Comparing Players Through Common Opponents
Most pairs of club players have played each other once or never, so their head-to-head record says very little. What they usually do share is a pool of opponents: the same club regulars, the same local tournaments. Comparing how each scored against those shared opponents is a natural indirect comparison. USCF Compare Players shows it for every pair, and the Chess.com and Lichess comparisons do the same with online games. For US Chess Regular and Quick, though, the headline estimate comes from the two players' ratings, calibrated and backtested: why, below.
Three views, from most to least direct
- Head-to-head — the games the two have played against each other. The most direct evidence, and usually the thinnest.
- Common opponents — each player's results against the people both have played.
- Side by side — each player's own ratings and record, with no shared reference point.
US Chess comparisons are made within one rating system at a time. Regular includes dual-rated events, Quick does too, and Blitz and the online systems are kept separate. Mixing a player's Quick results with another's Regular ones would compare different things.
Performance against the shared field
In USCF Compare Players, for each player the comparison takes the average rating of the shared opponents they played — counting an opponent met twice as two games — and adds a figure from FIDE's performance table for their percentage score. The result is a performance rating: roughly, the rating they played like against that field.
| Shared opponent | Rating | Player A | Player B |
|---|---|---|---|
| Opponent 1 | 1700 | 1 win, 1 draw | 1 loss |
| Opponent 2 | 1550 | 1 win | 2 wins |
| Opponent 3 | 1600 | 1 draw | 1 win, 1 draw |
- Player A: 3 from 4 games (75%) against an average of 1638. Performance 1638 + 193 = 1831.
- Player B: 3.5 from 5 games (70%) against an average of 1600. Performance 1600 + 149 = 1749.
A performed 82 points better against the shared field. Fed back into the rating curve, that gap would give A an expected score of about 0.62 in a game between them. USCF Compare Players shows this under its headline estimate, as information, so you can see when results against the shared field and the ratings disagree.
How much to trust it
A performance rating is only as steady as the number of games behind it. Here is how far it moves if a player who had scored exactly 50% wins one more of their shared games:
| Games against shared opponents | One more win moves the performance by |
|---|---|
| 1 | 800 points |
| 2 | 193 points |
| 3 | 122 points |
| 4 | 91 points |
| 6 | 60 points |
| 10 | 36 points |
| 20 | 18 points |
The example above rests on four and five games per player, so its 82-point gap could easily reverse with one more result. Check the games column before believing a comparison, and treat one built on a handful of shared games as a hint about recent form rather than a prediction.
Why the USCF estimate uses ratings instead
The common-opponent figure was tested as a predictor on about 21,000 Regular and Quick results from 2026 (about 13,800 games; a dual-rated game counts in both), games it had never seen, almost all in Texas. Plain ratings turned out to be overconfident: between two established players 200 points apart, the favourite scores about 70%, where the rating formula says 76%. The common-opponent figure was even more overconfident: in its most confident tenth of games it gave the favourite 98–99%, and the favourites scored 70–73%.
So the headline in USCF Compare Players now comes from a model built on the two live ratings. It shrinks the rating gap to what results actually show, and adds each player's recent rating trend and a small allowance for any games between them. On those same 2026 games it was closer to the real results than plain ratings, which in turn beat the common-opponent figure. Blitz and the online systems were not tested, so for them the headline is plain ratings. The common-opponent results are still shown for every system, as information.
Things that skew a comparison
- Time. Results from two years ago describe a different player, especially a junior. Shared opponents met in different seasons are a weaker reference.
- Opponents' ratings move. The calculation uses each opponent's current rating, not their rating when the game was played.
- Different pools. A US Chess rating, a FIDE rating, a Chess.com rating and a Lichess rating are four separate scales. Two players who both hold "1800" in different systems have not been shown to be equal.
To compare two players, open USCF Compare Players, or look either player up on the Chess.com or Lichess tab and press Compare Players.