Forget set pieces – England’s Euro 2020 knockout game against Germany could be settled by spreadsheets. While the players have been practicing penalties and working on their pressing, behind the scenes, data scientists from both teams have been preparing for the match for months – building opposition dossiers, assessing fitness levels, and sifting through the huge volume of data that’s now available at the elite level of the sport.
Basic information on match events such as passes and shots has been available for more than two decades, but in the last five years new technology has made it cheaper and easier to track a player’s movement during a game, using sophisticated video processing algorithms; and in training, through the use of wearable technologies.
The number of data points generated by a single match has swelled from three or four thousand to up to four million, as it’s become possible to pinpoint the position of all 22 players several times a second. Add in information collected in training, and data on nutrition, hydration, workload and so on, and there’s an avalanche of material to work with.
“One thing teams have lacked historically is the capability to analyse and understand that data,” says Omar Chaudhuri from sports intelligence first Twenty First Group. As they try to move on from just “counting things” to actually understanding what that data means, elite clubs and national teams have embarked on a new transfer scramble, hunting not for natural finishers or great free-kick takers, but for data analysts with a background in maths and statistics.
For England’s last-16 opponents at Euro 2020, data is informing every aspect of the national game, right from grassroots youth football all the way to the top. The German Football Association (or DFB, for Deutscher Fußball-Bund) is using data to try and assess, for instance, whether it’s better for a young player to stay with a smaller club for longer or move to the academy of one of the big Bundesliga teams. Since 2004, it has been conducting regular skills tests with its young players, and now has a database of more than 175,000 players – it can compare the performance of a potential future star against current elite players to see what they were like at the same age.
As you go up through the levels, the amount of detail increases. “We have data on every run, pass, sprint, acceleration,” says Pascal Bauer, a DFB data scientist in charge of data science and machine learning applications. When it comes to match preparation, data is mainly used to speed up the work of video analysts. Instead of having to manually scroll through hundreds of hours of game footage to find key clips of how an opponent reacts to two-on-one situations, for instance, machine learning and data analysis tools can be used to automatically flag such incidents to the analysts. These tools are still in their infancy in football – DeepMind is working with Liverpool to develop AI tools for football further. Right now, it’s largely about making life easier for video analysts – “95 per cent of what we are doing is process automation,” Bauer says.
On Wednesday, June 22, before Germany’s final group game against Hungary, the DFB’s data team prepared preliminary reports on all five of the team’s potential next round opponents – as soon as the final whistle blew and it was clear they were playing England, a report was sent to the coaches so they could begin their preparations. Data helped the video analysts do their job more quickly in the short period between matches.
Match data is also being combined with training data (a technical challenge which it’s worked with database firm Exasol to handle) to provide assessments of a player’s workload and if they’re nearing the “red line” where they’re at greater risk of injury. That’s been particularly key this season because of the packed football schedule necessitated by the pandemic – Germany had some players from Chelsea who played in the Champions League final and arrived a week later than the rest of the squad, as well as players from Bayern Munich who had two weeks off, so there were lots of differing workloads to balance. “Data helps them make an informed decision,” says Sebastian Koppers, innovation lead at the DFB academy. “For us, it’s just about providing that dashboard so that they can make the decisions.”