The business of football: the big data arms race

You can enable subtitles (captions) in the video player

Burnley Football Club training ground, host to the annual International Youth Tournament.

And nestled next to the pitches is a lab that is part of the wave of change that over the last decade has been transforming every aspect of the sport.

All of the players across the course of the weekend are going to come into the lab, are going to do a series of testing, and we’re going to give some insights.

Like Formula One, basketball, and cricket, football has embraced the world of big data. Teams are using technology to gain the edge in all areas of the game. Stadiums across Europe are covered in cameras tracking players’ every movement. Wearable sensors and GPS units monitor their overall physical performance, meaning each 90-minute game collects over a million data sets, giving backroom staff keys to possible success.

Huge amounts of money are flowing into the sport, and with financial rules putting the squeeze on spending, the search for talent has become a numbers game. AI-based sports science company, ai.io, is using data to help teams identify the perfect signing.

I think there’s almost 300mn football players worldwide who are all looking for an opportunity. They just want to be scouted. They want to be seen by a club. And actually, that process requires a lot of luck. You need scouts to come to your game, to watch you, to have a good day. And we saw this world where… how do we put the power into the players’ hands?

Budding players upload videos of various drills, and using their technology, ai.io is able to evaluate their abilities.

When we’re analysing a video we’re looking for, basically, football and athletic-based metrics. So how fast can you move? How high can you jump? How quickly can you change direction? If we’re putting a ball in the scene, then how well do you dribble? Use your left foot, your right foot? With this AI-powered computer vision the video goes to the cloud. That’s where we start to work with Intel, use a lot of AI-powered compute. It analyses all the movement-specific characteristics. Where do they move within the scene of that video?

What’s really important is we have professional benchmarks so we will analyse all the players that we work with across the clubs. So we analyse the Chelsea players and those Burnley players again. And that creates a unique scoring system for each of those clubs, so it makes it really meaningful for the players to get some feedback. It also makes it really meaningful for the scouts. And they’ll see a score that’s relevant to the players that they already know about within their own organisation.

You know, there was no data analytics, anything like that. There was no artificial intelligence. It was experienced operators who knew what talent looked like. And it was all very subjective. Each football club will probably have their own metrics on how they would be looking to benchmark players, but ultimately it’s potential. And that’s the most difficult thing to identify, is you can see what a player has at that time, but it’s trying to identify what they’re going to look like in the future.

The amount of money flowing into football from TV and sponsorship has never been higher, yet across the world few clubs make a profit. Instead, high revenue goes straight into player wages and transfer fees. To help address this clubs are increasingly investing in youth development in the hope of creating, rather than buying, their future stars.

It gives us the ability to measure a player with objective information. And what it is is a really beneficial tool for Burnley Football Club because academy players can go into the lab. They’ll go through a set of series of tests.

Thank you. What position are you in, Maurice?

Right wing.

Right wing? Under 14. So if you now want to go on the computer, I’ll be with you in one minute.

We get information back on what they currently are in terms of their levels. And what then that allows us to do is to develop an individual learning plan for that player over the coming months.

Good scores. And we move on to the next test.

We know what the player needs to work on. And hopefully within that four-month period, as an example, they go back into the lab, and it’ll inform us of what the progress of that player has made.

Come on. Big jump. Up, up, up.

I don’t think that we’ll ever not have scouts the traditional way.

Big jump. Jump.

It’s just another aid and another tool in your recruitment process.

Grassroots and amateur it’s always been difficult because you don’t really have much opportunity to collect data unless you’re in a professional environment. We’ve got all the longitudinal data, so as we collect every single year, every single time we work with our clubs, the power of that predictability increases.

As hard as you can, all right?

Now we’re getting this data. I think we’re getting millions in a mobile phone? Then we’ve got our labs adding to the pro data, is that level of predictability of talent, but I think also helping support the talent going, are you in the right position? Are you in the right sport? Giving them other opportunities to navigate sport in life. You might enjoy this a little bit more because you’ve got a great profile match to it.

Lead with your hips. Good. Pull up. Pull hard.

Every football club in the land is trying to find that next diamond. If you’re open-minded to the tools that are around, it’s only going to increase the football club’s ability to recruit players as well. You know, when I started many years ago we didn’t even think about it.

Green, green, green. Good. Speed, speed.

And actually, even, I would say 10 years ago, data analytics was just becoming a real prominent part of recruitment process. We don’t just use data now and AI in terms of recruitment. It’s also used in match days and game models, and analyse the opposition. It’s just a broad spectrum. So at the top level I’d be amazed if nobody’s using it.

As football looks to get a grip on runaway spending with tighter rules, no club can afford to waste tens of millions of pounds on signing the wrong player. That’s where many hope that big data can help. For elite teams finding the very best player for a particular position can be the key to unlocking a title-winning season.

This is why they brought me, of course.

And for smaller clubs, making more limited budgets stretch further has never been so important. Many are also looking to unearth a hidden gem who can be sold on for what can sometimes be huge profits.

In football money talks, and the best way to predict the league is to look at the income and the wage spending of teams.

65 per cent of revenue spent on player wages, and that varies by the size of the club. The general pattern is that larger clubs with larger revenue spend a slightly smaller fraction than 65 per cent, and smaller clubs spend more than that. In terms of transfer fees, about 25 per cent of revenue goes on transfer fees.

Ian Graham was Liverpool FC’s director of research until 2023 and has attributed the team’s Premier League title in 2021 to their use of data.

The question is, how do I get more value per pound that I spend? A transfer can go wrong in many different ways, even when a player looks really good, and even when the data analysis says this player will be really good, that transfer can go wrong.

World’s most expensive keeper.

If you look at the history of Premier League transfers, 50 per cent of the transfers fail, so the efficiency is really bad. And data analysis can help with that because we can benchmark and measure players on the same scale as one another. Then in the past 10 years or so tracking data has become widely available.

For teams in the professional game, in big leagues, like the Champions League or the Premier League, the league pays for that data collection. You can see 29 co-ordinates per player, so you know the positions of their ankles, their knees, all of their joints, even their head and their eyes. You can infer which direction a player is looking in. So it’s really gone from a small, one-man-band sort of business to a business where you need a team of software engineers.

Over the last decade data analytics in football has grown, but many are still struggling to use it to gain a competitive edge.

Lots of teams have got their own data departments, and they invest in this area. Every investor has got access to stock market prices. Isn’t that a level playing field? But not everyone gets to be Warren Buffett, right? And so the difference is in the discipline and the implementation. So a few teams have really implemented a decision-making process that’s based on data. So Liverpool, from the ownership and from the sporting director, it was demanded that the club would be evidence-based, and data analysis for understanding player quality was a big part of that.

Brentford, owned by professional gamblers, who made their fortunes by understanding data analysis of sports teams and players, and they demand a data-driven approach. So with the revenue that really shouldn’t have supported them being able to rise to the Premier League, they managed to do it.

European football has had an influx of American owners, many of whom have seen the impact of the data-focused approach found in most American sports. That may mean that football clubs will increasingly be competing off the pitch in a technological arms race.

The future of technology – football – I think, obviously, there’s lots more still to come. I think what we’ll see more and more, and I do believe we’re at the forefront of this, is actually combining those AI tools with that human intuition and skill.

More teams are realising the importance of data analysis. It’s difficult because data is not traditionally an expertise that you see at clubs. Clubs find it difficult to hire talent. They find it difficult to understand how to implement decision-making, using data into their processes. But there’s certainly been an explosion of interest because of those success stories that we’ve seen.

In terms of the future, you’ve already seen the transfer market get a little bit more rational. I think in order to compete, teams would be forced to adopt more of a data-led approach because you need to be as efficient as your rivals.