Carlos Alcaraz and the Fruits of Shot Tolerance

Carlos Alcaraz making it look easy

Tennis people increasingly talk about “shot tolerance.” What does it mean?

There’s no standard definition. A Google summary settles on this: “a player’s ability to get to and return a given ball as desired.” Other definitions focus more on avoiding unforced errors: How long can a player stay in a rally without making a mistake?

I think of it like defense, in a very broad sense. Usually when we talk about defensive skills, it’s a lunging Andy Murray, recovering balls that nearly end the point, like big serves or would-be winners. In pro tennis, though, nearly every shot has an offensive component. That is to say, almost every shot tests the defensive skill–or resourcefulness, or shot tolerance–of the other player. A heavy Jannik Sinner forehand right at the feet, a Novak Djokovic backhand angled wide: Can you handle that and keep the point alive?

Another way to conceptualize shot tolerance is to imagine a win probability stat that updates with each shot. When Sinner hits that heavy forehand, his chances of winning the point against an average opponent increase to, say, 70%. A player with high shot tolerance will somehow save more than 30% of those points. A player with low shot tolerance will fail to get as many back (or hit weaker replies), and he’ll win fewer than 30% of those points.

We don’t have that all-knowing win probability stat, so we can’t measure shot tolerance so directly. Because the concept is fuzzy, I don’t imagine we’ll land on a fully satisfying way to quantify shot tolerance. But it’s worth the attempt, and it will help us explain how Rotterdam champ Carlos Alcaraz works his magic.

Long rallies

Start with the basics. Players with high shot tolerance should win more long rallies, right?

I drew the line at six shots, including rallies where the sixth stroke was an unforced error. I don’t want to muck things up by mixing surfaces, so we’re sticking with hard courts today. Based on Match Charting Project data, here are the men who have won the most of these “long” rallies on hard courts since the beginning of 2024:

Player            6+ W%  
Jannik Sinner     56.1%  
Carlos Alcaraz    55.6%  
Alex de Minaur    55.1%  
Grigor Dimitrov   55.1%  
Joao Fonseca      55.0%  
Learner Tien      54.5%  
Andrey Rublev     54.2%  
Novak Djokovic    53.7%  
Daniil Medvedev   53.7%  
Alejandro Tabilo  53.2%

The top of the list is as expected: Sinner and Alcaraz can outlast most opponents and have the ability to end the point. Fonseca and Tien probably won’t sustain these numbers, since they haven’t played the same level of competition as the others. Tabilo’s position is dicey, too, as we don’t have as many charted matches of his. Alexander Zverev is next on the list, if you’d like to promote him in Tabilo’s place.

Complicating matters is how these points end. The goal isn’t to sustain the longest rally possible. At some point shot tolerance gives way to power and calculated risk-taking. Some players are particularly strong on the pure shot-tolerance side of things, avoiding unforced errors in these long rallies:

Player            6+ Rally UFE%  
Casper Ruud               15.8%  
Bu Yunchaokete            17.9%  
Lorenzo Musetti           18.5%  
Daniil Medvedev           19.5%  
Alex Michelsen            19.5%  
Frances Tiafoe            20.0%  
Karen Khachanov           20.2%  
Alejandro Tabilo          20.4%  
Learner Tien              20.7%  
Novak Djokovic            21.0%

There’s some overlap between the two lists, but not much. Sinner’s error rate is better than average, at 22.3%, while Alcaraz’s is worse, at 24.1%. In the Rotterdam first round against Botic van de Zandschulp, Alcaraz committed unforced errors on 40% of points that reached the sixth shot. He still somehow won half of the long points.

There’s a relationship between win rate and error rate on long points–there pretty much has to be, since errors are points lost. But error rate explains less than 30% of the variation in long-rally winning percentage. Alcaraz, for one, breaks the mold by committing a lot of errors yet winning the majority of the points:

Alcaraz’s errors don’t usually expose a weakness of shot tolerance. They reflect a gamble. (Sinner is similar, though his groundstrokes are so imposing that he can do more damage with less risk.) We can’t just count errors and create a shot-tolerance metric, but we also don’t have the ability to ask players what they were thinking when they attacked every shot. Isolating shot tolerance requires a different approach.

Accepting errors

Let’s shift from points to shots. Again for hard-court matches since the start of last season, I tallied each player’s baseline strokes starting from the fourth shot of each rally. Shot tolerance is useful for serve returns and plus-ones, but those shots are so often out of a player’s control. And since most points are short, returns and plus-ones end up dominating the data. To get a sample of shots that reflect what we think of as “rallying,” we need to discard those.

(The word “baseline” is doing a ton of work here. Shot tolerance isn’t usually about making volleys or smashes, or about executing passing shots. So I’ve excluded every shot at the net, as well as every shot when the opponent is at or approaching the net.)

As with the long rallies, we can start by getting a sense of the shot-tolerant all-stars, the guys who are best at avoiding unforced errors:

Player                UFE/Shot %  
Learner Tien                7.4%  
Alexander Shevchenko        7.9%  
Lorenzo Musetti             8.2%  
Alejandro Tabilo            8.2%  
Tommy Paul                  8.7%  
Frances Tiafoe              8.9%  
Carlos Alcaraz              9.2%  
Jannik Sinner               9.2%  
Matteo Arnaldi              9.2%  
Casper Ruud                 9.7%

Musetti, Paul, and Ruud are names you’d probably expect to see here. Alcaraz and Sinner are more bracing. They don’t stand out as error-avoiders when we look at long rallies, but on a per-shot basis, they do.

In general, there’s a predictable trade-off. Players who hit more winners (and force more errors) commit more unforced errors. But the relationship between the two numbers is not the same for everyone. Here’s the same UFE-top-ten list, with winner rates added:

Player                UFE%  W+FE%  
Learner Tien          7.4%   7.3%  
Alexander Shevchenko  7.9%   9.1%  
Lorenzo Musetti       8.2%   9.1%  
Alejandro Tabilo      8.2%  10.6%  
Tommy Paul            8.7%  11.1%  
Frances Tiafoe        8.9%   8.0%  
Carlos Alcaraz        9.2%  10.1%  
Jannik Sinner         9.2%  14.1%  
Matteo Arnaldi        9.2%   8.3%  
Casper Ruud           9.7%  11.1%

Holy Sinner! The typical ATP regular hits slightly more winners than UFEs at these stages of the rally. Tien, Tiafoe, and Arnaldi are on the wrong side of the scale. Tabilo, again, is probably favored by a limited (and biased) sample. And Sinner … well, you need to go 15 more players down the list before you find anyone who cracks as many winners as he does, and Karen Khachanov coughs up a quarter more errors to accomplish the feat.

Here’s the full scatterplot:

This is a tighter relationship than the one pictured earlier. The player-to-player variation in winner rate explains half of the difference in error rate. Yet again, some players defy the usual tradeoff. The closer they are to the upper left corner of the graph, the more risk-free their aggression.

Controlled (for) aggression

It feels weird to quantify shot tolerance by considering winners, but that’s exactly what we’re going to do.

As a simplification, imagine that we put every shot into one of two categories: aggressive or defensive. Aggressive shots really aren’t about shot tolerance. Unless the aggression is just a last-ditch effort from a hopeless position, it’s a shot that the player more or less knows he can make. He hits hard, aims for the line, and it ends the point one way or the other.

Shot tolerance isn’t about the unforced errors that come from that kind of risk. We’re interested in how steady a player is on every other shot.

To get there, we’re going to string together some assumptions. You’ll probably disagree with some of them, and you’ll almost definitely disagree with some of the results. But bear with me for a minute anyway.

Say that the “cost” of winners (and forced errors) is half as many unforced errors. The usual ratio is closer to 1:1, but that doesn’t count forced errors, and it counts the kind of “bad” errors that come from low shot tolerance. So the 2:1 ratio means that if a player is going to hit two winners, the cost of doing business is one unforced error. For Alcaraz, his 10.1% winner rate implies that even if he’s playing flawlessly in non-aggressive situations, he’ll still have an error rate of about 5%.

From there, we come up with a “non-aggressive” error rate. Since Alcaraz’s total error rate is 9.2% and we’re writing off 5% as the cost of his aggression, that leaves us with 4.2%. We’ll divide that by the number of non-aggressive shots–that is, 100% of his shots, minus his winners, minus the 5% of aggressive errors. So: 4.2% divided by (100% – 10.1% – 5% =) 84.9%. Punch it into the calculator, and we get a non-aggressive error rate of 4.9%.

In more positive terms, that’s a “shot tolerance” of 95.1%. That is, when he has a reasonable chance of making a shot (meaning his opponent didn’t hit a winner or generate a forced error), and he doesn’t go big, he makes the shot 95.1% of the time. Average among tour regulars is 93.7%. Here are the top ten, excluding the names like Tien and Tabilo that I quibbled with above:

Player           ShotTol  
Jannik Sinner      97.3%  
Tommy Paul         96.2%  
Lorenzo Musetti    95.7%  
Andrey Rublev      95.4%  
Carlos Alcaraz     95.1%  
Grigor Dimitrov    95.0%  
Casper Ruud        95.0%  
Daniil Medvedev    94.9%  
Frances Tiafoe     94.5%  
Alex de Minaur     94.4%

So, do you agree that Alex de Minaur doesn’t have the shot tolerance of Rublev, Dimitrov, or Tiafoe? What about leaving Djokovic out of the top ten entirely? (His figure over the last 13 months is a below-average 93%.) Of course you don’t. That’s what I’m here for.

Djokovic is easy enough to explain. His last 13 months have been rocky. If we expand the time frame back to 2020, his number shoots up to 97.1%.

Dimitrov is higher than expected because of his reliance on the slice. Dan Evans did well by this metric, too. Slices are a good way to keep balls in play, even if they don’t generate a lot of offensive opportunities. I hesitate to exclude slices from the metric, but some kind of adjustment is probably in order.

De Minaur may expose another limit of this approach. One of the assumptions I strung together is that everyone’s winners come at the same cost. But the Aussie is relatively small: He can’t just wave a magic wand and generate winners like Sinner can. He needs to take more risks to end points. His winner/error ratio for this set of shots is 10.7% to 10.0%. My model assumes that a bit more than half of his errors are aggressive shots that missed the mark. But what if it’s more? In the bizarro world where players really did register their intention before every shot, we might find that many more of de Minaur’s errors fall in that category.

Or, maybe, he’s not quite as sturdy as we think he is. Either way, it’s something to watch next time you tune into a match of his.

Return to Alcaraz

Why, then, is Carlitos’s name in the headline? Sinner (or Tien, or Ruud) is better by most of these metrics.

What fascinates me about shot tolerance is what it doesn’t explain. If shot tolerance determines anything, it should tell us who’s going to win long rallies. And to some extent it does: Sinner tops the shot tolerance list, and he wins more long rallies than anybody else. (Though that doesn’t prove much: Sinner is better at just about everything, related or not.)

Yet Alcaraz isn’t far behind in long rallies. The Spaniard comes in second with room to spare. Tommy Paul does better on my shot tolerance metric, yet he wins only 51% of long rallies.

The X-factor, I think, is that shot tolerance is instrumental. You can win some points by out-shot-tolerancing your opponent, because yes, eventually they will miss. But nearly as often, you will keep the rally alive, even slightly in your favor, and they’ll take a risk that pays off. If that opponent is Sinner, that’s the most likely outcome. Just ask de Minaur, who has lost all ten of his career meetings with the Italian.

Alcaraz’s long-rally magic doesn’t fully show up in the shot tolerance metric because it isn’t confined to the baseline. The signature Carlitos point is a ten-stroke rally that he puts away at the net or polishes off with a drop shot. His baseline prowess isn’t quite a match for Sinner, and his net skills probably rank behind those of Federer or Nadal. But has there ever been a player who could go from gutbusting rally to all-court acrobatics with such success?

The Spaniard approaches the net half-again as often as Sinner does. He wins nearly three-quarters of points when he does so. In today’s game, a mid-rally net approach has to be earned, and many strong forecourt players don’t have the baseline skills to create those chances.

Shot tolerance, then, is necessary but not sufficient. (And that’s even ignoring short points. Impregnable rallying doesn’t count for much when the serve is unreturnable.) Sinner earns his point-ending chances with sturdy baseline work, then converts them from the same position. Alcaraz is nearly as good at keeping the point alive, and he has more options than anybody when it comes to finishing it.

* * *

Subscribe to the blog to receive each new post by email:

 

The Tennis 128* Welcomes Aryna Sabalenka and Jannik Sinner

She’s number one (hundred and seventeen)

A few young players are working their way up the ranks of the all-time greats.

In 2022, I published the Tennis 128, my ranking of the 128 best players of the previous 100 years. It was (accidentally!) well-timed, as there were only a few legends in mid-career at the time. That has changed. At the end of 2023, I ran the calculations again and found that Iga Swiatek had earned a place just outside the top 100.

Now it’s the turn of Aryna Sabalenka and Jannik Sinner. Both number ones posted outstanding campaigns in 2024, enough to move them into my top “128.” (It’s now 131, since I’m not going to give anybody the boot.) Several other players improved their cases, though not enough for inclusion. We’ll probably have at least one new addition a year from now, as well.

74a. Jannik Sinner

Sinner posted a season for the ages, losing just three matches on hard courts and picking up two majors. He’s still at it, adding the 2025 Australian Open and reaching a new peak Elo rating, placing him among the top ten of the Open era.

My plan was to update the ranking at a logical time, like the “offseason” in December. Had I done that, Sinner would have landed in the mid-80s. I can’t yet do much with his 2025 Australian Open title, but I can use his ever-higher peak Elo rating. That’s the difference between a mid-80s ranking and his new position of 74th.

The Tennis 128 algorithm is based on three components: peak Elo, year-end Elo rating in the player’s best five seasons, and year-end Elo rating for the player’s entire career. The latter two factors make it difficult for a 23-year-old to achieve a high ranking–and understandably so, as it takes time to build an all-time-great résumé. But peak is different. Only a couple dozen men have ever played as well as Sinner is playing right now.

Appropriately, he slides into the list just ahead of Lew Hoad, another man who reached stratospheric peaks. Rod Laver was one of many who long considered Hoad the best ever. Now, to the extent fans still think of him, the assessment is more that on his best day, Hoad could outplay anybody. I’m not sure if Sinner makes quite the same impression, but he’s getting there.

93a. Iga Swiatek

Swiatek lost her place at the top of the WTA rankings, but she added a major in 2024 and posted another solid season. That was enough to move her inside the top 100, to a new position just ahead of Simona Halep.

117a. Aryna Sabalenka

I wrote last year that Sabalenka needed a “particularly dominant season” to crack the list, and she delivered. Two majors, two more notable titles, and a new peak Elo rating to kick off 2025.

One might argue that Swiatek, Sabalenka, or both, should rank even higher. Iga has five majors and Sabalenka has three. There are a lot of names ahead of them on the list with fewer. Part of the issue is that they are young: If they keep playing at the same level for another half-decade or more, they’ll move up quite a bit. I designed the Tennis 128 to assess careers, not careers-in-progress, so it’s a bit awkward to apply the same algorithm to players in their mid-20s or younger.

The other factor is that Elo doesn’t hold the current era of women’s tennis in particularly high regard. Swiatek reached a peak rating of 2,287, good for a place in the top 30 all-time. But Sabalenka has yet to crack 2,200, and Iga has dropped back to 2,154. The Tennis 128 algorithm has a mild era adjustment, because I don’t entirely trust Elo to compare eras. But the adjustment depends on several years of data, so it won’t fully affect 2023 or 2024 for some time.

There’s a reasonable case that I’m underrating the current era. Top players lose early more than the greats of the past, and that is, at least in part, due to an ever-stronger field. Accordingly, we don’t get a lot of head-to-heads among the handful of top women, and that makes it harder to assess just how good they are. I’m keeping an open mind about this, but that won’t affect the rankings for some time: We simply can’t judge the early 2020s from the vantage point of 2025.

On deck

Carlos Alcaraz and Alexander Zverev both rank around 140. Another 2024-like season from Alcaraz will be enough for him. Zverev will need a step forward, as his peak Elo is holding him back.

Daniil Medvedev is still around 150. The way things are looking at the moment, that’s where he’ll remain.

The next active player on the list is Naomi Osaka, around 180th. She’ll need to return to top-ten form to climb the list, and she’ll need more than that to break into the top 128.

After that comes Coco Gauff, the woman most likely to join the Tennis 128 a year from now. She sits around 200th, about where Sabalenka was a year ago. If she can turn in a season like Aryna’s 2024, increasing her peak Elo rating in the process, she’ll probably make the cut. If not, it would take (at least) one more year.

No one else is really on the horizon. Elena Rybakina ranks around 250th. Madison Keys is also in the top 300. As Sinner has shown, a monster season with a historically great peak suffices to rocket up the list. So it’s always possible that someone like Rybakina or Qinwen Zheng will come out of nowhere, even if their timetables are likely more conservative.

Check back in December for 2025’s tweaks to the list–I might even do the next update on schedule!

* * *

Subscribe to the blog to receive each new post by email:

 

The Second-Serve Woes of Arthur Fils

The outcome of this point depends a great deal on whether Arthur Fils is about to hit a first or second serve.

Arthur Fils is a bit of a forgotten prospect. A year younger than Carlos Alcaraz and two years older than Joao Fonseca, he isn’t considered to have the awe-inspiring talent of either. Then again, who does? By any other standard, the 20-year-old Frenchman has made tremendous progress. Standing at #19 on the ATP computer, he’s the top-ranked player under 21, and he’s even within the top five of under-23s.

There’s no secret to the Fils game. He hits hard, smacking serves over 130 miles per hour and occasionally connecting with a forehand that might even cause Fonseca’s eyebrow to twitch. In Saturday’s Davis Cup match against Thiago Seyboth Wild, he scored 11 aces in only 54 service points. He won 29 of 32 points when he landed his first serve.

The Frenchman has plenty more to his game, as well. He is sturdy off both wings, unafraid to battle from the baseline. He rates +5 in Forehand Potency, hardly a tour-leading number, but better than the likes of Daniil Medvedev and Holger Rune. By the same metric, his backhand is neutral. That’s another number that looks better in context: He won’t win any awards, but the shot isn’t holding him back.

At first glance, then, his biggest challenge is making inroads on return. He has won 36.2% of return points over the last 52 weeks, a number just below the usual minimum for an elite player. It ranks him above Rune, Hubert Hurkacz, and Taylor Fritz, but behind most of his fellow top-20 players. It isn’t an obstacle to cracking the top ten, though, as Hurkacz and Fritz have shown.

Fils’s problem now is what happens when he misses that big first serve. Among the ATP top 50, he ranks 16th in first-serve points won, just ahead of Karen Khachanov. By second-serve points won, he ranks a dire 43rd, several places down the list from Sebastian Baez. Even in Saturday’s Davis Cup rout, the Frenchman failed to win half of his second-serve points.

What’s going wrong? Is it something that young players tend to improve? What does a second-serve weak point say about a prospect’s future trajectory?

Second to many

Let’s get a sense of the typical relationship between first- and second-serve win rates. This scatterplot shows the ATP top 50 over the last 52 weeks:

Players who are successful behind one serve are generally successful behind both, with Jannik Sinner leading the way in the upper-right corner. Carlos Alcaraz and Novak Djokovic rank among the second-serve win rate leaders, though their first-serve results aren’t as strong. Grigor Dimitrov and Giovanni Mpetshi Perricard pop out as men with effective first serves who don’t do as much behind the second.

Fils hardly sticks out as a disaster, especially with Alexander Bublik there to distract us. Still, for a man who can win more than three-quarters of his first-serve points, he sits in the wrong part of this grid. There’s virtually no difference between his and Ben Shelton’s first-serve results, yet the American wins five percentage points more often with his second serves. Shelton is top-five in hold percentage, while Fils languishes outside the top 30, despite a near-identical first-serve success rate.

The diagnosis

It’s one thing to say that a player doesn’t win enough points behind his second serve. Can we figure out why?

Start with two things it’s not. Fils commits more double faults than average: 4.1% of serve points versus the typical top-50 rate of 3.4%. But even when he lands his second serve, the results are disappointing. Instead of 43rd among the top 50, he’s 37th. Fewer doubles might be nice, but they would barely move the needle.

We can also cross particularly soft second serves off the list. At the US Open last year, Fils averaged 115 miles per hour on first serves and 92 miles per hour on seconds. That’s a gap of 23 miles per hour–exactly in line with the norm among players who reached the third round. Jannik Sinner averaged 91 miles per hour with his second serves, so raw speed isn’t the problem.

That leaves us with where the second serves are landing. Unscientifically, I get the impression that Fils’s second serves don’t land particularly deep. Khachanov, for instance, hits a bog-standard topspin second, yet it’s fine because he consistently drops it deep in the backhand corner. The Frenchman sometimes hits that serve, but just as often his not-enough-topspin delivery lands in the middle of the box. In 2025, against this field, that’s not going to cut it.

Match Charting Project data can’t tell us how deep the serves are landing, but it does tell us what direction they go:

           Dc-Wide  Dc-Body   Dc-T   Ad-T  Ad-Body  Ad-Wide  
Fils 2nds    17.5%    38.1%  44.4%  11.3%    20.4%    68.3%

First, the Frenchman rarely goes for the forehand. With a really good second serve, that’s smart. But with less imposing strikes, it gives opponents options. Last week I linked to an analysis of Alcaraz’s loss to Djokovic in Melbourne, where Alcaraz went to the backhand side on second after second–probably too often. That gave Novak the flexibility to position himself differently and play the returns more aggressively. Fils is making a similar offer to everyone he plays, especially on the ad side.

Second, that’s a lot of deuce court serves down the middle. “Body” is misleading–usually when we talk about body serves, we mean the really good ones, flying 125 miles per hour at the returner’s left hip. That’s not what’s happening here. A more appropriate name for the category is something less inspiring, like “serves that aren’t in a corner.” Whether the returner has move a couple steps or none at all, those are booming forehand returns waiting to happen.

Here are the directional breakdowns of a few men with strong second-serve results, for comparison:

           Dc-Wide  Dc-Body   Dc-T   Ad-T  Ad-Body  Ad-Wide  
Fils         17.5%    38.1%  44.4%  11.3%    20.4%    68.3%  
Shelton      27.3%    40.3%  32.4%  21.6%    35.0%    43.4%  
Djokovic     35.5%    38.8%  25.6%  30.6%    27.2%    42.2%  
Sinner       14.1%    61.5%  24.4%  15.0%    49.8%    35.1%

Both Shelton and Djokovic mix things up a lot more. Sinner does everything I’ve just criticized about Fils and, at least in New York last fall, he did it at the same speed. Sheepish grin emoji. Surely there’s more going on there, but we’ll have to save it for another day.

In development

In one way, clear weaknesses are a good thing. It’s easier to identify a clear target area for practice than to vaguely aim to get a little better at everything. I’m sure Fils and his team are aware that he should get more out of his second serve, and we’ll see him change things up.

The question, then, is if those efforts are likely to bear fruit. Many players struggle to meaningfully improve their stats once they’ve established themselves on tour. It’s hard enough to tread water. Opponents figure out how to counteract their weapons, age and wear-and-tear take their toll, and–if the player is lucky enough to climb the rankings–the average opponent in later rounds is stronger.

Fils has compared himself to Jo-Wilfried Tsonga. Tsonga never had a sub-50% second-serve win rate, but he did have middling figures around 51.5% in his first two years on tour. On the other hand, Tsonga was a late bloomer: He was barely playing Challengers when he was 20.

For a more comprehensive answer, I found 75 men who played full seasons on tour at age 20 since 1991. All but five of them eventually bettered their age-20 second-serve win rate, and the average improvement was 2.7 percentage points. If that holds true for Fils, he’ll peak around 52% on second serves. That’s not bad, though I suspect he would be disappointed if he never does better than that. It would still be an underperformance for someone with his first-strike weapons.

Indeed, the one-third of these players who posted sub-50% second-serve win rates at age 20 improved more than the larger population. They gained 3.4 percentage points. Among them are some names that might inspire even greater feats from the young Frenchman:

Player           Age 20 2ndW%  Peak  Peak 2ndW%    
Andrey Rublev           49.3%  2020       53.9%    
Nicolas Almagro         49.1%  2012       55.1%    
Sam Querrey             48.2%  2015       54.1%    
Tommy Haas              47.7%  2006       54.8%

All four won fewer second-serve points at age 20 than Fils did in 2024. Each one fully shored up the weakness– at least for one season. 54% isn’t elite, but it’s more than good enough, especially when paired with a top-notch first serve.

To be clear, I’m not forecasting a Haas-like, seven-point improvement for Fils. Repairing weak points at tour level is exceptionally difficult, and even the three-percentage-point norm will take a lot of work. It is nonetheless a challenge the Frenchman will need to undertake. He has come a long way on the back of a big first serve, but to breach the top ten, he’ll need a more well-rounded attack.

* * *

Subscribe to the blog to receive each new post by email:

 

Monthly Roundup #1: January 2025

Indoor tennis in 1908, at the 7th Regiment Armory in Manhattan

Hey, look, another new thing! I’m aping TheZvi and ACX in an attempt to clear out my notebook and help you find the better tennis stuff out there.

It will be a mix of worthwhile links, historical tidbits, and trivia that doesn’t quite merit the full Notebook treatment. There will also be some self/site promotion, partly because it’s my blog and I can do what I want, and also because I regularly find that readers don’t know about features on the site.

As with the Trivia Notebooks, please send me things if you think they belong here.

1. The Match Charting Project has reached 15,000 matches. (And counting!) I still remember getting excited when we reached 100. Now we have 100 matches each for several dozen individual players, not to mention 500-plus for each of the Big Three. Thank you to all contributors.

2. If you’d rather read my analysis in Italian or Spanish, there’s an ever-growing body of translations, too. Edo published Italian versions of my latest pieces on Jannik Sinner and Madison Keys in the last few days, and Ángel recently translated my post from last year on Ugo Humbert and Surface Sensitivity into Spanish. (These two guys are adding to the Match Charting Project totals, too!)

3. New academic paper on serve strategies from Axel Anderson, Jeremy Rosen, John Rust, and Kin-Ping Wong:

Do the world’s best tennis pros play Nash equilibrium mixed strategies? We answer this question using data on serve-direction choices (to the receiver’s left, right, or body) from the Match Charting Project. Using a new approach, we test and reject a key implication of a mixed-strategy Nash equilibrium: that the probability of winning the service game is identical for all possible serve strategies. We calculate best-response serve strategies by dynamic programming (DP) and show that for most elite pro servers, the DP strategy significantly increases their win probability relative to the mixed strategies they actually use.

I haven’t read the paper, partly because I am not optimistic that I would understand it. Still, it is heartening to see a treatment along these lines find that players are not optimizing. Most of the ones I’ve seen tend to conclude that they are, or that they are close.

4. Gill noticed that Frances Tiafoe was the “leader” in aces-against, making him a fun matchup for Giovanni Mpetshi Perricard in Brisbane:

It’s a good reminder that (a) this is a real phenomenon, and (b) you can find it on the TA leaderboards. Tiafoe was indeed the vAce% leader in 2023 as well. His last-52-week rate is up to 13.5%, in part because two of his last five matches were against Mpetshi Perricard. Arthur Rinderknech posted a 20% ace rate against him in Melbourne, too.

When I say this is “real,” I mean that it’s persistent, not just the year-to-year fluctuations of luck. Some players are better at ace prevention than others. It’s not necessarily a good thing: Returners who get more balls in play win those points at a lower rate, so it can come out in the wash. That said, a remarkable example of the persistence of ace prevention is Gael Monfils. I wrote about the stat all the way back in 2012, and Monfils was then the least aceable player on tour. Thirteen years later, he’s still in the top three.

5. I love finding footage of historical players I’ve never seen before. Here are ten minutes of Julie Heldman (and Billie Jean King) in the 1974 US Open semi-final:

There’s a wealth of historical match footage, but what has survived is (understandably) heavily concentrated on the biggest names, like BJK. A couple more exceptions I’ve come across lately are this 1977 clip of Martina Navratilova against Marise Kruger, and (compare and contrast!) a full match from 1990 of Martina against Halle Cioffi.

6. Hugh gives us a great technical breakdown of Holger Rune:

[H]e’s leaving speed on the table in numerous areas. He could keep the left hand on the throat for longer like Sinner and Alcaraz to coil the torso more, he could lift the hitting elbow higher and invert the racquet without adding the left hand like Berdych, he could flex the elbow more in the setup like Verdasco, or he could simply have a higher takeback but keep the racquet tip up like del Potro. All these solutions add a little more length and would help provide more and smoother racquet speed.

And:

You can’t fire a cannon from a canoe.

7. RIP Rino Tommasi (1934-2025). The Italian journalist covered tennis (and boxing) for well over a half-century. He was also a passionate record-keeper. He is at least indirectly responsible for the fact that Italians are disproportionately represented in pretty much every historical and data-oriented tennis project.

8. Sad to see that we’ve lost Czech coach Jiri Fencl, as well. He was probably best known for his work with Lucie Hradecka, and he even wrote a bit about analytics, in connection with his Resultina app.

9. Giri asks, Has the Tennis Ball Gotten Worse? Tough to say:

As for the guys playing on TV, I wonder if some of the changes in game style they decry have just as much to do with factors beyond the balls, such as the advances in technique or strength and conditioning that enable the likes of Alcaraz and Sinner to hit as big as they do on every single shot. I also wonder if I’d helped to bring their malaise into existence by raising the topic of the ball at all.

It sounds like the ATP, at least, is moving toward more standardization, which is long overdue. However different, worse, or otherwise the balls are, they ought to be the same from one week to the next.

And:

Medvedev and Rublev both seemed to think that “tactics” were gone in men’s tennis, but for nearly opposite reasons: Everybody can rally forever, or nobody can rally for two shots.

I’ll probably have more to say about whether “tactics are dead.”

10. Speaking of footguns, Madison Keys won’t be able to play in next month’s Austin event, as planned:

“With World No. 6 Jessica Pegula already committed to the event, WTA rules prevent us from having a second Top 10 player in the draw. When we entered an agreement with Madison, her ranking was World No. 21. Now with her title wins in both Adelaide and Melbourne, her ranking has moved to World No. 7. As a result of her new ranking, Madison will, unfortunately, not be able to compete in this year’s ATX Open.”

My first reaction was not just that this was idiotic, but that it was nonsensical: Wasn’t the entry list for Austin finalized six weeks before the tourney started? But no, apparently WTA entry lists are settled four weeks before, in this case just after Keys’s ranking boost.

(Also, do I have this right? Had Keys leapfrogged Pegula in the rankings, with both remaining in the top ten, would Pegula have been the one forced out?)

The motivation for the rule is clear: Austin is a 250, and it runs concurrently with a 500 in Merida. The tour wants the best players to compete at the higher-stakes events and, more importantly, to face each other. In general, 250s are getting sidelined on both tours, and this is just one small instance of that. I’m not sure exactly how the rules should be rewritten, but it seems foolish to prevent an American star from appearing in the US in favor of an event elsewhere that is only technically more prominent.

11. James Gray explored Emma Raducanu and Elo rankings, with an assist from me:

But according to one statistical analysis, Raducanu has no reason to fear clashes with top players, and in fact is a good enough player to be seeded, if she can just become consistent enough.

The Elo rankings show she is in fact the 13th best player in the world right now.

And now 12th, after the Australian … and something worse than that on Monday, once her Singapore loss to Cristina Bucsa (Elo ranking: 138) goes on the books. This was not a good month to use Raducanu as an example of the value of Elo. So, of course, Simon Briggs covered it too.

12. Ahead of the 1919 Wimbledon final, former champion Blanche Hillyard wrote to Dorothea Lambert Chambers, who was set to face Suzanne Lenglen in the Challenge Round:

My rheumatism is too damnable for words. How on earth [Phyllis] Satterthwaite has got to the [all-comers’] final, I don’t know. One of the worst styles of players and I always feel I could have given her ½ 30 in my best days. Well, my dear, again good luck and I do hope you won’t have the ‘curse’. I wish she may have it if she does have it at all?

(“½ 30” referred to handicapping, which was considerably more common back then. A handicap of 15 meant letting the weaker player start every game at 15-0 on serve or 0-15 on return. “½ 30” was halfway between a handicap of 15 and 30, so one point in half of games, two points in the other half. Believe it or not, it gets immensely more complicated from there.)

Hillyard said she’d try to “Evil Eye” the Frenchwoman, but Lenglen won–10-8, 4-6, 9-7–in a match for the ages.

13. At the start of the season, I updated my datasets of rankings, results, and stats for both ATP and WTA. I also updated the dataset of Match Charting Project data.

14. Gill and Ben Ornstein send out weekly editions of The Draw, which is a great resource for finding tennis coverage worth reading. Better than this, for sure.

15. Michal argues that Carlos Alcaraz lost to Novak Djokovic because he was too predictable on second serves:

By completely avoiding the Djokovic forehand return, Alcaraz gives Djokovic an invitation to set up the point however he likes. Djokovic can either step in, and take his backhand return early – made easier by the fact that he can wait in his backhand grip and give up a third of the box. Or, he can back up to get a forehand return anyway – here he can start moving early, while Alcaraz’ ball toss is still in the air, because he knows that the serve will be aimed toward a particular area of the box.

This seems like something a lot of players would let you get away with. Djokovic is not one of those players.

16. Ana gives us the context we need to understand the student protests in Serbia that have prompted Djokovic to speak up:

First, violence against young people—particularly the mass shooting at a Belgrade elementary school (OOS “Vladislav Ribnikar”) in May 2023—has almost certainly contributed to Novak’s new outspokenness. Second, both his age and his parenthood have shaped his thinking: he wants his kids, now 7 and 10 years old, to “also grow up in Serbia,” like he did, but it’s not certain that they will—not merely because they have the resources to live elsewhere but specifically because Serbia may not provide what they need to thrive. Third, while the exodus from Serbia over recent decades (essentially, since the breakup of Yugoslavia) saddens him, he understands it.

17. I probably could have done an entire Trivia Notebook on Gael Monfils’s title in Auckland. He’s awfully old to be winning titles, he’s won 9 of 15 finals (and six of his last seven) after starting his tour-level career 4-16, he beat top-50 opponents in the first two rounds but none in the final three … it was a week of oddities.

The weirdest bit of all how La Monf beat Nishesh Basavareddy in the semi-finals. The score was 7-6(5), 6-4, despite the American outpointing him by one. Most striking is the dominance ratio (DR), the ratio of Monfils’s return points won to his opponent’s return points won. That number was 0.75, low for any victory, let alone a straight-setter. Last year, the lowest DR for a straight-set win was 0.79 (Halys d. Gasquet in Gstaad), and that came from a pair of tiebreaks.

The trick here is that while Basavareddy won a higher rate on return (34% to Gael’s 25%), he had to play a ton of return points: 88 to Gael’s 59. He kept return games close, but couldn’t bunch enough points together to earn a single break. Monfils earned fewer break points, but he converted one.

18. Ravi Ubha asks, What is an Unforced Error?

…Mitchell pointed to probability.

“The speed of the ball coming in, the height of the ball, those things also impact on whether it’s going to be an unforced error or a forced error,” said Mitchell, who estimates he has charted 7500 matches over 30 years. “If it’s a short ball and it is missed, does the player have the ability to make the shot? And if they had the ability to make the shot and missed it, then we can label it as an unforced error.”

Every tennis fan goes through a phase where they discover the limitations of the unforced error stat. (Ravi knows his readership!) Unforced/forced is a simplistic way of looking at a point outcome, and scorers will not always agree.

Probability is the key word here. In a better world, those of us who cared would be able to calculate the chances that a player would miss any shot. Automated systems like Playsight have been doing that for years, even if you can’t see what’s going on internally. The charitable interpretation of “unforced error” is something like, “a shot that a typical tour player would have made more than 60% of the time.” Put in those terms, you might want better stats, but there isn’t anything inherently wrong with what we’ve got.

19. Ben Rothenberg covers the anti-Zverev protest at the Australian Open final:

An aggressive security guard, however, came between us and told me to back away. He told me to stop typing on my phone, and then repeatedly asked me to move further and further away from this woman, continually telling me to move further back meter by meter. He asked me repeatedly to leave the entrance area and return to the seating bowl, but I wanted to keep the woman in my eyeline until I knew what her fate would be. He also photographed my accreditation as well as the accreditation of another journalist who arrived a few minutes later, and reported my presence there to Tennis Australia.

Gross. Not the first time Tennis Australia has overreacted to protest or tried to halt media coverage of it. Apparently it’s only the Happy Slam if you stay in an increasingly narrow lane.

20. Non-tennis, but still sports: Michael Crawley’s book Out of Thin Air, about his time running with elites in Ethiopia, is fantastic. I learned a ton. Here’s an excerpt:

Throughout my time in Ethiopia, in fact, I never heard anyone mention “talent” or “natural ability.” Runners use the word lememed to refer to training as a runner, which literally means “adaptation” or getting used to something. Runners are either good at managing the process of adaptation or they are not. A good runner is most likely to be described as gobez, which means some kind of combination of cleverness and cunning, denoting an ability to plan and manage their training well. As Meseret, a local coach, frequently puts it to the runners, “You can be changed.”

21. I’ll leave you with this 1974 performance of Swing Low, Sweet Chariot from Sonny Rollins and Rufus Harley, the world’s grooviest bagpiper:

* * *

Subscribe to the blog to receive each new post by email:

 

The Locked-Down Serve of Jannik Sinner

It’s all business for Jannik Sinner, and business is good.

There are a lot of things we could talk about after Jannik Sinner’s latest display of dominance. With a second Australian Open title, his exceptional span of hard court performance has stretched to 13 months, and according to my Elo ratings, only eight players in the Open era have ever been better than the Italian is right now.

A year ago, I wrote that Yes, Jannik Sinner Is This Good. If anything, he has improved since then. He holds serve more often than anyone on tour–yes, even more than Giovanni Mpetshi Perricard. Yet he is far from one-dimensional. He couples the serve with elite groundstrokes on both wings. He breaks serve more often than all but six players on the circuit.

One particular aspect of Sunday’s victory stands out. In three sets, spanning 15 service games, Sinner did not face a single break point.

Zero break points faced is one of those stats that sounds really good–then it gets even better the closer you look. Alexander Zverev a top-ten returner. He breaks nearly one-quarter of his opponents’ service games. He earns approximately 0.6 break points per game, or three per set. In three sets, the Italian didn’t allow him one.

Every season there are a couple hundred matches in which the winner doesn’t face a break point. Yet most of those are short best-of-threes, many of them involving overmatched wild cards and qualifiers. In 2024, there were only five matches in which a top-ten player failed to generate a single break point over more than ten return games. Twice the victim was Hubert Hurkacz, probably the weakest returner among the elites. Another hapless outing came from Casper Ruud, who was held to zero by Zverev at the Tour Finals.

The other two standout matches belong to Sinner. In last year’s Australian Open semi-final, then again in the Shanghai title match, he shut down Novak Djokovic. The only recent precedents for Sinner’s plastering of Zverev are the Italian’s own performances.

What can we learn from the latest episode of the Sinner show?

Situational awareness

Twelve months ago, Sinner was developing a reputation as an escape artist. He beat Daniil Medvedev in a five-set Melbourne final, winning just one more point than the Russian. He snuck past Djokovic twice in November 2023, once winning exactly the same number of points, a second time despite winning four points fewer.

Sinner had a knack for erasing break points. Most players are less successful facing break point than on other service points, because stronger returners generate more break points. (Also, break points tend to crop up amid rough patches, at least for players human enough to occasionally slump.) The Italian, though, saved break points at a better clip than his other service points. He hit bigger at those moments, and it worked.

Recognizing that trend a year ago, I tempered the celebration with a dose of reality:

…I can tell you what usually happens after a season of break-point overperformance: It doesn’t last. Taking over 2,600 player-seasons since 1991, 582 (21.7%) of players saved more break points than they won serve points overall. 183 (6.8%) matched Sinner’s mark of saving at least two percentage points more than their serve-points-won rate.

That sound you hear? That’s Darren Cahill laughing at us. Sinner didn’t quite continue at the same pace, but he still serves better facing break point than otherwise. He wins a tour-leading 71.5% of serve points overall, and the number climbs to 72.5% when an opponent has him on the ropes.

(71.5% is circuit-best, but 72.5% is not: Ben Shelton stands at 73.1%. No one told Jannik, though. In their semi-final, Sinner won 6 of 13 break points against the Shelton serve.)

Of course, this specific skill didn’t come into play on Sunday. Sinner didn’t have break-point-faced results, because he didn’t face any break points. If you have an even bigger weapon to deploy at key moments, why wait until it’s absolutely necessary?

Beyond the escape room

Some players–especially left-handers–are more effective serving to the ad court than the deuce court. That helps them save break points, because most break points (30-40 and ad-out) take place in the ad court.

Sinner is right-handed. Across the 250-plus matches for which I have sequential point-by-point data, he is slightly more effective serving to the deuce court:

COURT    A%   SPW%  
Deuce  7.6%  66.6%  
Ad     6.5%  66.3%

The different in points won isn’t anything to build a strategy around, but the ace-rate gap suggests there might be a meaningful difference.

In Melbourne, the success rates took on a new look. Based on the five Australian Open matches in the Match Charting Project database, here are the same metrics:

COURT    A%   SPW%  
Deuce  8.9%  74.6%  
Ad     9.4%  70.8%

The ace rates flipped, but holy four percentage points! 70.8% serve points won is very good, better than anyone else on tour over the last 52 weeks. 74.6%, though: That’s from another planet. It’s better than John Isner’s best season.

It may be a fluke. After all, it’s just five matches, and Sinner’s deuce- and ad-court results were nearly identical in 2024. But I suspect there’s strategy at work here.

Up to eleven

I noted last year that Sinner served bigger facing break point than he did otherwise: 125 miles per hour compared to 122. What about other key situations?

At the US Open last year, the Italian showed some deuce-court preference, winning 72.7% of deuce-court points compared to 71.3% on the ad side. It wasn’t a matter of power, as he averaged almost identical speeds (117.9 to 117.7 mph) in the two directions.

Speed differences turn up at a more granular level. Here are Sinner’s first-serve speeds at the most common point scores he faced:

Score    MPH  
15-30  121.0  
40-40  120.3  
40-15  120.0  
0-15   119.7  
30-30  119.4  
40-30  118.0  
30-15  117.7  
30-0   117.0  
15-0   116.6  
0-0    116.6  
15-15  115.8  
40-0   113.0

(Yes, I know it’d be nice to have 30-40, 40-AD, 15-40, and so on. But this is Sinner we’re talking about. He didn’t face many of those.)

With the exception of 40-15, this list is an awfully good approximation of point scores listed by importance. With more at stake, the Italian hit harder. There’s no apparent trade-off, either. He made more first serves than average at 15-30 and deuce, even with the faster strikes.

The point is that when Sinner feels the need to go big, he has the ability to do so. Most players don’t: Their results don’t get better in critical moments, either because they’re already maxing out their skills on routine points, or because their opponents can raise their levels, as well. But the Australian Open champ has more in the tank than anybody who dares to stand across the net from him.

As the Italian rose through the rankings, he saved that extra oomph for break point. Now that he enters every match as the favorite, he can be even more aggressive. By serving harder at 15-30, or 30-all, Sinner probably doesn’t risk overexposure. With the possible exception of Alcaraz or a time-traveling Djokovic, no one can do much with an accelerated Sinner first serve–even if they’ve seen one recently.

The full arsenal

Put all of this together, and it’s hard to avoid the conclusion that Sinner will continue to climb the all-time list. In terms of peak level, Andy Murray is next:

Rank  Player          Peak Elo  
1     Bjorn Borg          2473  
2     Novak Djokovic      2470  
3     John McEnroe        2442  
4     Ivan Lendl          2402  
5     Roger Federer       2382  
6     Rafael Nadal        2370  
7     Jimmy Connors       2364  
8     Andy Murray         2347  
9     Jannik Sinner       2325  
10    Boris Becker        2320

Sinner wins more service points than anyone else in the game–even when he isn’t particularly trying. At the same time, he is an elite returner.

We’ve seen a few players with a similar service profile. Pete Sampras also had a knack for coming through at the end of sets. But he leaned on his tiebreak skill more than Sinner ever will, because his return game was mediocre. Roger Federer was better: a dominant–and clutch–server, and a more competitive returner than Pistol Pete. Nick Kyrgios had one of the most electric serves in the game’s history, coupled with the ability to focus at critical moments. Yet he was the most one-dimensional of the bunch. His return would have held him back even if he had stayed healthy.

Sinner has all the serve dominance without the drawbacks. He has no apparent weakness. He won’t win all the time, and he won’t stay on top forever, but at the moment, it’s hard to see how anyone will stop him.

* * *

Subscribe to the blog to receive each new post by email:

 

Trivia Notebook #2: Slam-Winning Nations and Fifth-Set Momentum

Lorenzo Musetti’s backhand is powerful, but it’s getting lonely.

Previous: Trivia Notebook #1

Two more weeks of tennis, another two weeks of trivia. I’ll get to a fraction of it, anyway. In this installment, we’ll look at countries with grand slam match victories, fifth-set momentum, and another sign that the one-handed backhand is dying.

Lebanon and countries with major wins

At the start of the Australian Open fortnight, Hady Habib made history by becoming the first Lebanese player to win a match at a grand slam. I have many questions. How many countries do have wins at majors? The number of countries represented on the women’s side is probably less than the men’s right? And with Lebanon off the board, who’s next?

First off, a clarification: Habib is not the first Lebanese player to win a match at a slam. He’s the first in the Open era. Nadim Hajjar reached the second round at Roland Garros in 1956 before losing to Jozsef Asboth. (For a great Asboth story, see my Roy Emerson essay from the Tennis 128.) That is not to detract from Habib’s accomplishment: He’s the first Lebanese man to appear in a slam draw since 1962.

Let’s focus on the Open era, then. I’m not going to get bogged down in definitions, except to acknowledge that you easily could. No two sources are going to agree on how many countries exist. Conventionally, tennis follows the International Olympic Committee: For example, Puerto Rico is treated as an independent entity, but Scotland is not. There are also plenty of tennis-playing countries with slam victories that have changed their names (e.g. Rhodesia) or ceased to exist (e.g. Yugoslavia). This is all a database-keeper’s nightmare, too. Lebanon has gone through three different IOC codes, and I realized in the course of this research that I’m using different ones for Habib and his countryman Benjamin Hassan. Oops.

Before Lebanon joined the list, 78 present-day countries had scored men’s main-draw slam wins. So Habib makes it 79. Another seven nations appeared in at least one main-draw match but failed to win any.

As I guessed, the numbers are a bit smaller for women. 66 countries have gained a main-draw slam win, and another six have made an appearance. Those figures will soon shift to 67 and 5, respectively, thanks to newly-aligned Armenian Elina Avanesyan. She has gone 0-2 in her first two majors sporting her current flag, but it’s just a matter of time until the top-50 player gets back in the major win column.

Aside from Avanesyan, who’s on deck?

Here are the top-ranked men who would be the first from their countries to win a slam main-draw match:

Rank  Player                  Country                   
171   Coleman Wong            Hong Kong                 
250   Abedallah Shelbayh      Jordan                    
332   Hazem Naw               Syria                     
336   Eliakim Coulibaly       Ivory Coast               
619   Nam Hoang Ly            Vietnam                   
623   Mitsuki Wei Kang Leong  Malaysia                  
694   Samir Hamza Reguig      Algeria                   
731   Colin Sinclair          Northern Mariana Islands  
852   Jesse Flores            Costa Rica                
887   Seydina Andre           Senegal                   
932   Petar Jovanovic         Montenegro

Wong is only 20 years old, so he’s on track to add Hong Kong to the Open-era list. (The city-state tallied three victories in the amateur era.) Shelbayh is only seven months older, so he’s in a good spot as well. The others are considerably more speculative.

The women’s list is deeper. The eleven men listed above are the only ones in the ATP top 1000 from a country without a slam match win. There are 23 women who meet the same standard on the WTA table. Here are the first 13 of them:

Rank  Player                       Country        
136   Alexandra Eala               Philippines    
151   Victoria Jimenez Kasintseva  Andorra        
160   Kathinka von Deichmann       Liechtenstein  
201   Raluka Serban                Cyprus         
251   Justina Mikulskyte           Lithuania      
281   Lina Gjorcheska              Macedonia      
341   Sada Nahimana                Burundi        
404   Malene Helgo                 Norway         
433   Francesca Curmi              Malta          
492   Klaudija Bubelyte            Lithuania      
552   Noelia Zeballos              Bolivia        
620   Aya El Aouni                 Morocco        
678   Angella Okutoyi              Kenya          
687   Patricija Paukstyte          Lithuania

Eala and VJK, both still teenagers, seem like all-but-sure bets. Von Deichmann has been knocking at the door for years. Lithuania has three representatives on this list, so it’s probably just matter of time–though probably more time than it will take Eala or VJK.

Fifth-set momentum

This one is from Eric J via email:

Inspired by Learner Tien’s 5-set win over Medvedev (and his post-match interview in which he confessed to giving up on the fourth set so he could get a bathroom break): Is there a discernible trend in matches that last 5 sets as to whether the player who wins the fourth set when down 2-1 is more likely to win the fifth than his opponent?

Yes, there is. There are three ways to go down 2-1 in sets: lose the first two (LLW), lose the first and third (LWL), or lose the second and third (WLL). At majors in the Open era, for all of those permutations, players who come back to win the fourth have a better-than-neutral chance and taking the match:

Sequence   Win-Loss   Win%  
LWLW        522-395  56.9%  
WLLW        541-455  54.3%  
LLWW        637-490  56.5%  
Total     1700-1340  55.9%

The edge has gotten a bit smaller in recent years. Here are the same numbers for slams since 2000:

Sequence  Win-Loss   Win%  
LWLW       142-105  57.5%  
WLLW       168-144  53.8%  
LLWW       182-158  53.5%  
Total      492-407  54.7%

Not a big difference, and I don’t want to read too much into it. But I’m still surprised the shift is in that direction. Tanking–usually to save energy, not to race for a toilet break–used to be more common, even standard advice. If the effect is real, it might be that players are generally more fit than they used to be. A half-century ago, a player who lost the fourth set might be gassed, with little hope of recovery. That’s less often the case now.

Either way, there’s a small momentum edge for the fourth-set winner. It’s possible that controlling for matchups would change the results–perhaps stronger players are more likely to be the ones who come back from a slow start to force a fifth set.

One one-hander

Only a few men remain on tour with one-handed backhands, and they combined for a bad fortnight. Stefanos Tsitsipas and Giovanni Mpetshi Perricard lost in the first round. Grigor Dimitrov was forced to retire. That left Lorenzo Musetti:

The decline has been remarkably steady. I have backhand types for all 3rd-round players at slams back to 2000. In that time, we’ve gone from one-handed backhands for a full half (16 of 32) of third-rounders at Roland Garros in 2000 and 2002, down to one in Melbourne this year.

This graph shows the number of one-handers in the third round at each of the last 100 majors:

This year’s Aussie is indeed the first major with just one one-hander in the third round. We’ve been getting close for years, with two at Wimbledon in 2019 (Federer and Dan Evans), at Roland Garros in 2022 (Dimitrov and Tsitsipas), and at last year’s Australian (again, Dimitrov and Tsitsipas).

Oddly enough, things swung all the way back up to seven at Wimbledon in 2023, with the surprise runs of Christopher Eubanks and Christopher O’Connell, alongside stalwarts Dimitrov, Musetti, Shapovalov, Tsitsipas, and Stan Wawrinka.

Musetti lost his third-round match to Ben Shelton, so the round of 16 was a two-hander-only club. At least that’s not a first: The same thing happened at the US Open in both 2022 and 2023.

That’ll do it for today. Keep the suggestions coming, and we’ll do some more trivia in a few weeks.

* * *

Subscribe to the blog to receive each new post by email:

 

Elina Svitolina Plays Hard

Elina Svitolina at the 2020 Australian Open. Credit: Rob Keating

Here’s a stat to get us rolling: In yesterday’s fourth round match at the the Australian Open, Elina Svitolina won 67% of points when she hit a backhand. Tour average is a neutral-as-its-gets 50%. Svitolina, with her resourcefulness off that wing, averages 51%.

It gets even better. Usually, if a rally statistic comes out much above 50%, it’s because it’s loaded with winners. But against Veronika Kudermetova yesterday, Svitolina hit just one backhand groundstroke winner. She induced two more forced errors, but committed five unforced errors of her own. All of which is to say, those point-winning backhands came thanks to point construction, not swinging for the fences. (An erratic Kudermetova helped, too.)

The backhand masterclass, on the heels of another strong baseline performance against Jasmine Paolini in the third round, recalls Svitolina’s peak. The Ukrainian is now 30 years old, veteran of innumerable injuries, a pregnancy, and the perpetual distraction of her country at war. My Elo ratings suggest that she was at her best nearly seven years ago, after she upset Simona Halep for the the 2018 Rome crown. That’s an eon in tennis time: She hasn’t held a place in the top ten since 2021.

Yet here she is, in the Melbourne final eight. It’s her twelfth major quarter-final, her fourth since becoming a mother. She might have made it one more a year ago. After a grueling runner-up finish in Auckland to open her 2024 campaign, Svitolina raced through three rounds at the Australian before retiring to Linda Noskova in the fourth round. This year, she skipped the warmups and has sustained her best tennis on the bigger stage. With Madison Keys across the net tomorrow, she has a chance to go even further.

Svitolina 2.0 will probably never reach the level she showed in the mid-2010s. She’s a half-step slower, and she needs to manage her schedule with care. But like all players with successful second chapters, she has changed her game in response to her limitations. She is, and always will be, a counterpuncher, leaning on one of the game’s sturdiest backhands. Yet the new Elina has first-strike weapons that her younger self could only dream of.

Hitting big

While no one is about to mistake her for Aryna Sabalenka at the service line, the 30-year-old does more damage than she used to.

Yesterday against Kudermetova, Svitolina averaged almost 103 miles per hour (165 kph) on her first serves. I have first-serve speed for more than 70 of her career slam matches, and she hit that level in only five of them, mostly at Wimbledon. When she reached the quarters at the 2021 US Open, she averaged less than 100 miles per hour in all five matches.

It’s a small improvement, but coupled with increased precision on the first serve, it is paying off. In the sample of Match Charting Project-logged matches, she converted more plus-ones behind her first serve in 2024 than her career rates. Even her improved 2024 marks pale in comparison to what she has done in Melbourne:

Span               Unret%  <=3 W%  
Career              27.2%   38.6%  
2024                27.5%   41.9%  
R3 vs Paolini       36.1%   60.6%  
R4 vs Kudermetova   39.4%   63.9%

The second column, showing the percent of first-serve points in which Svitolina won the point with her serve or second shot, is where players make their money. No matter how good your ground game, it's tough to make up a deficit in the cheap-points category. Through 2024, the Ukrainian was middle-of-the-pack in both of the stats. The form she has shown in the last few days is something entirely different.

The numbers are particularly impressive against a defender like Paolini. While the Italian probably isn't as strong as her #4 ranking suggests, she is certainly an elite returner. In charted matches last year, she put nearly three-quarters of first serves back in play. Svitolina allowed her only 64%. And as we see with the serve-or-second-shot stat, Paolini couldn't do much when she did get the ball back. On average, the Italian wins more than half of the first-serve points in which she lands her return. On Friday, she won just 6 of 23.

There's just one reservation about the third-round performance. Svitolina got those results by taking chances, and she made fewer than half of her first serves. It was a day of extremes: 48% of first serves in, 83% of first-serve points won, and 42% of second-serve points won. Had she explicitly targeted a more typical 60% rate of first serves in, she wouldn't have posted the same win rates. But with a first-class returner across the net, Svitolina's tactics were proven correct.

That half step

So far we've talked about what the Ukrainian can control. Just as important is what she can't: The aging process and its effect on the rest of her game.

Here's an overview of how her current level compares to her 2017-20 peak, measured by first-serve and second-serve win percentages, along with return points won:

Span     1st W%  2nd W%   RPW%  
2017-20   66.6%   47.2%  46.1%  
2023-25   65.5%   46.8%  45.2%

Approximately a one-percentage-point drop across the board. That makes sense as an explanation of the difference between a player ranked around #5 and one who should be hovering around #20. (Elo is more optimistic than Elina's official ranking of #27.)

Now try the same stats, hard courts only:

Span     1st W%  2nd W%   RPW%  
2017-20   67.5%   47.3%  46.1%  
2023-25   68.9%   47.9%  43.6%

At her peak, Svitolina was basically the same player on all surfaces. Now, she sports a different type of surface profile. The service aggression is paying off, while she seems to be struggling to do return damage on faster courts.

The 30-year-old has always aimed to get a lot of serves back. According to MCP data, she has put 77.5% of serves in play, a number that fell to 76.4% last season. Both marks are near the top of the table. Players who adopt such a defensive posture usually don't win a particularly high rate of those points, and Svitolina is no exception: Her 51.5% win rate when she puts the return in play is roughly tour average. It's a low-risk, fairly-low-reward strategy.

Against strong servers, the results can be bleak. She got fewer than 65% of serves back against Sabalenka in Cincinnati last summer, and even Ons Jabeur held her to 67% at Wimbledon. In Adelaide three years ago, Madison Keys was so strong from the line that Svitolina put only half of returns in play. At her peak, it was rare for the Ukrainian to fall below 70%.

Her hard-court results, then, depend a great deal on the matchup. The relatively punchless Paolini was a lucky draw in this regard. Once the serve and the plus-one are past, Svitolina can go toe-to-toe with anybody. She won 60% of points that lasted four or more strokes against the Italian. Facing Kudermetova, as we've seen, it was even easier. Once Elina got her racket on a backhand, the Russian basically gave up.

It's a mad, mad quarter-final

Svitolina's next challenge is entirely different. In five career meetings, she has two victories against Keys, coming at the Australian Open and US Open in 2019. Even when she was younger, she couldn't neutralize the American's serve. She has won about 36.5% of return points in their head-to-head, never topping 43.3% in a single match.

The margins for the Ukrainian tomorrow will be slim. While Svitolina has boosted her serve, she has gained more plus-one points than unreturned serves. That works against opponents like Paolini, but Keys swings big at everything, including hard serves. In the Adelaide final against Jessica Pegula--a broadly similar player to Svitolina--Keys held her opponent's serve points to an average of 3.1 strokes. That's a lot of short "rallies," and it means fewer chances for Elina to put away a weak second ball.

Svitolina will find herself even more powerless on return. As we've seen, Keys is responsible for one of the worst performances of her career. In that 2022 Adelaide match, the 30-year-old won a miserable 22% of return points. The longer the rally, the better for Svitolina. But Keys will try to prevent the commentators from using the word "rally" at all. In the Adelaide title match, the American's service points lasted, on average, a mere 2.6 strokes.

The Ukrainian may not have much control over the proceedings, but that isn't to say she doesn't have a chance. My forecast leaves her plenty of room, giving her a 42.5% shot at reaching the final four. Keys's aggression often goes astray, and nagging injuries could hamper either player. If the American can't serve at 100%, or if she falls back on more passive tactics, the underdog will pounce. In longer points, Svitolina is the clear favorite. She'll have to hope she gets the chance to play some.

* * *

Subscribe to the blog to receive each new post by email:

 

Matteo Berrettini and Best-of-Five Puzzles

Matteo Berrettini is a man you don’t want to face in best-of-five. (Unless you’re Holger Rune, in which case you beat him in four sets yesterday.)

Some players seem to play better in best-of-five matches. Maybe they are fitter than average, or they take time to get into the rhythm of a match, or they are particularly good at managing their preparation to peak for grand slams. There are many plausible explanations. I suspect most fans assume that there’s some kind of “best-of-five” factor that makes certain players better or worse than usual at the majors.

I took a first crack at the question during the 2014 Australian Open. Then, Jo-Wilfried Tsonga stood out as a best-of-five specialist. With more data and better tools, it’s time to try again. Gill Gross and Alex Gruskin teed it up:

Challenge accepted!

Note that the stat here is awfully specific: hard-court matches since 2019. Matteo also lost to Holger Rune in the second round yesterday, so the numbers have slightly changed. We’ll come back to this particular puzzle in a bit.

First, it’s important to remember that many players should have better records in best-of-five than in best-of-three. It’s the same reason that in other sports, best-of-five (or best-of-seven) series are more likely to go to the favorite than more luck-bound best-of-threes. Given more sets, players have more time to turn things around. The stronger of the two competitors is more likely to do so.

A fantastic illustration of this comes, appropriately enough, from the (all-surface) career numbers of Berrettini himself:

Format      Set%   Win%  
Best of 5  62.1%  68.9%  
Best of 3  62.0%  63.8%

The Italian wins sets at almost exactly the same rate, regardless of format. But the match win percentage is different. There’s still something to be explained: A player who wins 63.8% of best-of-three matches should, all else equal, win approximately 67% of best-of-five matches. Matteo has done better than that, but the format alone explains much of the gap.

Better at best of five?

Identifying players who outperform expectations requires that we define “expectations.” As usual, Elo makes it easy to do this. For every individual match, we can use the Elo ratings of the two men to generate probabilities that each will win.

For Berrettini at hard-court slams since 2019–excluding retirements and the 2025 Australian Open–I have him at 27-9, good for a 75% win rate. Based on pre-match Elo ratings, though, he “should” have gone just 22-14 (technically, 22.5-13.5), taking 62% of the contests.

That’s noteworthy but not astonishing: A player expected to win 62% of the time has about an 8% chance of winning at least 27 of 36 matches. Given the number of ATP tour regulars, it stands to reason that somebody would post a stat like that. We can’t cast aside the Italian’s case yet, because we haven’t talked about his best-of-three results. Again, I’m going to kick it down the page because I want to show you some other numbers first.

Before looking at the narrow set of 2019-24 hard-court numbers, let’s see how everybody fared at grand slams, on all surfaces, since 2000. Out of 154 men with at least 50 grand slam matches this century, here are the top dozen overperformers:

Player               W-L     W%   Exp%  Ratio  
Pablo Andujar      23-39  37.1%  27.9%   1.33  
Denis Istomin      34-41  45.3%  34.7%   1.31  
Frances Tiafoe     45-34  57.0%  46.2%   1.23  
Mario Ancic        40-20  66.7%  54.4%   1.23  
Victor Hanescu     29-35  45.3%  38.6%   1.17  
Karen Khachanov    59-30  66.3%  56.6%   1.17  
Simone Bolelli     25-32  43.9%  37.6%   1.17  
Leonardo Mayer     33-38  46.5%  40.1%   1.16  
Marat Safin        79-32  71.2%  61.5%   1.16  
Nick Kyrgios       54-28  65.9%  57.0%   1.16  
Bernard Tomic      40-35  53.3%  46.7%   1.14  
Matteo Berrettini  49-20  71.0%  62.6%   1.13

The first percentage is actual win percentage, followed by expected win rate (based on Elo ratings). The ‘Ratio’ column is simply the ratio of actual to expected. These are the guys who have played better at slams than their track records would have implied.

This ratio starts to identify overperformers, but we can go one step further. Sample size really counts here. It’s one thing to win seven of ten matches when you’re expected to win five. It’s wildly different to win 70 of 100 when you’re expected to win 50. The odds of the first are 17%, while the chances of the second are a fraction of one percent.

Since this next metric accounts for sample size, I’ve expanded our view to the 334 men with at least 20 slam matches since 2000. Here are the twenty players who have most defied the odds with their overperformance in best-of-five:

Player                   Record     W%   Exp%  Ratio  Odds  
Novak Djokovic           364-45  89.0%  84.3%   1.06  0.4%  
Rafael Nadal             304-41  88.1%  82.9%   1.06  0.5%  
Tennys Sandgren           16-17  48.5%  27.6%   1.76  0.9%  
Marin Cilic              133-56  70.4%  62.4%   1.13  1.4%  
Stan Wawrinka            151-66  69.6%  62.4%   1.12  1.6%  
Marat Safin               79-32  71.2%  61.5%   1.16  2.2%  
Frances Tiafoe            45-34  57.0%  46.2%   1.23  3.5%  
Mario Ancic               40-20  66.7%  54.4%   1.23  3.6%  
Denis Istomin             34-41  45.3%  34.7%   1.31  3.7%  
Jo-Wilfried Tsonga       120-43  73.6%  66.8%   1.10  3.7%  
Karen Khachanov           59-30  66.3%  56.6%   1.17  4.0%  
Carlos Alcaraz            57-10  85.1%  76.5%   1.11  6.1%  
Nick Kyrgios              54-28  65.9%  57.0%   1.16  6.4%  
Tomas Martin Etcheverry   12-12  50.0%  33.1%   1.51  6.4%  
Lukasz Kubot              20-20  50.0%  37.3%   1.34  6.9%  
Pablo Andujar             23-39  37.1%  27.9%   1.33  7.3%  
Andrey Kuznetsov          18-21  46.2%  33.8%   1.37  7.4%  
Thomas Fabbiano           10-13  43.5%  27.7%   1.57  7.6%  
Matteo Berrettini         49-20  71.0%  62.6%   1.13  9.2%  
Joachim Johansson          15-8  65.2%  49.4%   1.32  9.5%

I’ll admit it, I mostly lowered the match minimum so that we could have a top five consisting of four slam winners and one Tennys Sandgren. Djokovic and Nadal don’t stand out in the “Ratio” category: They were expected to wins lots of matches, and they did. But not only that, they slightly exceeded expectations for a very, very long time. There’s only a one-in-two-hundred chance that a player expected to win 83% of matches would win 88% over such a long stretch.

Enter the skeptic

Even highlighting these outlier performances–many of them in the hands of players we’d expect to see on the list–it’s not clear whether there’s really a best-of-five factor. As noted, we’re working with a population of over 300 players. Three of them gave us one-in-one-hundred performances. Fewer than 10% turned in one-in-ten performances. Isn’t that what we’d expect?

This isn’t a laboratory: We can’t run tests on Novak Djokovic to see if he would keep winning at the same rate in his next 410 best-of-five matches. We certainly can’t do it 100 times to be sure. We can, however, wring a bit more from the data we have.

If there is a special best-of-five skill–above and beyond a player’s general tennis ability–we’d expect players to show it with some consistency. (If they didn’t, could we call it a skill?) Here are two tests to check whether it’s a skill:

  1. Career halves: Split each player’s list of best-of-five matches into halves. Tommy Paul, for instance, went 13-13 in his first 26 best-of-five matches–worse than expected. Since then, he’s won 19 of 27–better than expected. If there’s a best-of-five skill, we’d expect those numbers to be persistent. Sometimes they are, but in general, they are not. Statistically, there’s virtually zero correlation.
  2. Odd and even matches: Maybe career halves are the wrong way to do it: Players improve and tendencies change with age. Instead, take each player’s list of best-of-fives and put them in two buckets, one for the first, third, fifth, etc. matches on the list, the other for the second, fourth, etc. Different tack, same results: no reliable relationship.

To be clear, this doesn’t tell us that everyone’s results are a luck-driven mirage, or that no one has any noteworthy best-of-five skill. Across 350 or 400 matches, I’d bet that Djokovic and Nadal probably do. (Heuristic: If a trait is good, they probably have it.) But in general, a player who is winning more best-of-fives than expected is probably due for a correction. There’s no basis to expect the trend to continue.

The Berrettini double

With that bucket of cold water thrown on our dreams, let’s return to the head-scratcher we started with.

Hard-court matches since 2019. Here are the best-of-five overperformers, minimum 20 slam matches:

Player               Record     W%   Exp%  Ratio   Odds  
Frances Tiafoe        27-12  69.2%  49.6%   1.40   1.0%  
Matteo Berrettini      27-9  75.0%  62.4%   1.20   8.1%  
Adrian Mannarino      15-12  55.6%  41.1%   1.35   9.3%  
Alexei Popyrin        13-11  54.2%  39.9%   1.36  11.2%  
Taylor Fritz          27-12  69.2%  58.6%   1.18  11.6%  
Novak Djokovic         52-4  92.9%  87.2%   1.07  13.9%  
Rafael Nadal           31-5  86.1%  79.7%   1.08  23.4%  
Daniil Medvedev       56-11  83.6%  79.5%   1.05  25.6%  
Pablo Carreno Busta    21-9  70.0%  63.1%   1.11  27.9%  
Marin Cilic            17-8  68.0%  60.9%   1.12  30.7%

Holy Tiafoe! Elo would have predicted a 50% win rate, and instead he went 27-12. Berrettini comes next, but by this metric, it’s a distant second. Only Tiafoe really stands out in this sample.

But wait–there’s more to the Gross/Gruskin puzzle. The Italian has not only overperformed at slams, he has notably underperformed on hard courts elsewhere. Excluding Challengers, retirements, and Davis Cup, Berrettini’s record is even more mediocre than the one listed above: It works out to 42-42. Elo would have predicted a 58% win percentage, not a mere break-even rate.

Of the 35 players with at least 20 hard-court slam matches in this span, only David Goffin more severely underperformed in best-of-three. Only Goffin, Jannik Sinner, and Gael Monfils posted more unexpected numbers in best-of-three. Berrettini is as odd in best-of-three as he is in the longer format, just in the opposite direction.

Using the “Ratio” numbers, we can compare best-of-five over- (or under-) performance with best-of-three, for a kind of “super-ratio.” While Matteo is unique is the unexpectedness of his two numbers, Tiafoe still comes out ahead:

Player             bo5 Ratio  bo3 Ratio  bo5/bo3  
Frances Tiafoe          1.40       0.98     1.42  
Matteo Berrettini       1.20       0.86     1.40  
Adrian Mannarino        1.35       0.98     1.38  
Alexei Popyrin          1.36       1.07     1.27  
Marin Cilic             1.12       0.91     1.23  
David Goffin            1.04       0.86     1.21  
Daniel Evans            1.12       0.97     1.15  
Dominic Thiem           1.10       0.97     1.13  
Davidovich Fokina       1.14       1.01     1.13  
Taylor Fritz            1.18       1.05     1.13

The odds that Berrettini would give us such an unusual pair of stats are 0.3915%. Tiafoe’s number is 0.3969%. Let’s call it a tie.

Are we there yet?

After all this, I’m not sure that I’ve “explained” what’s going on here, per Gruskin’s request. We’ve seen that where best-of-five results differ from a player’s overall results, it’s mostly luck. I assume it’s the same with best-of-three. Maybe there are some additional factors in Berrettini’s case: Perhaps he’s more likely to play non-slams when he’s physically less than 100%.

There’s also this:

Tiebreak records are definitely luck-driven. These splits account for much of the difference in Matteo’s match-level results. A few points here or there, and we wouldn’t be having this conversation. Or, more likely, we’d be overreacting to unexpected numbers from somebody else.

Poor Hubi

One last thought. We’ve looked only at overperformers so far. Of course, there will always be underperformers as well. Hubert Hurkacz has disappointed a bit at slams: 34-25 before the Australian Open, compared to an Elo-expected 37-22. The subset of hard-court slams since 2019, where you’d expect the big-serving Pole to excel, has been far worse.

In a dozen majors, Hurkacz has gone 14-12, a 54% win rate compared to an expected mark of 73%. The odds of such a wide gap are 0.9%, slightly more extreme than Tiafoe’s happier results. In the same span, Hubi has outperformed in best-of-three, winning 63% of those matches instead of 58.5%. He is the anti-Berrettini.

We’ve learned today that outlying best-of-five records are probably not predictive of future results. For a statistician, such findings can be a bit disappointing. For Polish fans, though, it’s reason to rejoice. Hurkacz didn’t turn things around in Australia, winning one match and losing his second. Still, a correction remains in the cards. If apparent best-of-five specialists like Berrettini and Tiafoe can lose in the second round, a laggard like Hurkacz could–eventually–give us a deep run.

* * *

Subscribe to the blog to receive each new post by email:

 

This Is What a Dangerous Madison Keys Looks Like

Madison Keys in 2023. Credit: Hameltion

Last year, Madison Keys missed the Australian Open with a shoulder injury. She ended up playing barely half the season, missing more time after tearing a hamstring at Wimbledon. She still won enough matches to head to Melbourne as the 19th seed at this year’s first major.

She’s better than that.

In one sense, I’m just stating the obvious: She beat Jessica Pegula for the title in Adelaide on Saturday and moved up to 14th on the WTA computer. Beyond that, anyone who can hold on to a spot in the top 20 despite missing so many events is probably better than their ranking. Elo agrees, rating Keys ninth among women, a modest 33 points behind fifth-place Elena Rybakina.

Even more striking is the way the American won the Adelaide championship. She served as well as she has in years, indicating that the shoulder is fully healed. She played extremely aggressively, a style that she has never shied away from, but that she sometimes struggles to maintain. Finally, Keys did all that while posting excellent return numbers. The 29-year-old is a two-time semi-finalist at the Australian Open, and if she keeps this up, she could easily make it three.

The serve is back

When everything clicks, no one on tour–with the possible exception of Rybakina or Aryna Sabalenka–makes serving look so easy. Keys doesn’t just slam flat serves down the tee: She adds a bit of side spin, so her inch-perfect deliveries look like they’re sailing slightly wide until after they cross the net. Then she employs the same spin to send wide serves even wider. When she misses, she can fall back on some of the heaviest topspin seconds in the women’s game.

Whether the shoulder was still shaky or the hamstring compromised her motion, the American struggled to maximize her serve as late as last year’s US Open. In her third-round loss to Elise Mertens, her average first-serve speed was just under 99 miles per hour. Out of nearly 100 grand slam matches for which I have serve speed data, it was only the second time–the other was 2017 Roland Garros–that she hit firsts so slowly.

Today in her Melbourne opener against Ann Li, her average first serve was 109 miles per hour. That’s the fastest I have on record for her since 2015.

I don’t have serve speeds for Keys’s victories last week in Adelaide, but the results hint at numbers well into triple digits. In the final against Pegula, she hit 10 aces, good for 13% of her serve points. Facing Liudmila Samsonova in the semis, she smacked 12 aces–17% of serve points. In a short quarter-final against Daria Kasatkina, she tallied 11 aces, an eye-popping 21% of serve points. It was only the fourth time in the 2020s that Keys topped the 20% mark and the only time in her career she managed it against a top-ten opponent.

Adelaide marked the first time since 2019 that the American aced at least 10% of her service points in three consecutive matches. She hadn’t done so at a single event since 2016.

Aces are great in themselves, but the stat is particularly useful for representing the serve’s effect on even more points. Yes, Keys won 13% of her serve points against Pegula with aces, but 41% didn’t come back. That’s another sign of a revival: In dozens of Match Charting Project-logged matches, it’s the first time she’s topped 40% in that category since the Australian Open in 2022–her most recent semi-final run Down Under.

The American mitigated her shoulder woes last year by starting points more conservatively. She wasn’t as deadly with her first serve, but she landed more of them. Among the WTA top 50, only Elina Avanesyan and Yulia Putintseva missed fewer first serves. If Adelaide is any indication, it’s back to business as usual, taking a few more risks and wreaking absolute first-serve devastation:

Span        SPW  1stIn  1st W%  2nd W%  
Adelaide  65.4%  63.4%   71.6%   54.8%  
2024      60.6%  68.2%   66.7%   47.4%

It’s not an apples-to-apples comparison, because the 2024 line contains plenty of clay-court matches, including two against Iga Swiatek. But the difference is sufficient to tell the story anyway. 60.6% of serve points won was good for 8th-best on tour last year. 65.4%, on the other hand, is almost two percentage points better than anyone posted on hard courts. The 71.6% first-serve win rate would have put her in the top five, and no one came close to winning 54.8% of their second serves.

I don’t want to put too much emphasis on a single tournament–everybody looks good if you turn the microscope on a great week. But it’s worth offering one more tidbit in Keys’s favor. She posted those numbers against extremely strong opposition. Her five victims in Adelaide were ranked 16th, 17th, 9th, 26th, and 7th, respectively. That’s a tougher schedule that any player faces over the course of an entire season. If Madison does reached the Melbourne semis, it’ll be an easier path than she faced to collect the trophy in Adelaide.

Swinging freely

Keys has improved her return game over the years, and she’s gotten more comfortable playing long rallies. One of the more surprising numbers on her stat sheet is that she has a better winning percentage on clay than on hard courts.

Still, she’s an aggressor at heart. Her serve isn’t the only shot she can hit as hard as anyone, nor is it the only weapon she can land on the line. Generally speaking, the more aggressive she is, the better her results. The shoulder and hamstring injuries forced her to play more conservatively. That is now over.

In less than an hour on court with Kasatkina, she crushed, by one count, 38 winners. Facing Pegula on Saturday, she tallied 40. I have winners and unforced errors for about one-quarter of her career matches, and the Adelaide final was the first time she cracked 40 winners since 2019. It wasn’t uncontrolled either. The opposite side of the ledger was a respectable 27 unforced errors, good for a ratio of 1.5. Even in her Auckland loss to Clara Tauson the previous week, she recorded 38 winners against 30 unforced, a ratio that would win most WTA matches.

The best indicators of the American’s renewed attack are the various metrics for aggression. By Rally Aggression Score–a measure of how often a player ends points for good or ill after the return of serve–she rated +147. (Average is 0, and almost all players fall between -100 and 100.) Return Aggression Score–the same idea, but strictly for returns of serve–put her at +137. Her career averages are around +100, but in 2024, she fell below +60 in both.

The last time that Keys reached +137 or higher by both measures was the 2019 Cincinnati quarter-final, when she beat Venus Williams en route to the title.

We keep finding things that Keys has done for the first time since 2019. They almost all go back to that week in Cincinnati. (Coincidentally, she straight-setted Kasatkina there, too.) With the possible exception of her 2017 US Open final run, the Cincinnati effort was the best of her career. She has found that form again.

Keys to the match

One difference between 2019 Cincinnati and 2025 Adelaide: The American returned a whole lot better last week. She won 48.1% of return points in Adelaide, compared to 43.7% in Cinci.

It’s rare for players to substantially improve their return game once they arrive on tour. The rest of the tour learns how to beat you, the opposition gets stronger, and age slows you down. Yet Keys, in her late 20s, has gotten better:

While the 2025 data point probably won’t stick above 47%, the 2023 and 2024 results demonstrate the trend. Last year, Keys’s 44% mark was better than half of the top 50, a strong showing for a serve-first player. Return points are an extreme case of tennis’s small margins. By top-50 standards, 43% of return points is weak, 44% is adequate, and 45% is strong.

47%–the American’s success rate in Auckland and Adelaide–is beyond elite. Only two players–Coco Gauff and Marketa Vondrousova–did better than that on hard courts last year.

It will take some time before we know whether Adelaide was an outlier or a harbinger of a resurrected career. Keys’s 2025 season will surely fall somewhere in the middle, at least if she remains healthy. There are certainly reasons for optimism. For the most part, she’s done all of this before, serving and attacking her way into the top ten as far back as 2016, and returning better in the last two years. If those two halves come together, we won’t see a (19) next to her name again for a long time.

* * *

Subscribe to the blog to receive each new post by email:

 

Trivia Notebook #1: Ranking Leaps and Marathon Men

Tomas Martin Etcheverry, three-set marathon expert

I run a lot of queries, and people often ask me arcane trivia questions. Has this ever happened before? Is that a record? My beat seems to be the super-niche stuff that no one would ever bother to include in the official media notes.

The Trivia Notebook is my attempt to put more of the answers in one place. I’m thinking I’ll do one of these every two or three weeks. If you have a question or topic you think would fit well here, please send it. No promises–most of my ideas don’t end up making the cut.

It’s a new year, so we’re all bursting with energy to start new projects. Most of them are long gone by April, and odds are the same thing will happen to this one. But hey, you never know, right?

In this first installment, we’ll look at 100-point ranking leaps, marathon man Tomas Martin Etcheverry, seedless quarter-final lineups, and single-country duos that conquered a tournament.

100-spot ranking leaps

Joao Fonseca ended 2024 ranked 145th in the world. My Elo ratings put him 45th, and after the Canberra title last week, his place on that list climbed to 27th.

As with many trivia questions, we’ll need to be a bit more specific. Tons of players move up 100 places each year, but going from 845th to 745th–while impressive!–is presumably not the sort of thing we’re looking for. Same thing with injury recoveries. While Pablo Carreno Busta finished 2024 ranked 196th, it won’t be momentous if the former top-tenner bounces back to the top 100.

A narrower question, then: Which players have jumped at least 100 ranking places in a single year, ending with their first year-end top-100 finish?

Here are the biggest single-year improvements that ended with a top-100 debut:

Player               Year  Prev YE  New YE  Jump  
Kenneth Carlsen      1992      835      69   766  
Leonardo Lavalle     1985      745      87   658  
Guillermo Coria      2000      722      88   634  
Pablo Carreno Busta  2013      654      64   590  
Marco Chiudinelli    2009      605      56   549  
Jacob Fearnley       2024      645      99   546  
Josef Cihak          1987      613      77   536  
Andreas Vinciguerra  1999      633      98   535  
Andre Agassi         1986      618      91   527  
Alex Michelsen       2023      599      97   502  
Arnaud Di Pasquale   1998      572      81   491  
Radek Stepanek       2002      542      63   479  
Ben Shelton          2022      573      96   477  
Fritz Buehning       1979      555      81   474  
Jannik Sinner        2019      551      78   473

Pablo made it! A few other names there you might recognize, too.

If Fonseca skips forward 100 spots, he’ll do something that sets him apart from everyone on that list: He’ll leap into the top 50. Still, a 100-spot move is hardly historic:

Player               Year  Prev YE  New YE  Jump  
Marc Rosset          1989      474      45   429  
Ronald Agenor        1985      418      49   369  
Goran Ivanisevic     1989      371      40   331  
Vincent Van Patten   1979      374      43   331  
Sergi Bruguera       1989      333      26   307  
Juan Carlos Ferrero  1999      346      42   304  
Jim Courier          1988      346      43   303  
Horst Skoff          1986      299      42   257  
John McEnroe         1977      264      18   246  
Ulf Stenlund         1986      274      34   240  
Mark Philippoussis   1995      274      38   236  
Peter Lundgren       1985      265      31   234  
Ricardo Cano         1975      274      42   232  
Jack Draper          2022      265      42   223  
Mel Purcell          1980      245      27   218

About 80 players have made a 100-plus-spot jump into the top 50. It’s harder to do so now than it was in the days of McEnroe or Courier, but men still manage it with some regularity. Fonseca will have to settle for breaking other records.

Marathon men

This was the Adelaide second round. Thanasi Kokkinakis decided this was enough for his Australian Open prep, as he withdrew from the quarters. Headlines about this match tended to focus on Thanasi’s penchant for marathons. He’s well-known for his 5h45 battle with Andy Murray two years ago in Melbourne. Last year, he went 3h15 against Aleksandar Kovacevic in Houston, then 3h29 a week later at the Sarasota Challenger against Gabriel Diallo.

But… the name that caught my eye was Tomas Martin Etcheverry. While he doesn’t have a marquee marathon to his name like the Murray tilt, he spends a lot of time on court. Just three months ago, he muscled through three hours and 43 minutes to beat Botic van de Zandschulp in Shanghai.

Etcheverry doesn’t have a ton of slam experience, and the best-of-five format lends itself to memorable marathons. But in best-of-three matches, the Argentinian has now crossed the three-hour mark more than any other active player:

Rank  Player                   Bo3 Marathons  
1     Tomas Martin Etcheverry             27  
2     Albert Ramos                        26  
3     Novak Djokovic                      25  
4     Pedro Martinez                      24  
5     Carlos Taberner                     23  
6     Thiago Monteiro                     22  
7     Roberto Carballes Baena             20  
8     Mikhail Kukushkin                   19  
8     Timofey Skatov                      19  
8     Juan Pablo Varillas                 19  
8     Thanasi Kokkinakis                  19  
12    Lorenzo Giustino                    17  
13    Jordan Thompson                     16  
13    Alessandro Giannessi                16  
13    Marton Fucsovics                    16

This is an imprecise measure, because it’s really “three-hour matches I know about.” It includes tour-level matches back to 1991, tour qualies and Challengers going back a decade, and Challenger qualies for the last few years. So it’s biased a bit toward younger players, who have played more in the “Jeff knows about their match times” era. Still, it’s an impressive tally for Etcheverry–and he’s only 25 years old.

The Kokkinakis match also tied Etcheverry for first place on the all-time list with Nicolas Massu. Here’s that leaderboard, again with the caveat that older players do not have Challenger matches counted:

Rank  Player                    Bo3 Marathons  
1     Tomas Martin Etcheverry              27  
1     Nicolas Massu                        27  
3     Albert Ramos                         26  
3     Carlos Berlocq                       26  
5     Novak Djokovic                       25  
5     Rafael Nadal                         25  
5     Andy Murray                          25  
8     Pedro Martinez                       24  
9     Carlos Taberner                      23  
10    Thiago Monteiro                      22  
10    Paolo Lorenzi                        22  
12    Adrian Menendez Maceiras             21  
13    Roberto Carballes Baena              20  
14    Mikhail Kukushkin                    19  
14    Timofey Skatov                       19  
14    Juan Pablo Varillas                  19  
14    Thanasi Kokkinakis                   19  
18    Gilles Simon                         18

Legends one and all. It’s continually amusing to me that Djokovic, Murray, and Nadal have landed on the same number. Roger Federer, for his part, only reached three hours in six of his short-form matches.

Seedless quarter-finals

At the Nonthaburi Challenger this week in Thailand, the quarter-finals featured a wild card, two qualifiers, and an alternate… but no seeds. One of the seeds withdrew, five lost in the first round, and the remaining two fell in the second.

Let’s say it together: Has that ever happened before?

In fact, there were seedless quarter-finals five times at Challenger level last year, including once in Nonthaburi! Altogether, there have been more than 80 such tournaments in Challenger history. Even “two seeds in the second round” isn’t that special–it happened at Amersfoort last year.

What about one seed in the second round? For that, we have to go back to 2018 in Lyon, where a 17-year-old Felix Auger-Aliassime defended his title. Gotta love the Wikipedia summary of how things went for the seeds:

As far as I can tell, that’s the closest we’ve come to a Challenger with no seeds in the second round. Credit to Pablo Andujar, he did his best.

Lonely countrymen

Last one, this time from the archives. Last year at the Dobrich Challenger, two Dutchmen–Jelle Sels and Guy den Ouden–met in the final. It isn’t unusual to have players from the same country in a Challenger final, even outside their home country. What was odd about Dobrich is that Sels and den Ouden were the only Dutch men in the main draw.

You know the drill: Has that ever happened before?

I should know by now: Ask that question about the varied history of the Challenger tour, and the answer is almost always yes. It happened again in September, when two Japanese men met for the Columbus final. It also arose twice in 2023. Two Bosnians played for the championship in Sibiu, and the San Benedetto title match was contested between Benoit Paire and Richard Gasquet, the only Frenchmen in the draw.

Altogether, there have been 32 such Challengers. There were none in the first decade of the tour, but they’ve clicked off about once per year since. My favorite of the bunch is the 1998 Fürth Challenger, where Christian Ruud and Jan Frode Andersen saw off all of their non-Norwegian foes.

This scenario has also come up about as often at tour level. More than half of them were before 1980, and they’ve gotten progressively rarer. But we got one in 2024! Arthur Fils and Ugo Humbert were the only two Frenchmen in the Tokyo draw, and they were the last two men standing. That was the first such tournament in a decade, since Monte Carlo in 2014, where Stan Wawrinka upset Roger Federer for the title.

That, I think, is enough tennis trivia for one day. We’ll have some more–maybe!–in a couple weeks.

* * *

Subscribe to the blog to receive each new post by email: