Tennis statistics by surface: why a yearly average is useless
TL;DR. Tennis is one of the few sports where the playing surface changes radically between tournaments. Clay, hard courts and grass produce matches so different that a single rating — a whole-year average — describes an average player who does not exist. A per-surface rating is what fixes that.
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.
What changes on each surface
| Surface | Speed and bounce | Effect on the game |
|---|---|---|
| Clay | Slow, high bounce | The returner has time: more breaks, long rallies, rewards endurance |
| Hard court | Intermediate, regular bounce | The most neutral; rewards an all-round game |
| Grass | Fast, low and irregular bounce | The serve dominates: few breaks, short points |
The effects are large: the share of points won on serve, the frequency of breaks and the total games all shift at once when the surface changes. It is not a fine adjustment: it is a different sport.
The problem with a single rating
Blend the whole season into one number and a clay specialist ends up undervalued on their surface and overvalued on grass, with a pure server the other way round. The number is perfect for the average player and wrong for almost everyone else — who are precisely the interesting ones.
The obvious fix is a per-surface rating. But it has a problem of its own: sample. Counting only grass matches leaves many players with a handful a year, and a rating on that basis is very unstable.
Blending is the answer
Our model does not choose between the two: it combines them. The general rating supplies stability (many matches) and the surface rating supplies specificity (few matches but highly relevant), and the result is a weighted blend of both.
The weight on the specific part is calibrated from history, not picked by eye: it is the split that best anticipates results across the whole set of matches. Too high a weight would turn the rating into noise; too low would return it to the yearly average we were trying to avoid.
The details that slip through
- Not all hard courts are the same. There are fast and slow tournaments within the same category, and the speed of a given court can vary between editions.
- Indoor or outdoor. Under a roof wind and sun disappear; the same player performs differently.
- Altitude. At high-altitude venues the ball flies further and the game speeds up, regardless of the court material.
- The transition. The first tournaments after a surface change throw up more surprises than usual: every player's recent history is from a different surface.
Where to see it
On every tennis match page and on the tennis statistics page. The rating concept lives in the Elo glossary entry.
Frequently asked questions
Why is a single yearly rating not enough in tennis?
Because the three surfaces produce very different matches. A single rating undervalues a clay specialist on their own surface and overvalues them on grass: it fits the average player and misfits almost everyone else.
Why not use only the surface rating?
Lack of sample. Counting only grass matches leaves many players with a handful a year, and a rating on that basis is very unstable. That is why it is combined with the general rating.
Do all hard courts behave the same?
No. There are fast and slow tournaments within the same category, and factors like playing indoors or at altitude change the ball's behaviour regardless of the court material.