Return of Serve

07-20-2012, 01:59 PM
I'ev been crunching some numbers recently from the ATP Tennis site. I'm tracking world rankings over time and correlating these with any available statistics from the tour. It appears, at least in my preliminary analysis, that the best single predictor of a high ranking is win percentage on return of serve. The best players consistently finish near the top in this category. This is even more strongly correlated than aces, winning percentage on serve, unforced errors, and so forth. More to follow shortly.

07-20-2012, 02:03 PM
Interesting but it makes sense. Someone good at returning serve almost certainly has good ground strokes, which is the most important tool in today's game.

07-20-2012, 02:13 PM
Just as a side note, I believe throwing in the statistics on winners and unforced errors muddies the waters a bit, at least from the statistical aspect. These measures introduce subjectivity, that is, human judgements about what count as winners and unforced errors. I am running parallel analyses to see whether or not inclusion of these variables add anything significant to the analysis. So far they are slightly significant, but the variance is high, meaning that it appears these variables are not as stable as stats like first serve percentage, wins on first serve, second serve percentage, wins on second serve, double fault percentage, ace percentage, etc. Anyway, still, win percentage on return of serve, especially win percentage on opponent's first serve, is crucial.

One other thing to notice. Excellent players also tend to be great servers, and they don't have to face themselves, so they improve their own return of service stats by having their own serve taken out of the calculation. This may have a confounding effect on the statistics. I'm looking at possible multicollinearity in this regard.

Henry Chinaski
07-20-2012, 02:56 PM
winners shouldn't be subjective, but unforced errors clearly are.

Uncle Latso
07-20-2012, 06:55 PM
winners shouldn't be subjective, but unforced errors clearly are.
Winners can't be subjective, yet it's far from a reliable stat, as first of all winners should be counted as % of the number of strokes (as any mug can hit 30 winners if a match lasts 300 points, while 30 winners from 80 points is another story).

Then they need to be separated from the service winners, as Karlovic should have a minimum of 50% of his winners from this shot, while f.e. Nadal's are 90% from the FH, which is a very different thing.

About the magical statistic that the return of serve is the best indicator - i agree, but it's just as good of an indicator as the ranking is.

Better % of returns, points won on return or whatever - higher ranking. And vice versa. So this won't be very helpful if it's intended to be used in defining odds for betting f.e.

It's like saying that the higher average speed of the driver defines him as the better driver - yes, so does the ranking of the race...

To have any decent use of the stats in tennis you need much, much more precise and in depth data like - winners from the net against defensive shots/offensive shots, hitpoint of the winner - first third near the net/middle/last third till the baseline/behind the baseline, winners from serve, winners from return of first serve with oponent at the net/baseline, winners returning second serve, etc., etc.

That's for the winners only and you can have the same for UEs, forced errors, BH, FH, percentages of winners/UEs per type based on number of shots per type, etc., etc.

That would give you a clearer picture of who's doing what. The plain official stats are useless.