Best of three — ATP and WTA tour
Everything except men’s Grand Slam singles.
21.4Average games
20Median
30.1%Go to a deciding set
157,933Matches counted
| Line | Over | Under | Share of matches |
|---|
| 16.5 | 77.4% | 22.6% | |
|---|
| 18.5 | 59.5% | 40.5% | |
|---|
| 19.5 | 51.7% | 48.3% | |
|---|
| 20.5 | 44.9% | 55.1% | |
|---|
| 21.5 | 40.2% | 59.8% | |
|---|
| 22.5 | 34.8% | 65.2% | |
|---|
| 23.5 | 31.0% | 69.0% | |
|---|
| 24.5 | 29.2% | 70.8% | |
|---|
| 26.5 | 22.6% | 77.4% | |
|---|
Best of five — men’s Grand Slams
A different sport for this purpose: two extra sets change every number.
32.8Average games
31Median
15.1%Go to a deciding set
3,650Matches counted
| Line | Over | Under | Share of matches |
|---|
| 16.5 | 96.6% | 3.4% | |
|---|
| 18.5 | 91.5% | 8.5% | |
|---|
| 19.5 | 89.3% | 10.7% | |
|---|
| 20.5 | 86.9% | 13.1% | |
|---|
| 21.5 | 85.1% | 14.9% | |
|---|
| 22.5 | 82.2% | 17.8% | |
|---|
| 23.5 | 79.7% | 20.3% | |
|---|
| 24.5 | 78.1% | 21.9% | |
|---|
| 26.5 | 70.1% | 29.9% | |
|---|
How this is counted
- Retirements and walkovers are excluded: an abandoned match says nothing about how long two players take to beat each other.
- Best-of-three and best-of-five are never mixed. Averaging them would misdescribe both.
- Long sets count as the games they were. A "12-10" at Roland Garros is real tennis; a "10-5" super-tiebreak is not, and the two cannot be told apart from the score. This affects 1.1% of matches and makes the tail very slightly long.
- Last five years only. Tennis has got shorter — 21.9 average games all-time against 21.4 recently — so an older base rate would describe a game nobody is playing now.
- This is not a betting signal. We tested our own model against every one of these lines: it beat "always call the more frequent side" by between −4.1 and +0.2 points. We publish the base rate because it is true, not because we can beat it.
