What chess opening statistics can and can’t tell you

Opening statistics are some of the most-read numbers in chess, and some of the easiest to misread. A line that “scores 60%” sounds like a weapon, until you ask who played it, against whom and how many times. This guide explains what the opening numbers on StatChecked’s chess pages measure and how to get real value from your own report. StatChecked reads public games from chess.com and Lichess and is not affiliated with either.

What a score percentage means

Chess statistics use score rather than win rate. A win counts one point, a draw half a point and a loss nothing, and the total is divided by games played, so five wins, two draws and three losses from ten games is a 60% score. Two openings can share a 50% score with very different games behind them, one full of draws and the other all wins and losses. That is why our tables show a win/draw/loss bar next to every score.

Colour matters too. Because White moves first, it scores a little better than half in almost every large collection of games. A 52% score with White is unremarkable, and 48% with Black is no cause for alarm.

Popularity by rating band

The chess home page shows trending openings from games stored for players looked up on StatChecked in the last 90 days. Only rated games count, and each counts once, however many of its players were looked up. Each game is placed in a band by the average of the two players’ ratings: under 1200, 1200–1500, 1500–1800, 1800–2100, 2100–2400 and 2400+. The panel shows the band with the most games. For each opening family with at least five games in that band, it gives White’s score and the change in the opening’s share of games, comparing the last 30 days with the 60 days before, in percentage points.

Three caveats apply. These are the games of players people chose to look up here, not a random sample of everyone. All time controls are mixed together, from bullet to daily. And the ratings are chess.com ratings, which sit on a different scale from Lichess. A rising share usually reflects fashion, such as a popular streamer’s new favourite, more than a discovery about the opening’s strength.

Why scores shift with rating

The same opening can score very differently at different levels, for reasons that have little to do with the moves themselves.

Refutations. Traps and sharp gambits score well against players who haven’t met them, but higher up more opponents know the answer. Say a gambit scores 58% for the players who choose it in the 1200–1500 band; it could easily sit below 50% above 2100.
Draws. As ratings rise, more games are drawn, which pulls every score towards 50%.
Who plays it. An offbeat opening chosen mostly by enthusiasts who have studied it scores well partly because of those players. A beginner copying it shouldn’t expect the same results.
The clock. In bullet and fast blitz, many games are decided on time, so an opening that is easy to play quickly can outscore a sounder but more demanding one.
Mixed matchups. A band based on average rating can contain a 1300 playing a 1700, and the stronger player’s opening gets the credit for the win.

An opening statistic describes an opening in the hands of a particular group of players. It is not a verdict on the opening.

The small-sample trap

The rule of thumb from our guide to sample size works for chess too: divide 100 by the square root of the number of games to get a rough margin, in percentage points, either side of the score you see. Twelve games in the Caro-Kann at 40% carry a margin of about ±29 points, so your true score could be anywhere from about 11% to 69%. Draws make the real margin a little narrower, but not by enough to change the conclusion. Twelve games can’t tell you much.

Our tools build in minimums for this reason:

Openings tables hide openings below a minimum number of games (three by default, and adjustable). They sort within your 25 most-played openings, so a single lucky game can’t top the table.
Areas to improve flags an opening only after eight games, and only if it scores ten points below your overall score.
Repertoire marks a move as scoring badly only after five games under 40%.
Vs openings verdicts also wait for five games.

Treat every one of these as a prompt to look at the games, not a conclusion.

Score against expectation

Who you played matters as much as the result. The Elo formula gives an expected score for each game from the rating gap: against an opponent 100 points stronger you would expect about 36%, and against one 200 points stronger about 24%. The Vs openings tab compares your score in each group of games with those expectations added up. So 45% against much stronger players can be a good result, and 55% against much weaker ones a poor one.

The tab also shows a performance rating: the average opponent rating, plus 400 times wins minus losses, divided by games. Six wins, two draws and two losses against an average opponent of 1500 gives 1500 + 400 × 4 ÷ 10 = 1660. The “Against opponents rated…” table on the overview tab does the same comparison by rating gap.

Using your own repertoire report

A player report has three opening views:

Openings groups your games by opening family for each colour. It shows your score, your average and deepest theory depth, and your accuracy in games that were analysed on chess.com or Lichess. Theory depth is how many moves a game followed known theory, according to the Lichess opening book and chess.com’s own opening classification. The names describe the whole game, so a Sicilian among your White games is a choice your opponents made.
Repertoire is a move tree built from your most recent 100 games. From any position it shows what you, or your opponents, played next, how often, and how you scored. Its “Lines that cost you” list picks out your own move choices that scored under 40% over five or more games, from your second move up to move six.
Vs openings turns things round and scores you against each opening your opponents chose, judged against expectation.

A practical routine: