This is the final story in our series about the analytics boom in tennis. Part 1 examined the use of performance data in match preparation. Part 2 looked at wearables, tracking technology and training. This story projects how technology will shape the future of the sport.
* * * * *
The revolution of Big Data has prompted all sports to face a statistical reckoning and reexamine long-standing beliefs. For baseball, it was the sacrifice bunt. For basketball, it was the mid-range jumper. For football, it was always punting on fourth down. Those historically common plays didn’t hold up under the microscope of analytics. Now people are starting to ask similar questions about tennis, which is so steeped in tradition that even the players’ grunting is deemed controversial.
Quiet, please? Not anymore.
Everything is now under review in tennis, from rudimentary to advanced aspects of the game, and the evolution is being informed by a data set that builds upon itself with each new match. Consider the serve. It’s conventional wisdom that players should toss the ball the same way every time to disguise their shots. “But, actually, is that true?” asks Lorcan Reen, a Lawn Tennis Association performance analyst who has worked closely with Andy Murray. “If some players are getting away with their ball toss going everywhere, the notion of how we believe things should be—is that actually true?”
In the not-too-distant past, a player may have checked an opponent’s serve speed and received a few pointers from a coach’s video study. But many of the world’s best players now have a dedicated analyst or company on retainer tapping into new technologies that provide deep insights into statistics and trends.
PART 1: Advantage, Analytics—How Tennis Players Are Using Science to Elevate Their Artform
PART 2: In Change of Direction, Tennis Catapults Into the Future With Embrace of Technology
While the sport’s embrace of technology started slowly, it’s now accelerating faster than a John Isner serve. And Isner, for one, has benefited from challenging long-standing beliefs about tactics. The 6’ 10” American has a howling serve—routinely exceeding 140 mph—but he lumbers around the court less quickly than most of his opponents. Yet it was his willingness to attack the net rather than play it safe from the baseline that helped him ascend into the world’s top 10.
Craig O’Shannessy, the founder of Brain Game Tennis and a strategy analyst for the ATP World Tour, explains that when opponents used to return Isner’s serve, the point was a 50-50 proposition. Isner had lost his service advantage by remaining on the baseline. Most of those returns were typically defensive shots—ideal for Isner to go on the offensive and attack the net. Isner initially called that notion “crazy,” O’Shannessy says, but “if you look at John’s rise, it certainly has something to do with that, for sure.”
* * * * *
Inspiration for reimagining tennis has been found in seemingly unrelated fields. Golden Set Analytics has hired a game theory expert. O’Shannessy has partnered with a supply chain AI firm. An Israeli sports data provider is harnessing computer vision to turn any single-vantage point video feed into advanced data. A new tennis ball machine has taken inspiration from advancements in drone and 3D printing technologies, allowing it to be programmed a smartphone app to fire complicated sequences of shots at varying spins and speeds. Tech giants IBM, SAP, and Infosys are perpetually iterating new tennis metrics.
Advanced analytics have largely been limited to single-shot data points, but ever-growing sophistication in machine learning is expanding the scope of exploration. “So I don’t just want to understand where the serve lands, but maybe I want to understand more,” says Milan Černý, SAP’s innovation lead for tennis. “How does the rally evolve based on certain factors?”
Černý offers a few possible variables—game score, hand of opponent, time of day, court surface—and explains that much more can be done with the existing data that’s already being culled from Hawk-Eye’s optical tracking cameras at elite tournaments. “We are working on a new evolutionary step for analytics, which is going to go beyond analyzing just single points,” he says. “It’s a bit early to speak about this today, but just as a preview, we [already] analyze single points and patterns in serves, patterns in returns, and player positioning—and now it’s really going to be about taking that next step and looking into what’s happening during a rally?”
New technology, of course, can expand the tennis industry’s understanding of the sport and paint pictures of player tendencies, but the consensus among analysts is that existing datasets can be mined for further examination. Artificial intelligence can expedite that process and identify new insights. “The next frontier is more about slicing the data in different ways than we currently do,” O’Shannessy says. “I’ve actually teamed up with a supply management company called Rightchain. Supply chains are insanely complex. We kind of look at a point in tennis the same way—at all the different computations and different parts of the court.”
Boutique firms such as Golden Set Analytics are pioneering new areas of analysis. “Pretty much anything I ask, they’ve been able to uncover and show me numbers based on what I’m seeing or what I’m not seeing on video,” says Billy Heiser, who coaches GSA clients Alison Riske and Dominik Koepfer. “It gives me a lot of confidence as a coach.”
* * * * *
In preparing a player for an upcoming opponent, analysts must weigh recent performances with the historical record. O’Shannessy says certain shot patterns remain consistent, but players adapt. He gives an example of old Andre Agassi tapes in which the all-time great played distinct styles to either force opponents into certain shots or to run them side to side to wear down their conditioning. Over the course of days and weeks, certain skills can improve, informing the reports he prepares as the strategy analyst for Novak Djokovic, who currently holds the world’s No. 1 ranking over Rafael Nadal.
“We put a lot of weight in what’s happening now with an opponent,” says O’Shannessy, who lets no detail go unchecked. “You certainly want to see what the weather is currently.”
HIGHLIGHTS ON DEMAND: How the PGA Tour Modernized the History of Golf
The age of an opponent likely factors into that. Unless a veteran tennis player has to overcome an injury, the deviations are usually minimal. “Nothing formal has been done on this yet, but what the data tends to show is that players who are younger—say, between 18 and 22—there’s more chance they’ll be changing their game a bit more,” says Reen, the personal analyst for Murray. “That’s where a lot of developments can be seen in the stats, match by match. And it sounds like commonsense that, the older a player gets, like a guy in his 30s, is he really going to have changed his game so much?”
Use of technology for longitudinal tracking of player development is a key growth area. For now, Tennis Analytics founder Warren Pretorius says his firm still tags video of matches by hand in the Dartfish platform but expects to change over to automation as the AI accuracy for tennis improves and becomes more affordable. National federations are doing similar work—Tennis Australia still adds rich shot detail manually whereas the USTA benefits from its partnership with IBM to tap Watson for this automatic capability.
HIGH SEAS: Only Thing Faster Than SailGP Boats Is the Data Fueling the Fan Experience
LSports has been supplying the gaming industry and media companies with sports data for eight years, but its founding team has launched LVision as a tennis offshoot. Using only one camera angle, the system can analyze ball and player tracking with a latency of less than 400 milliseconds.
“We are able to do it for any video stream, from your mobile phone to broadcast TV, YouTube, security cameras—any video source,” says CEO Ido Lazar, who adds that more than a million points have been put into the system from international matches. “During the rally, we can tell you the chance of every player to win the point or how long would be the rally [last],” Lazar says.
PlaySight’s camera system, which can be found at the USTA National Campus in Orlando and is used by dozens of colleges, provides instant feedback on every shot, including metrics on type, speed and spin. Soon, the company may be able to provide automated feedback to guide training. “We’d like to take it to the next step of insights—someone goes out and practices for an hour, then we send them an email saying, ‘Hey, we noticed in this practice set that you missed more backhands in the net than the average player or than you typically do.’” says Josh Graves, who leads PlaySight’s North America Tennis Division.
Such tracking technology gleans a wide array of data from its camera systems, but more actionable insights remain elusive. “What doesn’t happen at the moment is really looking at what the player is doing on a micro level,” Černý says. “We have the position of the players, that’s fine. We have the ball, that’s also fine. But when are we going to be able to analyze what limbs are doing, for instance, during a match? How do movements affect efficiency and readiness, for instance?”
Startup company Hydrogen Sports recently developed its new Proton tennis ball machine that’s half the heft of traditional options. They used miniaturized motors inspired by drone technology to save weight, and 3D printers made robotic motion control more accessible. All of this leads to a machine that can shoot balls up to 70 mph at 3,000 rpm—“Federer-level top spin,” says founder Jonah Harley, a former Apple senior manager and primary inventor of the Apple pencil.
The machine comes with 10 pre-programmed shot routines but has a near-infinite capacity for users to design their own. In the future, Harley envisions partnerships with online tennis instructors. Users could watch a video, then download a shot program and maybe add a camera for feedback too. “Once you have a semi-smart machine, what else can you do?” he says.
* * * * *
Singles tennis can feel isolating, with competitors standing alone on the court for upwards of five hours in a Grand Slam. Concentration, focus and resilience are paramount. That’s why four of the experts polled for this story all indicated that the holy grail of tennis analytics would be finding a way to quantify the mental chess game within a match.
“When you can obviously see there’s a change in play, how does that register within a player’s mind?” says Tennis Australia performance analysis manager Darren McMurtrie, referring to his or her capacity to adapt to an opponent’s adjustment.
BEHIND THE SCENES: Inside ESPN’s High-Tech Operations at the U.S. Open
Heiser, the pro coach for Riske and Koepfer, offered one starting point for how to measure a player’s response to pressure. “The only way you could even start to look at that is if you could monitor a player’s heart rate during different situations in the match,” he says. “Then you could start to scientifically get it a little bit.”
No such wearable is currently on the International Tennis Federation’s list of technologies permitted in competition, although the recent approval of Catapult Sports’ GPS tracker could be a harbinger of a change in attitudes toward such devices.
“Really thinking next-gen, cutting-edge, futuristic, I think cognitive analysis and having a way of quantifying what’s happening with a player in terms of their body language, their self-talk, their use of rituals and routines, their breathing,” says USTA general manager of player development Martin Blackman. “That’s probably the next big step in this area because those are things that can also be trained.”
External analysis may help in such a process. IBM’s Watson already analyzes audio and facial expressions to help discern key points for inclusion in its automated highlights. “Even at the beginning of this project, we discussed that as being part of the end game,” Blackman says. “In my mind, that would be next-level.”
Question? Comment? Story idea? Let us know at talkback@sporttechie.com