How Chessbook works

Repertoire stats ​

Chessbook grades repertoires on three key stats because "is this repertoire good?" doesn't really have a single answer. Building a repertoire is always a tradeoff between these stats.

Effectiveness ​

Effectiveness estimates how well your repertoire should score in games at your level.

Effectiveness is measured by win rate according to the Lichess database. We follow your saved lines and look at the expected result of the positions you reach using games from players in your rating range. Those positions are then weighted by how often opponents actually reach them against your repertoire, so you get rewarded more for high-scoring positions that are also likely to happen.

Soundness ​

Soundness measures the objective strength of your repertoire according to Stockfish.

For each path through your repertoire, we measure how much your position degrades according to the engine. This considers the whole line rather than grading each move in isolation and gives extra weight to risky critical paths.

Learnability ​

Learnability estimates how easy your repertoire will be to remember.

The Learnability of any given move is predicted using an AI model trained on how often other Chessbook users got similar moves wrong in practice sessions. The model looks at things like:

  • How common the position is
  • How often people at your level play the move
  • Whether you play the same move in similar positions in your repertoire
  • How deep you are in the line

FAQ ​

How does the 1 to 99 scale work? ​

For Effectiveness and Learnability, the ends of the scale are set as follows:

  • A score of 1 represents a bad but plausible repertoire for your rating, side, and coverage goal. We use the Lichess game database to simulate these real repertoires.
  • A score of 99 represents the best repertoire Chessbook's auto-fill can construct when it focuses entirely on that one stat, for your rating, side, and coverage goal.

Unlike Effectiveness and Learnability, Soundness is graded on an absolute scale that means the same thing at every rating level: a path concession around 140 centipawns (roughly "playing a very dubious gambit everywhere") maps to a score of 1, and keeping the concession under about 10 centipawns maps to 99.

Are the scores percentiles? ​

No, they describe where you sit on a useful, achievable range. They're not a comparison with other users. A score of 50 means you're halfway between a bad-but-plausible repertoire and the best build we know how to make.

Will my coverage goal affect my stats? ​

Generally yes. Preparing more opponent responses adds useful coverage, but it can also require harder-to-remember lines or expose tradeoffs that didn't appear in a smaller repertoire.

There are also progress stats like completion and Mastery, which are more about how much of the repertoire you've built and learned.

One of my scores is low, is that bad? ​

Not necessarily. Building a repertoire is all about tradeoffs, and sometimes giving up points in one category helps you increase another one.

Low Effectiveness ​

High Effectiveness is only valuable if you actually learn the moves. If you don't have a lot of time or motivation to learn hundreds of unique lines, trading off raw win rate for Learnability might actually improve your results.

Low Soundness ​

Many gambits deliberately concede objective evaluation in exchange for practical chances. Most of your opponents aren't going to play like Stockfish.

Low Learnability ​

Sharp, tactical or ambitious openings can all be harder to remember. Many high-level players are prepared to put in extra work to play the objectively best lines.

Why did my stats change all of a sudden? ​

We periodically update the way these stats are calculated as we discover more accurate ways to do it. We also regularly import new games from Lichess, which can also affect the stats.