Dutch tech firm transforms real-life football through video game concept

A Dutch company that was created when gamers thought the Football Manager computer game concept could work in real life helped to support the Belgian national team at the last World Cup in 2018.

Now, with the European Championship finals coming up this summer, countries are looking for ways to optimise their team’s performance – and data can play an important role.

During the 2018 World Cup, Dutch tech firm SciSports, which specialises in the analysis of football data, helped the Belgian team by providing analyses of its opponents. Giels Brouwer, CIO at SciSports, said: “It can be very interesting to see where the weaknesses lie in the team you have to play against, because then you can respond to them.”

But it is even more interesting for coaches to see their own team’s weaknesses, said Brouwer.

The concept for SciSports originated seven years ago at the Netherlands’ University of Twente, where Brouwer and two fellow students noticed that footballers were often scouted purely on the basis of a gut feeling by scouts. “We thought that could be done more efficiently, so we started to focus on data that could be used to support decisions,” said Brouwer.

The students were inspired by the video game Football Manager, in which the gamer takes on the role of technical director of a football club, making selections, buying players and playing matches with the right tactics. “I thought it would be nice to be able to do that in real life,” said Brouwer.

After an internship at Dutch football club FC Twente, where he analysed players and matches based on various datasets, Brouwer completed his studies in industrial engineering. Then, with two fellow students, he built a platform with the vision that data would become indispensable in analysing players and enriching the recruitment strategy.

The SciSports platform, Insight, contains data on 90,000 football players worldwide, which helps clubs and countries to find the right footballers for their team much more efficiently.

Brouwer said: “If you, as a club, want to have the greatest talents from Slovenia, for example, then a scout has to go there for at least two months to watch all kinds of matches and players. As a result, they see a lot of players who are quite good at playing soccer, but are not good enough for what the club is looking for. By looking at our database, you can focus on only those players that are good enough. That focus offers a lot of efficiency.”

Scouts, clubs and national teams can search the database using all kinds of variables, said Brouwer. “Suppose I am from FC Utrecht and I am looking for a striker who has the same qualities as Klaas-Jan Huntelaar. Then, on the Insight platform, I can set a search query for players similar to Huntelaar, with, for example, EU nationality and for under €5m. Then the players who meet these criteria roll out of the system.”

The platform is powered by data that SciSports buys from an Italian data provider, which employs countless people who manually collect data from almost all the professional matches played. “That is what they call a click farm,” said Brouwer. “These people record from where to where a pass is made, how a cross is made, what movement leads to a goal attempt, and so on.”

SciSports builds algorithms on top of that data so that users of the platform do not have to plough through all the data themselves, but can easily gain insights.

“For example, we measure the influence of a player on a team, and the influence of actions leading up to goals,” said Brouwer. Machine learning plays a major role in this, he added: “That’s the basis of everything.”

Three years ago, one machine learning expert started work at SciSports; now there is a whole team working there predicting the potential of players. The influence of an individual pass on a game and a team’s scoring ability is compared to 100 million similar passes in the system, said Brouwer. “We can see how many of those passes ultimately resulted in a goal, and therefore say something about the probability that a particular pass will contribute to a goal.”

Although a lot of the platform’s data comes from the click farm in Italy, in 2015, Brouwer set up another company, BallJames. This is a camera system that converts video into 3D data automatically, reducing the need for a click farm. “For the time being, this is not yet fully operational for us, because for BallJames we are still dependent on our own camera images and do not, for example, have cameras in stadiums in Colombia,” he said.