The importance of analytics is on the increase across the football world. In England, Arsenal FC went so far as to buy the Chicago-based sports analytics company StatDNA and bring it in-house. Last November, Liverpool FC appointed as their sporting director Michael Edwards, who had arrived at the club originally as head of analytics.
According to Rob Esteva, director of The Stats Zone, we are still just at the beginning of the process of mining this rich seam of information. Esteva, who once worked with Matthew Benham at Brentford, says even the biggest clubs are still finding their way.
“Elite clubs are beginning to recognise the fact that the mainstream data available to them does not necessarily fully cater to their needs. I fully expect that more will follow the trend of building their own data collection systems and working to their own sets of definitions that suit their style of play, and players.”
“Historically you don’t have people with a mathematical background at football clubs; certainly on the technical side that is very rare,” he adds. “There are weaknesses within the data and a lack of expertise on the technical side helping us to interpret the data and make meaning out of it. It still feels like we’re five or ten years behind North American sports in terms of how we are using data for decision-making.”
According to Esteva, the long-term approach required for data – big samples take several years to accumulate – means this can be overlooked by some coaches, whose job insecurity leaves them thinking in the short term only. Yet his own work includes helping more curious clubs find specific answers. He explains: “We will go to a team and say, ‘What theories do you have that you want to support or blow out of the water?’ If we look at the high press, for instance, we’ll ask, ‘Are your players chasing the ball down efficiently from the time the opposition defender gets the ball? How many seconds does it take the nearest man to close him down? Is he positioning himself correctly so he can stop the defender from playing an easy pass?’”
Another example he proffers concerns the use of data in player recruitment:
“I worked on a project with Chelsea several years ago as they knew they had to replace Frank Lampard at some point and wanted to use data to look at how they could identify players who have similar attributes. You go through the process, look at which attacking midfielders score 15 goals a season, make five to ten assists per season, can run 11–12km per game and can average 70–80 passes per game. You can come up with a list of players that fit within that sphere but you’re looking at different leagues and different teams. That is difficult and there’s also the context of what happened around Lampard at Chelsea so you need to look at the data around two defensive midfielders sitting behind him that enabled Lampard to do what he did.
“So you have to look at what those around him did to allow him the freedom to do the attacking side of things. That was one aspect. Then you look at other countries and other players and, for instance Diego, the Brazilian attacking midfielder at Wolfsburg, had similar stats but in Germany, where he was in a very attacking team and where the average goals per game at that time was something like 2.9 per game. In the Premier League the average was about 2.6 and Chelsea were not creating to the same degree. So there’s nothing to say that Diego can do those things as well if you put him into a team that doesn’t have as many chances and doesn’t play as open. There are just so many variables but context is key and using techniques like cluster analysis to look at units within a team can be vital to project player performance when recruiting. “Everything with data and small gains is about minimising your risk,” he continues. “It is giving you a slightly better chance of succeeding. It might be 1% or 2% but you’re not changing the whole system. There’s no magic formula.”