Taking a shot at AI tennis coaching

Photograph by Andy Hall/ The Observer

Given I have not played tennis since I was a teenager, I am nervous to have my ability analysed by an app that cheerily declares it can compare you with professional players.

The format of this technology is increasingly common across amateur sport. From running to cricket and everything in between, there is an app that will use video to help you improve on everything from form to technique.

The buzzword around a lot of these apps is “AI”. The ability to process large amounts of information from massive datasets has undoubtedly been a game changer in sports analytics, but the amount of data available has been supercharged by something much more simple, and often overlooked: the camera.

“We want to be able to take in video from almost any source,” says Lauren Pedersen, the founder of Sport AI.

“It’s becoming much more common around the world to have cameras mounted at local clubs. That’s because the cameras themselves are getting cheaper and more powerful. We can take in that video and analyse everything.”

The development of camera technology means that it is now easier than ever for people at any level and in any sport to record themselves.

That in turn increases the amount of information that can go into datasets and allows people to upload their own recordings and compare themselves with those sets of data.

When sports technology company Hudl was founded in 2006, teams were burning DVDs for their coaches or players. At the time, that technology was accessible only to the biggest teams and limited by poor quality and time-consuming editing.

“Our goal was to give every athlete and every team access to high-quality content,” says Matt Mueller, who is the chief operating officer of Hudl. “They could just put a camera up and it automatically catches matches and practices.

“The next iteration was finding ways for teams to sort through this content, and find the key moments that matter. That’s where computer vision and AI have come in and helped move things to the next level.”

Mueller emphasises that without the development in camera technology, there would be little that those technological techniques could do.

“It has gone from being able to say a shot happened, to this person made a shot, to this person made a shot and here’s where all the other players on the court were.”

Crucially, cameras have got cheaper, with that quality improvement, making the analysis more accessible and therefore better even at lower ­levels of sport.

The next step from that analysis is adding the kind of in-app coaching advice that characterises what SportAI is able to do in tennis.

“You’ll often have coaches taking out a mobile phone, capturing video of someone playing and showing it back to them,” says Pedersen. “But there’s no real data being captured. Now, we can look at everything from biomechanics, angles, joints, how you’re bending, how you’re moving, and giving feedback that is really data driven.”

Pedersen emphasises that this feedback is the kind of quantitative data that coaches would not be able to know immediately just by watching an athlete.

“Being able to put data behind it helps coaches track progress and change. You can compare yourself to a pro, but you can also compare yourself to your game last week. Those kinds of comparisons can be really helpful for players.”

It is not just coaching where camera technology has drastically changed sports. The use of Hawk-Eye systems in cricket and tennis allow for different forms of adjudication.

But overreliance on cameras comes with its own issues. Tennis has almost entirely done away with human line judges, instead using electronic line-calling. One incident at Wimbledon, when Sonay Kartal hit a ball that was obviously long but was not called out, demonstrated the issue. Rather than losing the point, and the game, it was instead replayed and Kartal went on to break her opponent.

Proponents of technology in sport would argue that human error would be far more common than situations such as this, but it shows how blind faith in the objectivity of these systems can be misguided.

Mueller believes that advancements will continue to take place, with youth sports set to be the next big area of growth for camera technology.

“These systems used to be a $150,000 install,” he says. “You can’t do that as a high school, or a youth club. That forces people like us to continue to leverage tech to unlock things for the masses.

“People are going to chase this for a long time, and the more video they have, the more accessibility that they have to get real insights out of that, and the better results will be for every athlete.”

Back at Roehampton, my results are in. As I watch myself hitting a forehand in slow motion, I am given an overall score of 67 out of 100. My kinetic chain timing is my highest score – “fairly well-coordinated” – but my swing curve needs desperate improvement. All in all, I’ll take it.