There is a habit in reporting that treats a small number as nothing happened. The test came back flat, so the test is written up as a failure, filed, and forgotten. The large differences get the write-up and the small ones get deleted.
That is backwards when the small number sits next to a large one measured on the same thing. Tennis has a clean example, because the sport is played on three surfaces and keeps the same records on all of them.
What the surface changes
Grass, clay and hard courts play differently, and the record shows it plainly.
Players hold serve 74.43% of the time on grass and 65.98% on clay, with hard courts between at 69.95%. That is a spread of 8.45 percentage points on the most repeated event in the sport.
Tiebreaks follow. 15.91% of sets on grass reach one, against 10.72% on clay. Counted per match instead of per set, 31.73% of grass matches contain a tiebreak and 22.52% of clay matches do.
So the surface changes how points are won, how often serve survives, and how often a set runs to its limit. Nobody argues about that.
What the surface does not change
Now take the outcome. Across 116,382 completed matches, the player who loses the opening set goes on to win 16.52% of the time. Split by surface:
Grass: 17.20%, from 3,593 matches.
Clay: 16.79%, from 47,944 matches.
Hard: 16.47%, from 61,751 matches.
A spread of 0.73 points, against 8.45 on hold rate and 9.21 on tiebreak frequency. The conditions move everything about the texture of a match and almost nothing about who wins it.
Why the small number is the useful one
If the comeback rate had moved as much as the hold rate, the conclusion would be ordinary. Conditions matter, so results differ by conditions. Instead the two numbers disagree, and the disagreement is the information.
It says the match outcome is carried by the gap between the two players, not by the court they are standing on. A surface that helps the server helps both servers. Whatever advantage it creates, it creates at both ends, and it cancels.
That is a real finding about where the variance lives, and it is only visible because somebody published a number that barely moved.
What to do with a flat result
Three things, and none of them involve deleting the row.
Report the small difference with its neighbours. 0.73 means nothing alone. Next to 8.45 and 9.21 from the same matches, it says the effect is confined to one layer.
Ask whether the change applies to both sides. Most conditions do. A change that helps everyone equally shifts the texture of an outcome without shifting the outcome.
Treat agreement between groups as evidence. Grass has 3,593 matches against 61,751 on hard. Very different sample sizes, landing within three quarters of a point of each other. That agreement is worth more than either figure on its own.
Where the figures come from
All of the above comes from the Live Tennis API, a JSON service for developers. It covers live scores and fixtures, with players, match events and point-by-point history behind them. The figures are published openly at blog.livetennisapi.com/studies.json under CC BY 4.0, free to reuse with attribution.
Method: completed best-of-three professional singles matches, 1 January 2023 to 14 September 2026. Retirements and walkovers excluded. Tiebreak games excluded from hold rates. Surface breakdowns cover the matches whose surface the feed stated. Counted from the point-by-point record.
Also Read
