Is football so chaotic and random that it can only be enjoyed, but never understood? Or is it a problem that can be worked through and solved? Those are extreme positions but the debate itself is real, and is more contested than ever before.
Other sports are increasingly understood, analysed and predicted through numbers. And yet in football we are still told that only one statistic counts. Or that a statistic has never made a save, or scored a goal.
But neither the difficulties of football analysis nor the cultural resistance to it have put off some people. Because those very barriers only make the potential benefits gained from insightful football analysis far greater. Anyone who can build a model that can even start to understand and predict football has something very valuable on their hands.
There is a new generation of football analysts whose attempts to predict outcomes have been tested on betting markets and then sold to clubs. Analysts like Rory Campbell, founder of C&N Sporting Risk, for whom the difficulty of analysing and predicting football only makes it more attractive.
Campbell was at the Sloan Sports Analytics Conference in Boston earlier this year, listening to a discussion of golf analytics. “Someone said that golf is the sport most suited to analytics,” Campbell recalls to The Independent. “And I thought what that meant is that it is the easiest to solve, in terms of helping you get to the right answer.”
Football, with far more variables than golf, is the opposite. “Football is hard,” Campbell says. “But saying that it is not suited is missing the point. In my mind, hard is good. Because there is the most opportunity to gain a competitive advantage if you use it right. This is why poker players make more money than chess players. Because in chess the best player almost always wins, and in poker there’s variance and luck, and that’s a good thing.”
The challenge, then, for the next generation of analysts is to break the game down and measure it in such a way as to better predict what is likely to happen next. “It’s impossible to solve, impossible to get a 100 per cent answer, because of randomness and variance in the game,” Campbell admits. But how do we get closer?
“You’re trying to more accurately predict the percentage chance of things happening, more than someone who is just randomly picking,” Campbell says. The key is to know what to look for. “There is a big difference between data and analytics. Data are pieces of information – not necessarily numbers – about things that happened in the past. Analytics is using that information to help better predict the future. So if the data doesn’t have predictive utility, it’s useless. I don’t want to know happened in the past because it happened, I want to be better at predicting what is going to happen in the future.”
The big problem for football used to be the data itself. There simply was not enough information kept on what happened on the pitch. Not just goals and assists but everything else. But that is changing, with x,y coordinates of every action on the pitch now being recorded. It is not very hard to get hold of all the data on years and years of football. Analysts just have to ask the right questions.
Campbell runs through a list. “Are shots more likely to have a higher % of scoring if they come after a cross, a through-ball, a counter-attack or a long ball? Or are passes more valuable if they go directly forward or at an angle? Are teams more likely to play better when they’re winning, losing or drawing? Is it better for these 11 players to play long-ball, counter-attack or possession football? All these kind of things. You’re trying to test the premise, to get closer to solving a problem.”
Once these questions are tested it is easier to predict what will happen next. That is the root of much successful sports betting, which is what Campbell, now 29 years old, did while taking his football coaching badges after graduating from Oxford University. “I always had that focus that having analytically-based thought processes and strategies were the best way to maximise competitive advantage, in whatever field you’re in.”
And while this information can be useful to clubs about individual players, it can also help them to think more clearly about what they are trying to achieve. “Numbers enable you to define things. If I have a football club and our objective is to get into the Champions League, and finish in the top four of the Premier League, there are loads of things that are going to control whether or not that happens. It is too big a goal to take on its own.”
The goal has to be broken down. “So how many points do we need to get into the top four? We can work that out. Then how many goals are we going to need to score? What is our goal difference going to be? Then let’s drill down to think about exactly how many shots we are going to need to have, and how many shots we are going to concede. And work back from there. Get into the tangible elements of understanding a much bigger objective, which we couldn’t understand on its own.”
Not enough football clubs think like this, but it has been very successful in other sports. Campbell points to the example of how Team Sky won the 2012 Tour de France, breaking down what they needed for Bradley Wiggins to win the tour.
“Team Sky profiled that year’s Tour de France, with historic data of requirements on the climbs and the time trials. They knew it was an opportunity for competitive advantage, because it had never been done before. So they plotted a power curve, with time along the x-axis and power on the y-axis, marking what power output the Tour winner would need and when. Then they looked at Brad’s curve, looked at the gaps. And they had the techniques to bridge them. The analytics are what enabled them to the understand the difference between where they were, and where they wanted to get to.”
The most famous example from sport is the story of the Oakland Athletics told in ‘Moneyball’ but even that, Campbell argues, is not so much about numbers as it is about thinking clearly. “’Moneyball’ wasn’t a data thing, it’s about understanding,” he says. “We need to better understand what we’re looking at. ‘Moneyball’ is having an objective strategy, understanding where we are, where the inefficiencies are, and how to bridge the gap. Understanding what the objective is, and strategically, how are we going to get a competitive advantage.”
Slowly but surely, this analytical thinking is coming to English football. Campbell has advised West Ham and the successful transfer policy of Liverpool – Roberto Firmino, Sadio Mane, Mohamed Salah and Naby Keita – bear the marks of a data-driven approach. Mane had been lauded by the analytics community while he was still playing for RB Salzburg.
Then there is the fact that two successful clubs are owned and run by men who have made their money in betting, and applied those principles: Matthew Benham at Brentford and Tony Bloom at Brighton and Hove Albion. “Two very big betters have bought football clubs, and if something works for you in one area you tend to apply those tendencies to another. From betting you become very good at using underlying profiles to predict performance, and that translates into knowing why teams are good or bad.”
English clubs are starting to take advantage of these insights and the evidence suggests there are rewards for whichever teams move first. All it takes is for data-sceptical clubs to realise that just because they are not being offered a total answer, it does not mean they are being offered no answer at all. “If you’re saying you want to solve the problem, your mindset is wrong. Are we going to solve it? No. Are we going to understand it better? Yes.”