Italian translation at settesei.it
No matter how well a player serves, he still needs to win some return points. While one-dimensional ATPers such as Ivo Karlovic and John Isner have demonstrated that an unbreakable serve alone can get you a steady paycheck and some quality time in the top 20, their playing style has never translated into a prolonged stay in the top ten.
Nick Kyrgios isn’t quite as tall as Isner or Karlovic, but his numbers are similar. In the last year, he has won 31.7% of return points, third-worst among the top 50, ahead of only those two players. In fact, since 1991, only five players have lasted a full season at tour-level while winning a lower percentage of return points. To make an impact in the upper echelon of the men’s game, the Australian will need to improve his return game in a big way.
To win matches, you need to break serve or win tiebreaks, and most players don’t demonstrate any particular tiebreak skill. That leaves breaks of serve, and to break serve, you need to win return points. Almost all ATP tour regulars win between 29% and 43% of return points, so a single percentage point or two is a meaningful distinction. While Milos Raonic‘s rate of return points won over the last 52 weeks is a Kyrgios-comparable 32.1%, no other top-ten player is below 36%.
If Kyrgios is to crack the top ten without any substantial improvement in his return game, Raonic is the model. Last year, Milos finished the season at #8 in the rankings despite having won only 33.7% of return points. That’s the lowest rate on record for a player with a year-end ranking in the top ten, and only the seventh time since 1991 that a RPW% below 35% earned someone a spot in the top ten.
Even at 33.7%–two percentage points higher than Kyrgios’s current rate–it took a remarkable run of tiebreak success for Raonic to win as many matches as he did. Milos won 75% of tiebreaks last year, a rate that almost no one has ever sustained beyond a single season. In other words, if Raonic is to continue winning matches at the same pace, he’ll probably need to post better return-game results.
To earn a place in the elite of the top five, the return-game threshold is even higher. Only two players–Pete Sampras and Goran Ivanisevic–have finished a season in the top five with a RPW% below 36%, and only two more–Andy Roddick and Stanislas Wawrinka–have done so with a sub-37% RPW%. Roger Federer, the most serve-oriented of the big four, hasn’t posted a RPW% below 38% in fifteen years.
The difference between 32% and 36% is enormous. To use a baseball analogy, a similar gap in batting average would be, roughly, from .240 to .280. The effects are equally meaningful. At 32%, a player is breaking serve roughly once per eight return games–considerably less than once per set. At 36%, he’s breaking serve almost once per five return games. Improve a few more percentage points to 39%, and he’s breaking every fourth game, almost twice as often as Kyrgios is now.
Those break rates are simply a way of quantifying what we already know at a general level: Players with strong return games have the power to decide matches. The more one-dimensional the playing style, the more likely a match is decided by just a few key points. And the smaller that number of points, the more that luck plays a part.
Of course, luck cuts both ways. It’s what makes players like Isner and Kyrgios so dangerous. Someone like Novak Djokovic or Rafael Nadal can usually dictate play, but against an unbreakable opponent, it all comes down to a few points in a couple of tiebreaks. So big servers tend to rocket into the top 30 or 40. A fifty-fifty winning percentage, especially coupled with a big upset and an occasional deep run at a big tournament, is plenty good enough to earn a spot that high in the rankings.
But without at least a mediocre return game, it’s tough for a big server to get beyond that level. Isner has managed it by winning tiebreaks at one of the best rates of all time, and even he has barely dipped his toe in the top ten. Raonic is a substantially better returner than the American, and it remains to be seen whether he can sustain his impressive tiebreak winning percentage and keep a spot among the game’s best.
Fortunately, Kyrgios has plenty of time to improve and break out of the mold of a one-dimensional big server. If he hopes to make a mark beyond the occasional upset and a home at the fringes of the top 20, that’s exactly what he’ll need to do.
I find this post interesting, because personally I had decided that Kyrgios is not a “servebot” at all. However I based this on his ability to create some really nice groundstrokes – for a guy his size he can be creative & deft. I have never seen Isner hit as pretty a backhand as Kyrgios can produce fairly often. His latest loss, to Murray at the FO, seemed more about a niggling injury and a fairly to hit his groundies consistently, and less about not having a good return game. And this reveals the value of stat analysis such as yours – you are able to show that in fact, he *is* playing quite a bit like a servebot in terms of that woeful stat for return points won.
The one time I actually used stats to make an argument of this sort was in a blog post I wrote (for Craig Hickman’s blog, back when he was still running it) about Andy Roddick – I noted that for the various return stats maintained by the ATP, Roddick was wretched in all categories. And Roddick was clearly not seen as a “servebot.”
Also of interest to me is that Isner bores me as a player . . . EXCEPT on those rare occasions when he goes for broke on his return game & produces an upset. Why he doesn’t take more risk, more often, I don’t know; he seems an innately conservative personality.
FYI here is the post I wrote on Roddick if you want to see how I used the stats:
http://craighickmanontennis.blogspot.com/2010/04/return-on-risk.html
I agree — poor returners like Isner should be more aggressive. After all, they’re probably not going to break anyway. Isner does sometimes go for big returns, and will occasionally return-and-volley. Seems like their best bet is to take advantage of their opponents’ poor service games, and one or two successful points like that are sometimes all it takes to rattle the server and sneak away with a break.
Here is the part of that blog piece where I wrote about Roddick’s return game – this was back in 2010, and Roddick, who was being coached by Larry Stefanki, had recently upset Nadal at Miami by attacking Nadal’s serve much more aggressively than usual:
“I missed the crucial 2nd and 3rd sets of the Nadal match at Miami, but I hear that in fact, Roddick did start attacking the second serve, and with good results. But if we check out Roddick’s ATP stats for 2010 so far, we’ll see that for the stat Stefanki was talking about, break points converted, Roddick is as bad as ever––just 31 percent. That ranks him 63rd on the tour in this area.
“It would be interesting to know why Stefanki mentioned this particular stat. It doesn’t seem especially meaningful when taken in isolation, given that the current leader is the amazing Evgeny Korolev. I think it more telling that Roddick trails badly in all return categories compared to the very top players. The category that really shows the difference is for return games won: Murray, Nadal, and Federer are ranked 5, 6, and 7 here, while Roddick is out of sight at 53.
“So unless the Nadal victory represents a sudden breakthrough for Roddick, he has got a lot of work to do on his return game if he is to meet Stefanki’s expectations (not to mention ours). Can he do it the way he did against Nadal––flick a switch mid-set? And even if he can, does he want to, or will his innate conservatism win out and keep him passive?”
Kyrgios is definitely better off the ground than Isner, Karlovic and Raonic for that matter.. Had Murray deep behind the baseline at times during R3 at roland garros.. As Nick’s game continues to develop over next 12-18 months his return points won will be on the incline.. It’s more poor shot selection and taking too bigger cuts when returning serve.. 5% increase shouldn’t be a problem considering it’s only 5 out of 100 points! By the end of 2017 Nick will be top 5 in the rankings!
A slightly different way of looking at the matter is to suggest that Kyrgios is a “serve prodigy” – a young player whose service game has become world-class while the rest has lagged behind. While he may remain in the same category as Karlovich, Isner and Raonic, we may also see his return game develop steadily over the next few years. After all, nowadays it is increasingly hard to become a world-class player before the age of 21-22. Maybe Kyrgios has coasted on the strength of his service game until it became manifestly insufficient. Now he has “bitten on granite” for a while, I expect him to start returning better.
It could be. I’d imagine that will be the focus of his next few seasons, and you never know how or how much a player will improve. OTOH, while there are plenty of examples of players who fit the ‘serve prodigy’ mold, there’s very little precedent for them substantially improving their return game.
If anything, many big servers become *more* one-dimensional, in part because as their serve improves, they must play more difficult opponents. Ivanisevic is an interesting example–his first full season was 36.2% RPW%, and his career RPW% was a bit lower; he never reached 37% for a full season. Roddick: broke in at 36.3%, never exceeded 37.5%, career RPW% was lower. Ljubicic: broke in at 34.9%, only beat 36.0% once, and stayed right at 35.0% for his career. Those numbers probably hide some improvement, because they faced tougher opponents as their careers progressed, but Kyrgios will confront the same challenge.
Great stuff, Jeff! I really enjoyed reading it. You talked about likely-hood of breaking based on percentage of return points won. A little over a year ago, I created a basic java computer program that calculates just that. Unfortunately, all the numbers I have are rounded, so that accuracy isn’t perfect, but based on a 31.7% rate of winning return points, Kyrgios should be winning 12.1% of his return games. However, according to the ATP site it’s at 14% (I’m sure you have the more exact number). That is a fairly significant difference as you know, although still not as high as Kyrgios would probably like. My explanation is that Kyrgios throws away points on return much more often than on serve. In other words, he plays more risky shots, particularly when he falls behind in a return game, because he doesn’t feel the pressure that he has to win it. Meanwhile, on the more important points, he plays less risky, leading to more success. That’s evidenced by his 38% winning rate on break points.
My question then is: Would Kyrgios be better served by playing every point the way he plays a break point? If that were the case he would break 22.4% of the time. Or is that 38% stat misleading in that if he played every point that way, he would not be winning 38% of points?
In tennis, the score of a game certainly dictates the way you play a certain point, but I think maybe Kyrgios would be better off playing with a little less risk on return, so that his matches aren’t decided on just a few points.
Here’s my code for game probability based on serve or return points won:
https://github.com/JeffSackmann/tennis_misc
I’ve got his break rate at 13.4% (which I think includes last year’s Nottingham CH, hence the difference)
http://www.tennisabstract.com/cgi-bin/leaders.cgi?f=s91w1
Without looking too closely, an important thing to remember is that a disproportionate number of any player’s break points will come against weaker servers, so it’s to be expected that win% on BPs is better than win% on all return points. For instance, among the top 50, RPW% is 36.8%, but win% on BPs is 40%.
That’s a great point in the third paragraph. I had noticed that such was the case for many players, but I hadn’t thought of that exact reason. I always just thought of it as the server facing pressure and the returner using momentum from having just won the previous points. I’m sure that’s a factor too, but your explanation is probably a bigger factor.
Great stats, and very helpful in understanding what makes a top 10 player, versus a player who’s ranked 11-20, or 21-30.
At the same time, all the examples are players whose games don’t include great wheels. This must influence our understanding of why they struggle with a poor RPW: since they cannot count on winning points if they manage to set them at neutral with an assertive (but not outright winning) return, they suffer from missing too many 2nd serve returns on days when they’re playing at less than 95% (a guess, but you get the point), or possibly against someone far above them in the rankings, when only their 100% game will get them the W.
Among ATP players Federer, Nadal, Djokovic, Murray, Ferrer and a good number of other players can put a 2nd serve return in play that’s solid, or perhaps even a little forcing, and then count on their consistency and wheels to help bring the point to their side of the ledger. On the other hand players with lower rates of consistency, who’re frequently big bangers with slower feet, have to rely on exploiting 2nd serve return opportunities with huge returns. Which is to say, they’re more likely to miss and have a lower RPW.
I suspect that it would be even more telling to develop a stat that compared 2nd serve returns in play versus 2nd serve RPW. Then we’d get a better picture of *why* the more fundamental RPW stats fall out as they do.
Interesting analysis. I have to believe another reason for the BP vs RPW discrepancy is the deuce court vs. ad court bias. While there are more return points played in the deuce court, there are more break points played in the ad court. So rightie v. rightie you may have a guy with a monster kick serve getting it up high to a rightie’s backhand and the server immediately takes control of the point. Or you might have Nadal hitting the hook out wide and then unloading a forehand first ball. Would be interesting to see the variation in deuce court vs. ad court SPW and RPW. Not sure you can get this from the ATP stats but I think it could be derived from the ITF stats that Jeff has posted in the past.
Another area I’ve never seen analyzed is in whether or not there is a statistical difference in hold %age depending on if someone serves first after a changeover or not. It may be slight, but maybe there is something significant in getting up off the changeover and serving the first point after sitting down for 90 seconds. Or maybe it’s having to serve 30 seconds after losing a brutal rally after 8 deuces. Not sure who it might favor, but would be interesting to look at nonetheless.