Total games in tennis: what actually predicts it (and what doesn't)
TL;DR. Total games is to tennis what total goals is to football, with two differences. First: the format imposes a rigid floor and ceiling. The second is counter-intuitive and we measured it across almost 20,000 matches: within that frame, what best predicts the total is not how evenly matched the two players are, but the games tempo each of them habitually plays at.
Note. Informational and statistical content. It is not betting advice nor a promise of profit. No statistic predicts a single match: they are there to estimate probabilities over the long run. 18+ only. Play responsibly.
Format sets the frame
| Format | Minimum games | Maximum |
|---|---|---|
| Best of 3 sets | 12 (two 6-0 sets) | 39 (three sets with tiebreaks) |
| Best of 5 sets | 18 | 65 (five sets with tiebreaks) |
On the current professional tour the deciding set is settled by a tiebreak, including all four Grand Slams since 2022 — before that some events played it out by two clear games, which is where those marathons came from. The variation still alive is the super tiebreak replacing the third set in many doubles draws, which cuts the total sharply.
What actually predicts it: each player's tempo
Intuition says a close match gives plenty of games and one with a clear favourite gives few. It sounds impeccable and it is nearly irrelevant: measured across 19,595 matches of our history, how evenly matched the players are barely anticipates the total, and a model built on it does no better than simply taking the average.
What does work is far simpler: how many games each player's matches tend to last. Some players systematically stretch matches out — they hold serve, return well, force tiebreaks — while others resolve or collapse quickly, and that individual tendency carries from one match to the next far better than any measure of closeness.
The underlying reason is that a clear favourite has two ways to win: a 6-1 6-2 rout or two tiebreaks. Both are comfortable wins on the set scoreboard and they produce opposite totals. The gap in level cannot tell them apart; the players' habitual tempo can.
The surface moves the level (and not the way people think)
Another myth worth dismantling. These are the average games per match measured in our own history, not an impression:
| Surface | Games per match (average) | Why |
|---|---|---|
| Grass | 24.4 | The serve dominates, breaks are rare and sets go to tiebreaks |
| Clay | 22.3 | More breaks, but also more long, contested games |
| Hard court | 20.8 | The most decisive of the three |
So clay produces more games than hard courts, not fewer. People assume the opposite because clay brings more breaks, and breaks shorten sets; what gets forgotten is that clay also brings far more long, contested games at both ends. Grass is by some distance the highest-scoring surface for totals.
How the model estimates it
Without unnecessary sophistication, because none is needed. For each player we take the average games across their previous matches and combine the two; that number is blended with the average for the tournament's surface, weighting the player more the more matches they have and leaning on the surface average when the sample is thin. The probability of each line is computed around the resulting total.
Note what does not go in: the Elo rating, or who the favourite is. That is not an oversight, it is the result of the measurement: since the winner does not predict the total, adding it would only contribute noise. The games engine and the one estimating who wins the match are deliberately independent.
One last piece of honesty: the estimator is calibrated to best-of-three. In best-of-five matches it falls short by construction, so there we publish the estimated total with a format warning but not the line probability. We would rather not give a number we know is biased.
Two warnings
- Retirements and walkovers. Tennis has more unfinished matches than any other sport, and that affects how games markets settle. The rules vary and are worth knowing before comparing histories.
- Historical game averages blend formats. An average mixing best-of-three and best-of-five matches describes nobody, and the same applies if it mixes singles with super-tiebreak doubles.
Frequently asked questions
What determines the total games in a tennis match?
First the format, which sets a rigid floor and ceiling. Within that frame, what best predicts it is each player's habitual games tempo, not how evenly matched they are: we measured it across 19,595 matches and the gap in level barely contributes.
Does a clay-court match have more games than a hard-court one?
Yes, contrary to what is usually assumed: in our history clay averages 22.3 games per match and hard courts 20.8. The highest surface is grass at 24.4, because the serve dominates and sets go to tiebreaks.
Does a clear favourite mean a short match?
Not reliably. A favourite can win by a rout or by two tiebreaks, and those are opposite totals. That is why our games estimator does not use who the favourite is: when measured, it added nothing.