Serving up tennis improvements through AI

For the love of tennis

The company’s management combines technology expertise and a passion for tennis. In addition to extensive experience growing global tech firms, Pederson competed in NCAA Division 1 college tennis and represented Norway at the 2023 ITF Tennis Masters World Championships. As well as being an entrepreneur and sports technology product expert, Kittelsen was CEO of Sevensix Tennis, the provider of an app designed to analyse tennis technique to help players upgrade their game by comparing their technique to that of a professional player.

Kittelsen describes the team as tennis “nerds”, watching all the games and following all the stats. But on a serious note, he insists the company is on a mission to democratise all the insight for everyone, and to do that, it needed a great tech team comprising machine learning engineers, AI experts, mathematicians, and experts in physics and physiology. Added to these are people with a proven track record in commerce.

Pedersen is adamant that with her company’s solution, insight into how to play tennis effectively can now be delivered cost-effectively and in a way that is comparable to the likes of Strava and Fitbit for runners. She notes that until now, a vast number of participants, without access to complicated technology, could play many games of tennis, yet not have any idea about how they had hit shots or how to improve.

The basic principle of the SportAI platform is that every movement a tennis player makes matters. After taking in video of tennis action – by using a standard mobile device such as an iPhone, sophisticated TV setups, or court-mounted cameras – the software uses machine learning and biomechanical analysis to build detailed 3D visuals of playing style. Once ingested into the SportAI system, data is uploaded to the Amazon Web Services (AWS) cloud, analysed and made available in seconds.

The app uses computer vision to check a player’s limbs and joints, tracking movement and the load on the racket during shots and the follow-through of the racket after the ball impact. It can measure biomechanics, swing curve, power generation and where the players hit the ball. It shows clearly the kinetic chain in making a shot – that is the sequence of shot creation from hip, shoulder and wrist position – generating an analysis from which it’s possible to see what needs to be improved.

For example, ball speed is a function of wrist speed, and the SportAI app can generate a swing curve, comparing it to that of a professional player. The AI within the app can display the velocity and rotation of hips and shoulders. All of this can be used by coaches to improve performance.

The subsequent data generated can be provided to individuals or to sports federations, academies, or equipment providers and manufacturers to see how people play and what can be done to improve technique. The data can also be compared with that of elite players to receive personalised improvement recommendations.

The analysis can also automatically jump to key points if there is something specific to focus on. Stats could include how many forehand shots a player hits in a given time, or they can generate highlights such as the longest rally in a game or action with the highest intensity.

“If you take a tennis lesson today, it might cost $100 an hour anywhere in the world. And you might have a good coach, [but even if] you had three or four good coaches looking at your serve feedback, there would be no data to back it up. Now, with advances in computer vision and machine learning, you could change that,” says Pedersen.

“So instead of having to have sensors on your body to track movement and biomechanical analysis, now almost every pixel on the video starts to become something you can use to track and gather data from, and then use that [data] to power different experiences and feedback,” she adds.

“[Manufacturers] are potential customers for us to take on this type of technology. Sensors themselves are just not scalable – you would either have to put them on a body or on a racket. It is not as scalable as being able to have a video that can come from a mobile phone or from court-mounted cameras, [which] are becoming more common around the world.”

The SportAI business model is mass market and relies on subscription, available to individuals, federations or equipment manufacturers. Kittelsen adds that manufacturers are particularly interested in the biomechanics information that the video can generate.

“[The video can] track the rotations, the speed and the height of the ball, the precision of the ball. [Manufacturers] do not have a lot of data on biomechanics, and so now we can help them with that. It’s not just looking at the result of hitting the ball; it’s looking at how you get that result, and how you improve the swing. And instead of then [just asking] players how the racket felt, we can understand [how they perform] with data,” says Pedersen.

From Hawk-Eye to AI

In an expression of the confidence it has in the system, SportAI says in testing, it had a player serving a ball and captured data using a standard phone with a standard camera at 30 frames a second at 1080px resolution. This had 98% precision compared with data generated using Hawk-Eye, the ball-tracking technology that is currently used at all the major tennis tournaments.

Yet despite the high-tech involved, Pedersen also emphasises clearly that the solution is for everybody. “This is not just about supporting the top, elite players, because the elite players will often have a performance analyst coach on their team who’s manually doing this on video and can deliver it. But the other 90-something million tennis players typically have no access to this data, so we want coaches and players around the world to get it,” she says.

“It’s sort of universal how you create power around [shots], and [knowledge of that] is something we see that recreation players and beginner players [would want]. It’s super motivating to want to get better. And when you have some ground data, you can go out and improve. People then want to go back on court because they want to get better,” adds Pedersen.

In terms of development challenges, the company says a number of business and technical issues have had to come together to get the company to where it is. In addition to gaining investment, the company has had to educate its market by showing coaches and players how they can use the technology and how it can be simultaneously better for both of them.

AI is becoming a commodity – everyone is using AI in some form. Yes, it can make mistakes, but you can still train it to be smarter and better. We see it as a tool to help and assist tennis coaches

“In all businesses, in all verticals, there’s scepticism. It was the same with Hawk-Eye. Ten years ago, nobody believed Hawk-Eye to be accurate enough. Now they’re accepting it. That’s going to happen with AI. AI is becoming a commodity – everyone is using AI in some form. Yes, it can make mistakes, but you can still train it to be smarter and better. We see it as a tool to help and assist coaches. It’s not taking their place, because, like you see in other industries, it becomes much more effective and efficient, and makes better decisions.”

According to Kittelsen, one surprise the company found using AI in its system was discovering its basic power, how just a single camera with coded AI algorithms can detect and display complex rotations and velocities. “But also, I want to add that the AI is still doing [some things] wrong, so we have to teach it. We have to teach the machine to take away the error percentages. And with the new cameras [on new phones], the quality of video goes up. The processors are faster.”

Acing video capture

The rest of 2025 will see SportAI rolling out the system for its first customers. The company believes it is being helped by tennis clubs increasingly mounting cameras around their courts, aided by the more powerful and cheaper cameras on phones, resulting in better quality video being more accessible for clubs and federations.

The company has also forged a partnership with the Matchi booking system for racket sports venues worldwide. Matchi currently manages about 15,000 tennis courts, 2,000 of them camera-enabled. SportAI will be taking in video streams from these courts to analyse action. It is also working with some equipment brands to generate technique analysis and offer equipment recommendations.

A key technical development for the company will be moving from cloud processing of data to performing data processing at the network edge. In addition to cost savings, this is intended to make it even faster to analyse data and add the capability to perform 3D video analysis. There will also be work on creating AI agents that can be attached to the app, which could be aligned to a federation or an individual player.

Pederson is adamant that SportAI is in business for the long run, and that the data the app picks up could also be useful for injury prevention and healthcare in general. For example, it could show how players’ joints bend and flag any extreme styles that could lead to injury. “Our vision is to democratise access to this type of data. It’s about seeing that value happen worldwide. We’re passionate about sports and technology. We want to see the most kind of progressive coaches, academies and brands taking it on board and really changing the game.”