In any contest played in phases, the early phases get read as omens. Winning the first one feels decisive. The interesting question is whether the margin of that first phase carries information beyond the fact of winning it. In tennis there is a clean answer.
A dataset of 116,382 completed best-of-three singles matches runs from January 2023 to September 2026. It records how often the player who lost the opening set still won the match. The headline answer is 16.52%. The useful answer is what happens when you sort those matches by how the opening set finished.
Seven scorelines, in order
A first set lost on a tiebreak at 7-6 is followed by a comeback 21.92% of the time, across 14,265 matches.
Lost 6-4, a two game margin, the rate is 19.41% over 25,546 matches. Lost 7-5, also two games, it is 18.89% over 10,252. Lost 6-3, three games, 17.39% over 25,970 matches.
Lost 6-2, four games, 13.3%. Lost 6-1, five games, 11.06%. Lost 6-0, six games, 7.18% across 5,833 matches.
The order is the thing. Comeback rate falls as the margin widens, without a single exception across all seven scorelines. The two scorelines that share a margin of two games sit next to each other, at 19.41% and 18.89%.
Why that ordering matters more than any single figure
A pattern that holds across every category is much harder to explain away than a difference between two groups. With seven ordered scorelines and a mechanism you can state in a sentence, the finding stops looking like an accident of sampling.
The mechanism is simply that the margin measures distance. Six games apart says the two players were not close on the day. One game apart says they were. The comeback rate follows that reading faithfully.
So a 7-6 first set and a 6-0 first set both count as one set to nil. They sit three times apart in what follows. A scoreboard that records only the set count has thrown away most of the signal.
The phase-two version of the same question
The dataset also tracks the position within the second set. Take the second set 6-4, levelling the match, and the win probability is 50.16% across 8,132 matches. Taking it 6-1 puts it at 58.53% over 3,682 matches.
Those are set scores. Game scores inside the second set read differently. Leading 3-0 gives 45.22%, leading 3-1 gives 39.02%, and leading 2-0 gives 36.8%.
The two scales are easy to confuse and the gap between them is large. Levelling the match is worth far more than leading early in a set. Anyone quoting a figure here has to say which one they mean.
What to carry into any phased contest
Record margins, not just outcomes. A competition table that stores who won each phase and by how much supports questions that a win-loss column cannot answer.
Check whether your breakdown orders itself. If the categories have a natural order, as game margins do, the rates should follow it. When they do not, either the samples are too small or the categories are not measuring what you assumed.
Say which unit you are counting. The difference between a set score and a game score here is the difference between 50.16% and 36.8%.
All of the above comes from the Live Tennis API, a JSON service for developers. It covers live scores, fixtures, players and match events, with point-by-point history behind them. The figures are published at blog.livetennisapi.com/studies.json under CC BY 4.0, free to reuse with attribution.
Retirements and walkovers are excluded. The counting is done from the point-by-point tape instead of estimated, which is why the margin of every set is available in the first place.
