shown honestly, not hidden.