Repertoire Gap Finder
Which openings should you work on next?
How this works #
The Gap Finder builds your opening tree out of your own games, then asks a reference database how each of those positions really scores among players around your rating. Where the two disagree by enough to be worth trusting, that is a gap.
This page is being built in the open. It reads your games, builds the tree, looks up the positions you reach most often, and ranks what it finds. Still to come: a board for stepping through a line, and fuller notes under each section on how that part is judged.
What it reads #
Your games. Public Chess.com history, through the same caching service the Tilt Clock uses. Nothing is stored against you and there is nothing to sign in to.
The reference. The Lichess opening explorer, which holds hundreds of millions of games grouped by rating band and time control. It is the closest thing there is to a free answer to “what do players at my level actually do here”.
Positions, not move orders #
Every position in the tree is keyed by the position itself, not by the moves that reached it. So 1.d4 d5 2.Nf3 and 1.Nf3 d5 2.d4 count as the same node, with their games added together, because they are the same problem to study. Counting them separately would halve the evidence for each and could hide a gap that is really there.
Which positions get looked up #
Three rules, and they are stated here because they decide what the tool can and cannot see.
Within the first 16 plies. That is eight moves each. Past it, the position stops being theory the reference can help with and starts being a middlegame.
Reached at least 10 times. A position you have played three times carries no evidence, whatever the reference says about it.
Capped at 60 lookups per run, spent on the positions you reach most often. The reference database asks that it not be hammered, so the lookups go one at a time, and sixty is about as long a wait as is reasonable on a first run. What gets cut is by definition the part of your repertoire you meet least, so it is where a finding would have mattered least. The page says how many positions qualified and how many were looked at, so nothing is hidden.
Once a position has been looked up, it is cached for a month and shared with everyone else who reaches it. Common openings are therefore nearly instant, and a second run of your own is much faster than the first.
Rating bands are Lichess ratings #
The reference groups its games by rating band, and those bands are Lichess ratings. Your Chess.com rating is on a different scale, and for most club players it reads lower than their Lichess equivalent.
No conversion is applied, because any single number for that would be a guess dressed up as a measurement. Your Chess.com rating is matched to the nearest band as-is, the page labels it an estimate, and you can change it.
Colours are separate #
A repertoire is per colour. White and Black are analysed separately and never mixed, because the same position means opposite things depending on which side of it you sit.