A few days before last year’s Australian Open, Machar Reid, the head of innovation for Tennis Australia, was walking in downtown Melbourne when he heard someone call his name. He turned to see Alexander Zverev, the tournament’s No. 5 seed, who crossed the street and asked him about a new stat he’d just released: time pressure. After explaining that it measured a player’s ability to make an opponent feel rushed, Reid told Zverev there was another metric that the German ranked high in, but he couldn’t remember which. “Can you send it my way?” Zverev asked. It was a small request, but for Reid, it was a welcome harbinger of change.

Tennis has an unimpressive track record of using analytics. At the MIT Sloan Sports Analytics Conference, the preeminent forum of its kind since 2006, a paper on tennis didn’t appear until 2016, lagging behind not just the Big Four sports but also soccer, hockey and gambling. And even though statistics are commonplace on TV broadcasts, according to Reid, they’re used more to provide superficial support than to deliver complex insight.

In fact, tennis analytics have advanced to the point that statisticians like Reid can now produce metrics for every facet of the game, from color-coded diagrams of where a player served on each break point to graphs of the speed and weight of an opponent’s backhand. If used correctly, these statistics could not only make the sport safer and easier to coach, but also push it into a new era, where mental fortitude is considered just as important as physical ability.

German Alexander Zverev didn’t believe that he ranked first in time pressure, the ability to make an opponent feel rushed. He told Reid, “It should be Federer’s.”

Four years ago, French coach Thomas Drouet started an unusual nightly routine: updating a spreadsheet of the players in the women’s field. “In the beginning, it was crazy,” he says. “I didn’t know how to use Excel, so I was writing with my hand, and then with my phone, making the averages.” Since then, Drouet’s technical skills have improved, but he still spends 75 to 90 minutes a night updating his personal database, which contains entries for 140 players on three surfaces with information as granular as the average speed of the second-serve on break points.

“Sometimes I don’t want to do it,” says Drouet, “but I have to be disciplined.” This commitment is especially impressive considering that the information he records is already provided by the Women’s Tennis Association through its analytics partner, SAP. However, Drouet says its data isn’t as organized or exhaustive as his, and the quality of his coaching depends on reliable metrics.

“I think we’re all emotional,” he says, “but now, I immediately try to understand what’s happening with the data I have in front of me.” He gives the example of service returns: “We all say, ‘OK, you have to be a bit more aggressive on the second serve,’ but if we look at the statistics, the majority of girls hit slower on the second serve because it’s a bouncing serve, so it’s harder to block.”

Last year, Tímea Babos, whom Drouet was coaching when she became the No. 1-ranked doubles player in July, had more double faults than she did aces (215 to 170). So it was important for her to learn that during important points she could lean on her second serve. “There is a logic,” says Drouet, “but if you don’t see the numbers, how can you really see that there’s two, three miles different in the speed?”

In fact, Drouet has reconfigured his entire approach based on data from the women’s side about what wins matches.”They are not going to overpower you,” he told Babos. “You think that hitting 5 kilometers faster than Halep is going to help you? No, the game is more mental at this level.”

Reid has reached the same conclusion, which is why he’s taken aim at the four traditional styles of play: base-court, all-court, counter-puncher and serve-and-volley. “Those styles haven’t existed for the past 15 years,” says Reid, “but the sport continues to default to that.”

Instead, Reid proposes that a player’s style also include factors like how well he performs under pressure. Aptly named “clutch,” this metric gives greater weight to more crucial points (set points over deuces over love-alls), and it helped Reid and his team develop mental profiles of the pro players. In a paper for MIT Sloan’s Sports Analytics Conference, they analyzed 3 million competition points to group the men’s field into eight categories, one of which was populated by just one man: John Isner.

Not only do the massive serve and big forehand of the 6-foot-10 American defy conventional labels, but even under unfavorable match conditions, his serve stays consistent, suggesting “greater overall mental toughness … than any other player evaluated.” However, when it comes to defense, Isner shows a “lack of confidence” on important points. To fully understand Isner’s style of play, therefore, you have to know whether he’s serving at love-all or returning a match point.

This data-driven approach also strips away extenuating factors such as theatrics (although Reid is also studying those). For example, Australian Nick Kyrgios has a reputation for erratic behavior on the court: During the 2017 Australian Open, he famously committed 68 unforced errors, swore loudly and smashed his racket into the ground. At last year’s US Open, he stopped even attempting to return his opponent’s serve, prompting an umpire to leave his chair to give the Australian a pep talk. Despite this behavior, Reid’s analysis of Kyrgios’ record suggests a mental profile similar to other big servers. “We’re looking at the scoreboard,” says Reid, “when and what type of points are won, not how points are won.”

” I’m always looking to numbers for things I need to correct, not if I’m doing something well.”  Milos Raonic

Some players are more eager to adopt this kind of dispassionate approach than others. Milos Raonic, currently ranked 17th in the world, naturally gravitated towards it. “I come from a family of engineers, so it made sense for me to try to understand probabilities,” he says. “But I realized I wasn’t focusing on things I could control. That’s the thing in tennis. It’s not set plays. It’s what a guy likes to do, but a guy might adjust to you already.” Now, Raonic uses metrics mostly to confirm his instincts about a match, especially where he could improve his serve. “I’m always looking to numbers for things I need to correct,” he says, “not if I’m doing something well.” According to Reid’s data, this approach is paying off. His serve (along with Isner’s) is one of the most consistent on the tour.

Even for analytical players like Raonic, though, coaches must tread carefully, according to Jenni Lewis, the Global Sponsorships Technology Lead at SAP. For each player, she tracks a “magic number,” the first-serve percentage, below which that player is unlikely to win the match. She shares this information only with those who “are worried about their serve anyway,” she says. “Let’s make them worry on a fact, not a whim.” But even then, she pairs it with another number — say, a certain return percentage — so that a player can compensate when their serve falters. “That’s where a good coach can really help an analytical player,” she says.

In that sense, protecting players from data can be just as important as exposing them to it. “From day one, we thought that coaches would want to know about environmental things,” says Lewis, “and they go, ‘Do not ever tell my player anything about [how she performs under certain] conditions because she needs to play whether it’s day or night, 100 degrees or 5 degrees.'”

Analytics can also help players avoid injury. Because world rankings are determined by points accumulated during the entire season, players are incentivized to compete as often as possible. It’s easy to be on the court nonstop all year, and few players stay completely healthy. Last year, Zverev played 77 matches — more than anyone else in the top 100. Last week, he pulled his hamstring and had to withdraw from an exhibition in Adelaide. Because of injuries, Rafael Nadal has already pulled out of the Brisbane International this year, Juan Martin del Potro is skipping the Australian Open, and Andy Murray, still struggling to recover from hip surgery, has announced his retirement.

Reid’s solution? Personalized injury-prevention plans. Relying on physiological sensors similar to the ones used in the NFL and MLB, he measures inputs like a player’s distance covered, top speed, acceleration, changes in direction, and hitting frequency for the previous 21 days. Together these factors “provide you a reasonable sense of how resilient your body is at any moment in time” and could help a player decide whether to play through a sore shoulder or take a break. However, Reid cautions that these plans would be suggestive, not prescriptive: “Players are competitive beasts. If a player’s in a Grand Slam and going deep, it defies logic to pull him back.”

Reid’s analysis shows that American John Isner’s powerful serve stays remarkably consistent even under pressure.

If tennis analytics are so powerful, why has it been so slow to take root?

One of the major obstacles are the governing bodies. Because Tennis Australia produced great players from the 1980s to 2000s but slumped after that, Reid says they were more willing to experiment with analytics. That is not the case in the larger tennis culture. In fact, Rule 31 of the International Tennis Federation’s bylaws essentially prohibits players from using analytics during gameplay. “Simply by having the rule,” says Reid, “disincentives the use of technology from the outset.”

Another problem is the federation structure of tennis, which prevents the widespread, uniform adoption of analytics. “In general, I am skeptical that big data exists in tennis at all,” Jeff Sackmann, who runs a website devoted to tennis analytics, wrote in an email. “IBM likes to tout all of their data and everything they’re doing with it, but what they have is rather limited data that doesn’t even encompass every match for each [Grand] Slam they’ve covered. Even Hawkeye tracking data, which has a claim to be ‘big data’, is rigorously siloed, so no single organization ever has more than a fraction of the total possible dataset.”

Sackmann relies on volunteers to help him chart matches, which gives him enough data to answer fairly simple questions, like whether Djokovic a great returner. However, when it comes to questions like how a player’s style has changed in the last six months or whether certain players can read where a serve will land based on minor differences in the server’s toss, aspiring data crunchers are, in Sackmann’s words, “s— out of luck.”

Coaches like Drouet can maintain their own databases. But not every coach has the patience, and not every player has the resources to buy, collect, and crunch the data from various tournaments. The pay gap in tennis is wide — last year, the highest-earning man made 62 times that of the man in 200th place — and coaches and players are cagey about the cost of hiring an outside firm. When asked if all the players in the top 50 could afford the analytics company that he uses (but declined to name), Brad Stine, a former US National Coach for the USTA, said this: “That’s a determination for a player. If using analytics wins you one more round in a Slam, then it’s obviously worth it.”

Reid is quick to point out that analytics are not “the panacea to all of sport’s woes.” Still, he’s confident that his work will ultimately improve the game for players, coaches, and fans. “With these stats, you can better draw down on causalities,” he says. “Are you 100 percent sure that A plus B equals C? No, but you’re a hell of a lot closer than you were a year ago.”


Please enter your comment!
Please enter your name here