Testing AI at Chelsea’s training ground: How innovative technology is being used to improve pros and scout new stars across the globe

Chelsea and Burnley are just two clubs from across the world using AI to find future Premier League talents – and FourFourTwo were invited along to Chelsea’s Cobham training ground to test out whether or not we have what it takes to compete with top-tier footballers (spoiler: we don’t). 

Stationed at the Chelsea training ground was ai.io’s state-of-the-art elite performance lab, aiLabs, used by partner clubs to drill their own players from academy age groups right up to senior professionals. Along with Chelsea and Burnley, all 29 MLS clubs are partnered with ai.io in order to test their own players and, crucially, discover unhidden talent from all over the world.

Through aiScout, grassroots players can upload specific drills from their location to a mobile app, which the aforementioned 31 teams can access in order to scout new players of different ages. The world’s first fully automated football talent identification and recruitment platform, ai.io is currently the only digital scouting product invited into the FIFA Innovation Programme. 

As Richard Felton-Thomas, director of sport science and COO at ai.io, explains to FFT, the technology acts as a facilitator for clubs to see a lot more players across the globe for less resource, because there’s only a finite number of scouts available at any given time. 

“Our role is to say, ‘there is a talent there,’ and then the club has to determine whether it’s worth scouting that player further,” Felton-Thomas explains. “What we’ve seen so far is it’s not necessarily about going out and having to travel to see that because you’ve got the video, you’ve got the data. 

“So if the player is the right age where you can sign them, then clubs do engage with that. But it’s very territory specific, very club specific.

“Interestingly, the multi-ownership model for clubs is changing how Premier League clubs operate their scouting networks now, because they can find these players from different countries, put them into other teams, monitor them, develop them and then when they’re ready to come to the UK, they can do it.”

Data is, clearly, a key component of unearthing new talent, and that focuses on what Felton-Thomas describes as “three different pillars of testing”: physical, cognitive and technical. Results are position specific, too, meaning goalkeepers and strikers aren’t judged on the same metrics, for example. 

Physical tests included jump height, abductor and adductor strength, as well as an overall force measurement; cognitive tests were completed on a computer and related to picking a specific object (control) and selecting whether a certain shape appeared on screen (focus); and technical tests related to reacting to a green button with both our hands and feet, testing the agility to respond to certain stimuli.

Cognitively and physically, we performed well. Technically… let’s just say improvements can be made.

This, Felton-Thomas highlights, is where the AI truly works its magic. Separating all of the numbers and presenting it in a clear and coherent manner ensures certain attributes aren’t missed or, in some cases, are purposefully disregarded.

“We normally find that if players are devoid of one of those, they’re really good at the other two,” the ai.io COO explains. “For example, someone like Sergio Busquets’ cognitive testing would be off the charts. We know his technical ability would be off the charts. But even if his physical results are lacking, it’s actually on the club to still use their scouting skills. If they go into the database where we’ve collected data for say, the future Busquets, we’ve got all of his statistics and performance scores. If a club is specifically interested in power and pace, they will probably miss him.

“But another team might say, ‘actually, we don’t mind at this stage – he can learn that later,’ or ‘that ability isn’t a feature of our game so it doesn’t matter’. There is still that human input for scouting – even when all of the data is on the computer that we’ve collected for them, clubs still have to think about what they search for. 

“Going forwards, as we have more successful players come into the system, we will start to guide the teams around certain players. For instance, we might show them a player they wouldn’t normally look at, because their pace isn’t great for example, But everything else is so good that they shouldn’t miss the opportunity to give them a chance. Eventually, the AI will start to make some different recommendations.”

So FourFourTwo have a chance of impressing Chelsea without scouts having watched us play then? Maybe not.

“We’re not here to replace jobs, we’re here to work with the scouts and help them become more efficient, while also giving them insights into players that they might not be able to get otherwise. The key thing is that they’ve got their eye tests, but we also have data that can help those decisions.”

*Read 5 free articles per month without a subscription

Ryan is a staff writer for FourFourTwo, joining the team full-time in October 2022. He first joined Future in December 2020, working across FourFourTwo, Golf Monthly, Rugby World and Advnture’s websites, before eventually earning himself a position with FourFourTwo permanently. After graduating from Cardiff University with a degree in Journalism and Communications, Ryan earned a NCTJ qualification to further develop as a writer while a Trainee News Writer at Future.