Chess Rating (ELO) Estimator

Enter your current rating, your federation, how many rated games you have played, and then each opponent's rating with the result.

Take a 1,550-rated USCF player with 120 rated games in a five-round event: win against 1610, loss to 1720, win against 1480, draw with 1590, win against 1655.

The K-factor is 32, USCF's figure below 2100. Expected scores game by game are 0.415, 0.273, 0.599, 0.443 and 0.353, totalling 2.083 out of 5, or 41.7%. The actual score was 3.5 out of 5, or 70%.

Rating change is 32 × (3.5 − 2.083) = +45.3 points, giving a new rating of 1,595.3.

The per-game breakdown is where the system becomes intuitive. Beating the 1655 was worth +20.7 because it was the least expected win. Beating the 1480 was worth only +12.8, because that was expected to happen most of the time. Losing to the 1720 cost just −8.7, since that loss was largely priced in. And the draw against the near-equal 1590 was worth +1.8, because a draw against someone slightly stronger is a marginally above-expectation result.

This is the entire logic of the rating system: you are paid for surprising it. The board game Elo calculator applies the same core formula to any two-player game with a K-factor you choose yourself.

The performance rating answers a different question from the rating change. It asks: what rating would produce this result as an expected outcome? The linear approximation is average opponent + 400 × (wins − losses) ÷ games.

For the tournament above, the average opponent was 1,611 and the player scored 3 wins and 1 loss in 5 games, giving 1611 + 400 × 2 ÷ 5 = 1,771. The player rated 1550 performed like a 1771, 221 points above their rating, yet gained only 45 points. That gap is deliberate. Ratings move slowly because a single tournament is a small sample, and K is the dial controlling how much weight one event carries.

Now a stronger player under FIDE rules. A 2,050-rated player with 85 rated games: draw with 2210, win against 1980, loss to 2140. K is 20, FIDE's figure for an established player who has not reached 2400. Under FIDE's own probability table the expected scores are 0.29, 0.60 and 0.38, 1.27 in total, against an actual 1.5, giving +5 points after FIDE's rounding and a new rating of 2,055. The average opponent was 2,110 and the performance rating 2,110 exactly, since one win and one loss cancel.

Note how little that rating moved. The same over-performance under a K of 32 would have produced about +7.4 rather than +5. And a provisional player (1,200 rated with only 8 games) carries FIDE's K of 40, and FIDE's table expects a 200-point underdog to score 0.24, so a single win against a 1400 is worth +30 points. New players move fast by design, so their rating converges quickly toward their real strength.

One further FIDE rule worth knowing: once a player has ever reached 2400, K stays at 10 permanently, even if the rating later falls. A 2,380-rated player with a peak of 2,410 still carries K = 10.

This is a chess-specific tool, and the difference from a general Elo calculator matters. The board game Elo calculator computes a single game with a K-factor you supply, which is right for a club ladder or a tabletop league where you set your own rules. This page applies the published federation rules to a whole event, chooses K for you, and adds performance rating, the things chess players actually need and generic Elo tools do not provide.

What it estimates rather than reproduces exactly is the official calculation. FIDE's performance rating uses a non-linear table of rating differences rather than the linear 400-point approximation used here, and the two diverge at the extremes: a perfect score has no defined linear performance rating at all, since 400 × 1 understates a result that could imply any rating. For scores between about 10% and 90% the linear version is close.

Real federations also apply rules that no formula captures. Rating floors stop a rating falling below a level once achieved. Bonus points are awarded in some systems for exceptional performances. Provisional formulas differ from the established ones in ways beyond the K-factor. Games are rounded at different points in the calculation, and unrated opponents are handled by special rules. Online platforms use Glicko rather than Elo entirely, which adds a rating-deviation term so uncertain ratings move faster than confident ones, a mathematically better system that this page does not model.

For rating changes that actually count, the federation's own published figure is the authority. This estimate tells you what to expect, usually within a point or two.

The performance rating uses the linear approximation and will be wrong at the extremes. At a perfect or zero score it returns a figure that is not meaningful. FIDE's own table caps the implied difference rather than extending it indefinitely. Between roughly 10% and 90% scores it tracks the official table closely.

Rating floors, bonus points, provisional-rating formulas, unrated-opponent handling and the specific rounding rules each federation applies are all outside the model. So are the differences between rapid, blitz and classical rating pools, which are calculated separately with their own K-factors and their own numbers. Feeding a blitz result into a classical rating produces a meaningless answer.

The Elo system itself assumes performance is normally distributed around a player's true strength and that strength changes slowly. Neither assumption holds well for rapidly improving juniors, players returning after long breaks, or anyone whose rating is stale. Glicko-based systems, used by most online platforms, address exactly this by tracking how uncertain each rating is and moving uncertain ratings faster.

Finally, all of this assumes both ratings are from the same pool. Ratings from different federations are not directly comparable. The same player can carry noticeably different FIDE and national ratings, and online ratings are a separate universe again. Comparing across pools, or entering an online rating into a federation calculation, will produce a number with no meaning.

The Board Game Elo Rating Calculator computes a single game with a K-factor you set yourself, which is the right tool for a club ladder, a tabletop league or any two-player game outside chess. The Dungeons & Dragons Ability Score Probability Calculator covers a different corner of the same probability-in-games territory, and the Bowling Handicap Calculator handles the other classic approach to rating players of unequal strength: equalising the result rather than predicting it.