Roland Garros: How the AI ​​will unravel the secrets of tennis

Technology and sport go hand in hand, and tennis does not escape it. Upstream of Paris Olympic GamesInria researchers worked on the best-tennis project to optimize service and service return by mixing expertise in biomechanics, neuroscience and intelligence. But, on a daily basis, in one of the largest tennis tournaments in the world, Tech and IA are busy behind the scenes.

Three axes …

At the controller of the Roland Garros Tech, we find Infosys, giant of the Indian digital, who took over from IBM in 2019, and is a partner of the association of Tennis Professionals, ATP, since 2016.
Over the years, the service company has deployed technologies for three targets: the general public, the media and players and coaches.
In 2023, Infosys thus launched a museum in the metarers in partnership with the International Tennis Hall of Fame, in which fans could find legends like Martina Navratilova. Currently, between two matches of the 2025 edition of Roland Garros, you can go to the Infosys stand, in the fan zone, to face virtual reality Rafael Nadal. In any case, his avatar, managed by an AI driven from data collected on the player.

The center match, a part of the iceberg

But, Infosys also offers a center match, or rather three versions of this function. One is intended for the general public and visible on display panels on the Roland Garros site, and also on the tournament website. It allows you to have before you a whole set of data collected in real time by Infosys, in particular thanks to the Hawk-Eye cameras, deployed on the courses. These cameras, which film at 340 images per second, make it possible to locate, by triangulation, the ball with an accuracy of approximately 2.6 millimeters.
Journalists have access to a slightly different version of the Center match and especially to a brand new comment assistance system, generated by AI, which allows them to put information into perspective, to give more context during the match. Finally, players and their coaches have access to an even more in -depth version of this data.
But for these professionals, Infosys has developed an even more advanced tool in partnership with the French Tennis Federation (FFT): Le Players Portal.

Portal Players: French egalitarianism?

Launched in 2019 and improved with many smart functions since then, it is a secure interface, without log, so that no one can follow the use made by authorized people. Within the Portal, available online, the player can watch videos of his matches, parallel with a set of data and statistics thanks to AI. A colossal work since it is based on more than a thousand parameters taken into account in order to offer optimal analysis.

The particularity of this tool, wanted by the FFT, is its opening. “This kind of statistics and analyzes are generally only available to players in the Top 10 or Top 100,” explains Raghavan Subramanian, responsible for the Infosys tennis platform. Those who “can afford it”, some of which can move with “a whole entourage, not only a coach, but also a physiotherapist, a training partner, a mental coach and a nutritionist, and even sometimes their own racket cord”, lists the information manager.
A minority which therefore excludes all those who are out of the Top 100, “who share a hotel room with their coach”, puts Raghavanianiananian in perspective, clearly passionate about his subject.
“The entire FFT objective was that the Portal Players and all the data analyzed is accessible to all players, without restriction,” said the information manager. Before adding smiling and in French: “Freedom. Equality. Fraternity. »»
The total opening, this egalitarianism immediately implies that any player can watch another player’s match, even if he does not play against him, to study his game, his statistics of first or second service, his propensity to close the exchanges quickly, the number of times when he was taken on the wrong foot, etc.

The difficulty of a relevant analysis in a tactical game

Data that can be analyzed after the match or crushed to reveal trends, and thus set up training objectives. Hence the importance of analysis via artificial intelligence, which requires a lot of training as the criteria vary according to the context, the type of blow. Dozens of situation that must be qualified so that AI can adjust its analysis.
For example, “a powerful forehand that lands in the background to a greater chance of winning more quickly if it is powerful and long. But if you make a cushioning in forehand, it must fall flush with the net, be very slow and crossed preferably. So the criteria of a winning blow vary depending on the type of blow and where the opposing player is. », Puts Raghavan Subramanian in perspective. Especially since these sets of “data points”, as the representative of Infosys calls them, are only a departure. To this small dozen points given (three per player and three for the ball), we must add the derivative data points, which can be due to the context of the point as well as to the way it was won or lost. From then on things are complicated in an exponential way.
Raghavan Subramanian first insists on the fact that AI analysis is not 100%exact, and that the optimization work is containing. Then he details: “If you were in defense but still managed to gain the point, we comply a set of points that will show two things. “The first is very subjective and consists in estimating whether the player was in a favorable or unfavorable position to win the point. The second relates to the tracks of potential improvements. »A kind of assistance for coaches, or in any case of putting into figures of what coaches often feel instinctively without formalizing it or being able to support it by figures. Thanks to the Portal Players, all players and coaches have in the eyes of statistical data, IA analysis and video.

More data to unravel the secrecy of tennis

At the moment, there is a set of data that only athletes control and keep, it is their heart rhythms for example, their level of formal level or the data collected via the sensors placed in the snowshoes.
Raghavan Subramanian would like to see even more sensors to settle on lessons and on athletes. If he measures the path traveled in a few years, he likes to draw up a “list of realistic wishes” of data that could further improve his tool. “I don’t have the player’s heart rate,” he begins. “I would like to know exactly where the ball touches the racket, at what angle … and then it is a flat or slightly cut contact. He enthuses. Before bringing another clarification which illustrates all its quest: “I know the speed of rotation of the ball, but it is the result, not the realization. »»

“You see a huge amount of sensors is required,” he admits more calmly, before resuming. “You know players often take time to warm up. And even with a warm -up before the match, initially, you will see that their arm is always a little folded at the elbow. You have to move forward in the game so that the arm is completely stretched, gives all its power. That’s all we want to see in our analyzes … “
He knows that it is impossible for the moment, that is why he has a more realistic list of objectives, of data he would like to be able to collect to further improve performance and game monitoring.

The ultimate dream of Raghavan Subramanian would be to be able to quantify “the difference between what the player wanted to do and what he really did, between intention and realization”. To unravel the secret of the moments of grace of the great players, in a way …

Want to save even more? Discover Our promo codes Selected for you.