Artificial intelligence is now powering football clubs in ways few could have predicted
When Chelsea’s owners began a complete overhaul of staff and operations at the club, shortly after taking over in 2022, they identified data science as an area the Premier League was significantly trailing American sports.
NFL and NBA teams typically have large data departments containing double-digit numbers of experts and analysts who work with vast amounts of statistics to improve performance.
With Todd Boehly’s experience of processes across the Atlantic, as co-owner of the Los Angeles Dodgers and Los Angeles Lakers, and Behdad Eghbali and Jose Feliciano making many of their billions through tech investments, they funnelled knowledge into reimagining Chelsea’s data department, and The i Paper has been told the club is now leaning heavily into advancements in artificial intelligence.
Using one complex system they can assess every player in each position, using hundreds of data points, and make recommendations for who fits best together, based on strengths and weaknesses, in different formations and tactical systems.
Two recent hires have been key to developing the analytics side.
Sachin Gupta, a 42-year-old with a degree in computer science from MIT, joined the club last November to become chief analytics officer after almost two decades working for NBA teams. It was an appointment that came as a surprise to many in football circles.
Gupta is renowned for creating the NBA Trade Machine, a hugely popular platform that simplified basketball’s complex trading rules, while at ESPN, before moving into basketball teams, where he grew a reputation for combining the data side with a deeper understanding of the player and coach side – traditionally two conflicting areas.
“The smartest guy in the room” was how Minnesota Timberwolves general manager Scott Layden, who Gupta worked closely with, described him.
In January, Javier Fernandez, considered a world-leading mind in AI and football, joined the department as director of data science. Fernandez has a PhD in AI and worked at Barcelona as head of analytics for five years, spending the last four at Zelus Analytics, described on its website as “the world’s leading sports analytics platform”.
At Barcelona, Fernandez is credited with making data and AI a part of the very fabric of the Spanish giants in what is considered pioneering work. He used AI to change how they look at players and analysis and it even influenced tactical decisions.
Chelsea co-owners Todd Boehly and Behdad Eghbali are big proponents of AI (Photo: Getty)
In an excerpt on Barcelona’s website it stated that he was “involved in a continuous process of mapping football concepts and the club’s playing philosophy into actionable algorithms”.
The club won two La Liga titles while he was there.
At Zelus, one of his key areas was developing complex models to value players and teams.
“Working alongside some of the sharpest minds in sports analytics was truly transformative,” Fernandez posted on LinkedIn.
“It’s remarkable how much you can learn just by being immersed in a highly creative environment and surrounded by brilliant people [is that osmosis?].
“At Zelus, we pushed the boundaries of sports analytics across soccer, basketball, and American football.”
On Thursday, Fernandez shared two job opportunities for a full-stack web developer and data engineer which would “be at the heart of decision-making” at Chelsea.
AI is transforming the world of professional sport (Photo: Getty)
Some of the AI capabilities in football The i Paper has learned about from industry insiders are astonishing.
One AI platform working with Premier League clubs even claims to have “cracked the code” for promotion from the Championship to the Premier League.
There remains deep scepticism across the sport , but proponents of the technology insist that it is a brilliant tool that can assist head coaches, rather than replace them.
“We compare it to an X-ray,” Jan Wendt, founder of AI company PLAIER, tells The i Paper.
“Imagine the time before the first X-ray was delivered to a doctor. We’re not replacing a doctor. We’re just delivering one tool that helps him make a better diagnosis.”
Our test: Cole Palmer on the bench?
Away from the elite sport supercomputers, I wanted to find out how knowledgeable the freely available commercial AI systems are.
I put the same question – “What is Chelsea’s strongest starting line-up?” – to three of them.
ChatGPT, which sparked a trillion-pound tech arms race, was abysmal.
It included a player who left in the summer, a player out on loan, and a player currently suspended for failing a drugs test.
4-3-3: Sanchez; James, Silva, Colwill, Cucurella; Caicedo, Fernandez, Sterling; Madueke, Jackson, Mudryk.
Cole Palmer was on the bench. As was Ben Chilwell, currently on loan at Crystal Palace, and Kepa Arrizabalaga, who hasn’t played for Chelsea since 2023 and is currently on loan at Bournemouth.
What Chelsea’s starting XI should look like, according to ChatGPT (Graphic: The i Paper)
Copilot, Microsoft’s AI offering, was only marginally better.
4-3-3: Sanchez; James, Fofana, Badiashile, Cucurella; Fernandez, Caicedo, Palmer; Madueke, Nkunku, Mudryk.
Another AI currently unable to perform a quick Google search to discover that Mudryk is suspended.
Meanwhile, Google’s Gemini, after initially reluctant to offer an opinion, opted for a 4-2-4 formation.
4-2-4: Sanchez; James, Chalobah or Fofana (it couldn’t decide), Colwill, Cucurella; Fernandez, Caicedo; Palmer, Madueke, Neto, Jackson.
This is, of course, merely a bit of fun.
The systems and models – or “tools”, as those who build them refer to them – being used at clubs are vastly superior and targeted than the popular large language models.
Finger pricks and transfers
Teams rely on AI to optimise player performances (Photo: Getty)
AI is seeping into football everywhere.
Championship side Norwich City have one of the smallest staff teams of a category one academy but leverage AI to streamline efficiency.
They work with Hudl, an AI platform designed to enhance team performance, using their “Focus Flex” cameras that automatically follow action in training and upload it live. Combined with internet cameras and drones, by the time a player is back in the changing room they can review their performance in the session on a smartphone.
Premier League clubs are working with Orreco, a biostatistics company, that via simple finger prick blood tests can provide a detailed breakdown of health, rest and recovery. One club doctor said they found a 74 per cent reduction in soft tissue injuries using the technology.
The i Paper revealed last year that Liverpool have collaborated with Google Deepmind to predict how opponents will defend corners and devise the best tactics to score from them. Meanwhile, Manchester United are thought to be utilising AI in recruitment.
At PLAIER, Wendt broke down how his system works in a three-stage process.
He is reluctant to share who he is working with, beyond that it is Premier League and Championship clubs and others across Europe, insisting he doesn’t want the achievements of the coaches to be overshadowed.
But as an idea of their reach, in the January transfer window PLAIER was, he says, involved in 163 transfers, of which 103 were completed, all over the world.
1. They measure a league and club and tell them how good they are in comparison and what they need to do to fulfil their sporting goals. This includes measuring the coach’s performance.
2. They analyse the squad and reveal strengths and weaknesses.
3. Based on the results of the first two, they help recruit the right players.
“We make football measurable and then predictable,” Wendt says.
“In our system we have over 370,000 players worldwide, we buy every bit of data we can get, from every data provider.
“A footballer is probably the most transparent employee in the world, except for pilots and astronauts.
“Per player on average we get roughly 140 data points per minute, in event and tracking data. Combine them to derive a score – one figure for the player that very clearly says where they can play around the world.
“There’s so much data a human can’t compute it. An average Premier League player has a score of 5,000, the best players above 9,000.”
The technology can also provide a detailed breakdown of health, rest and recovery (Photo: Getty)
He adds, in an astonishing claim: “We cracked the code. There are five points you need to fulfil to be promoted from the Championship to the Premier League. We can clearly address them.
“If you do these five things, and you have the basic financials in place, you will be promoted to the Premier League. It’s all hidden in the data that’s already there.”
Silos and rewards
As it stands, clubs are in various stages of toe-dipping. Maybe trying one bit of technology, trialling another in the academy. But most operate in silos.
It is believed that when everything is combined into one mammoth system that permeates every area of a club and threads through its culture, that they will achieve spectacular results.
Several people working in the field in various roles at Premier League clubs told The i Paper they are convinced the first team that goes all-in on AI will reap the rewards.
“I fully agree,” Wendt says. “The first one always has a major advantage with almost no risk.
“In the end you quickly see if it works or not, then there will be a tipping point where you have to use it, if you don’t it will be a disadvantage. That will come.”
PLAIER has a team of around 15 people, mostly data scientists and AI specialists.
“In football there’s scepticism – I fully understand,” Wendt adds.
“Everyone working there has been trained by the system for decades. If you embrace AI you have a major advantage now, there are many hidden diamonds and insights on how you set up the team.
“We’d love to get one of the underdogs in the Premier League to qualify for the Champions League, and to get clubs from the Championship to the Premier League. The ones you never expect. That’s our mission.”