Last updated:
Chess.com's Game Rating is an estimate of the strength you played at in one game. According to the chess.com help center, it "is determined by comparing the quality of your moves to what is expected from a player at your rating level." Because your own rating is part of the comparison, and because one game is a small sample, the number can land hundreds of points above or below your Elo.
What Game Rating measures and what it does not
Game Rating looks at the quality of your moves in one game and expresses it on a rating scale. It is a description of that game, not a new rating. It does not change your chess.com rating, it is not a performance rating calculated from results, and it says nothing about how you would do over a whole tournament.
It also inherits every limit of accuracy. It measures how much winning chance your moves gave away, so a quiet, one-sided game and a sharp, double-edged game of the same real quality can produce very different numbers. For how accuracy itself is computed, see How Is Chess Accuracy Calculated?
Why the same game can get a different Game Rating
The help center definition compares your moves "to what is expected from a player at your rating level". Your rating is therefore one of the inputs. Two players who make moves of the same quality but have different ratings are not compared against the same expectation, and players regularly report different Game Ratings for similar accuracy. Chess.com does not publish the formula, so the exact weight of your rating is unknown.
The practical consequence: Game Rating is useful for comparing your own games with each other. It is much less useful for comparing yourself with another player, or for guessing your "true" rating from one game.
Why a single game is a noisy estimate
- Game length. A 20-move game with many book moves gives the estimate very little to work with. A long game gives more, but one late blunder can still sink it.
- Forced lines. Recaptures and only moves are easy to play accurately, so a game full of them looks stronger than it was.
- Opponent strength. A weak opponent hands you clearly winning positions, where almost every move scores well. A strong opponent keeps the position tense, where small errors cost more.
- Time control. Bullet and blitz games contain more mistakes made under time pressure than rapid or classical games at the same rating.
Accuracy itself swings a lot from game to game. In GM Larry Kaufman's study of chess.com rapid games, even with about 50 competitive games per player, predicting the ratings of top players from accuracy still had a median error of 48.5 Elo. A single game is far noisier than that. For the averages behind these numbers, see Chess Accuracy by Rating.
| Number | What it is based on | How stable it is |
|---|---|---|
| Your chess.com rating | Results of all your rated games | Stable, moves a few points per game |
| Game Rating | Move quality in one game, compared with your rating level | Very noisy, can move hundreds of points |
| Average over 20 to 50 games | Move quality across many games | Much more stable |
How ChessRank estimates game strength
ChessRank's Game Analysis also shows a rating estimate for each side of a game. This is how it works, as written in the code:
- Accuracy first. The game is analyzed with Stockfish in your browser and scored as described in How ChessRank analyzes your games.
- A ladder from accuracy to rating. The side's accuracy is mapped to a rating through a ladder of points fitted against the Game Ratings chess.com gave to real games.
- A pull toward your own rating. If the game's PGN contains the player's rating, the estimate is blended with it, and the pull is stronger the higher that rating is. If the PGN has no rating, the estimate comes from accuracy alone.
- Bots keep their rating. In games against the Play Bots, the bot's side is held to the rating the bot is advertised with.
- Range. The result is kept between 100 and 3500.
Known limits: the inputs are your accuracy and your rating, nothing else. Any error in the accuracy carries straight into the estimate, and a single game stays a single game. The number is calibrated to land near chess.com's Game Rating, but it is not the same system and will not always match it.
How to get a reliable estimate of your level
Average over many games. Twenty games is a reasonable minimum and fifty is better, which is the order of magnitude Kaufman worked with. Keep the time control fixed, because accuracy in blitz and in rapid is not on the same footing, and look at the trend rather than at any single value.
The Report does this for you: it analyzes your recent chess.com or lichess games in one batch and shows your average accuracy and results per time control. Chess Stats tracks how they change over time. Both are free, with no account needed and no daily limit.
ChessRank is an independent project and is not affiliated with, endorsed by or sponsored by Chess.com or Lichess. All trademarks belong to their respective owners.
Frequently asked questions
What does Game Rating mean on chess.com?
It is an estimate of the strength you played at in one game. The chess.com help center says it is determined by comparing the quality of your moves to what is expected from a player at your rating level. It does not change your actual rating.
Why is my Game Rating so low?
Usually because of one or two large mistakes, a sharp game where small errors cost a lot, or time pressure. One game is a small sample, so a low Game Rating after a bad game says little about your level. Look at the average over many games instead.
Why is my Game Rating higher than my Elo?
Short games, forced lines and opponents who give you a clearly winning position all make accurate play easier. A single game can therefore land well above your rating. If it keeps happening over many games, your rating is likely to rise.
Is Game Rating the same as a performance rating?
No. A performance rating is calculated from results against opponents of known strength. Game Rating is based on the quality of your moves in one game, compared with what is expected at your rating level.
How does ChessRank estimate the strength of a game?
It maps the side's accuracy to a rating through a ladder fitted against chess.com Game Ratings of real games, then blends it with the player's rating from the PGN when one is present. The pull toward that rating is stronger at higher ratings, and the result stays between 100 and 3500.
How many games do I need to estimate my level?
At least 20 games in the same time control, and 50 is better. In Kaufman's study, about 50 competitive games per player still left a median error of 48.5 Elo for top players. The Report on ChessRank averages your recent games for you.
Sources
The Game Rating definition comes from the chess.com help center. The figures on sample size come from GM Larry Kaufman, Accuracy And Ratings On Chess.com, chess.com blog, August 2023. Facts about ChessRank come from its own code.