The most quoted number in the table has the least evidence behind it

Comparisons get made between the numbers that are easiest to find. The headline figure for each group goes into the table and the groups get ranked. How much evidence sits behind each row is left out of the discussion.

Professional tennis shows how far apart those two things can drift, because its service records are published with the counts attached.

Table of Contents

Six groups, one measure

Tennis is organised in tiers. The ATP and WTA main tours sit at the top, the Challenger circuits below them, and the ITF circuits below that. How often a player holds serve is recorded on all six.

ATP: 78.62%, from 230,715 service games.

Challenger men: 74.10%, from 622,018.

ITF men: 70.26%, from 529,948.

WTA: 64.36%, from 266,503.

Challenger women: 60.40%, from 92,396.

ITF women: 58.93%, from 553,157.

Those six add to 2,294,737 service games. Read the rates alone and the ranking is clean, with a 19.69 point spread from top to bottom.

Now read the counts

The ATP figure is the one that gets quoted. It is the tour people watch, the number that appears in commentary, and the benchmark other rates are checked against.

It rests on 230,715 games, which is 10.1% of the record.

Women’s ITF, the tier almost nobody discusses, rests on 553,157 games. That is 24.1%, nearly two and a half times as much evidence as the headline figure. Men’s Challenger carries more still, at 622,018 games and 27.1%.

So the best known number in the table is built on the thinnest slice of it. The least known numbers are built on the thickest.

Why this is not a complaint about sample size

230,715 games is plenty. Nobody needs to doubt 78.62%, and the figure reproduces the rate published elsewhere, on a method tuned to neither tour.

The problem is what happens when the rows are compared. A ranking implies the rows are the same kind of object, measured the same way with the same confidence. Here one row is a careful measurement of a small, heavily watched population. Another is a much larger measurement of a group nobody examines.

Both are correct. They are not equally informative about their groups, and the table cannot tell you that.

Three habits that help

Print the count beside the rate, always. It costs one column and it changes how every other column should be read.

Check whether the famous row is the big row. It usually is not. Attention and evidence tend to run in opposite directions, because the small elite group is the one people watch.

Be careful with the smallest row. Women’s Challenger holds 60.40% from 92,396 games, 4.0% of the total and the least evidence of any tier. It will move most when the data is refreshed.

None of this needs a model or a correction. It needs the denominators printed next to the percentages, which is the cheapest thing in the whole table and the first thing cut for space.

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. Counted from the point-by-point record.