Hackathon explores AI-based match analysis

5 March 2024 – Artificial intelligence (AI) and Machine Learning are advancing at breathtaking speed, and DFL Deutsche Fußball Liga is supporting many initiatives to exploit the potential of these technologies for football match analytics. This year, a hackathon hosted by the Munich Data Science Institute at Technical University of Munich (TUM) under the leadership of Prof Dr Daniel Link investigated new opportunities to interpret match data using AI. DFL accompanies the project as co-initiator, partner and sponsor.

Five tough challenges

Under the heading “Analysis of spatio-temporal player and ball data of Bundesliga matches”, the participating teams were presented with a choice of five challenges. All tasks called for the development of models capable of identifying and analysing one of five specific match situations using Machine Learning and AI:

  • Line breaks (breaking through the opponent’s defensive line through a pass or dribble)
  • Quality of ball reception (predictive ball handling to set up the next action)
  • Winning the second ball (after a launch)
  • Off-ball player behaviour after turn-over
  • Prediction of the probability of certain events in e-football

“Nine teams of five members each took part in the hackathon,” reports Professor Link. “The teams received video recordings from five Bundesliga matches, along with detailed descriptions of the challenges and a compilation of position and event data from the same matches.” They entered a one-week retreat on the TUM campus to develop mathematical models for automated identification of the five predefined match scenes.

The members of the competing teams – all advanced students, some of them in doctoral programmes – normally specialise in developing industrial AI applications such as autonomous driving technology. The hackathon was a welcome challenge for them to test their know-how in a totally different field.

The quest for smart models

“The models developed during the hackathon were supposed to be able to identify the specific situations during a match,” explains Dr Hendrik Weber, Head of the DFL Sport Technology & Innovation department and member of the supervisory board of Sportec Solutions AG, who was involved in the project as a juror.  “For example, for the first challenge they had to develop code capable of identifying the ball-possessing team, the attack and defence lines as well as a break through the defence. Some of the teams did quite well in accomplishing that, and a few even built a visualisation demonstrating the capabilities of their model.”

Based on nine assessment criteria, the jury selected the three best-performing teams. These finalists are now preparing their solution proposals for presentation at this year’s SportsInnovation in Düsseldorf. First, second and third prizes will be awarded during the sport technology event on 20 and 21 March.

Visions for the future drive innovation

“A hackathon naturally doesn’t produce market-ready products,” explains Weber. “Rather, it is an explorative endeavour where DFL joins hands with science to test the potential of current technology. When it turns out, such as in this case, that the results are quite promising, it makes sense to pursue them further with a perspective to turn them into an actual product eventually. What is more, amid today’s struggle for talent this hackathon is a great opportunity to stir the enthusiasm of highly competent university students for a career in the sport technology segment.”

There is no shortage of options to put such solutions to good use. In a not too distant future, experts hope to use AI solutions to identify technical and tactical constellations that could help commentators and analytics experts “read” the game. Says Weber: “From the sport perspective, this means recognising configurations that could result in an advantage for the players; and from a scientific perspective, the idea is to make progress in decoding the DNA of football.”

As an organization with a strong focus on innovation, DFL considers competitions such as this one as an opportunity to steadily develop the practical benefits of technology for the sport world and bring them to fruition in as many applications as possible.