Who was the fastest player? Who had the hardest shot? And how many duels did my team win? Statistics have been around in sport for many years, but artificial intelligence and machine learning are being used to find new methods to process this data innovatively.
Performance analysis has developed remarkably in recent years and evolved into a veritable big data science. In particular, immense amounts of information are now meticulously recorded in the top leagues at every game. This data is not only used to answer fundamental questions such as the speed of players or the effectiveness of shots but also provides in-depth insights into the tactical and physical aspects of the game.
Prof. Dr. Martin Lames, Head of the Chair of Performance Analysis and Sports Informatics, gave a keynote speech on the topic of artificial intelligence in sport as well as the advantages and disadvantages of its implementation and its prospects at the 11th International Conference on Sports Sciences Research and Technology Support by icSPORTS, which took place in Rome from September, 16 to 17 and was attended by participants from over 20 countries.
"The current development in sport shows a constantly improving data situation, whereby a differentiated view is necessary. Especially in commercially orientated sports we find a high abundance of data. However, this contrasts with less commercial sports such as breakdancing or lacrosse, the new up-and-coming Olympic disciplines, which do not have a comparable big data situation. In lacrosse, for example, we have already started studies to prepare for the challenges of a new Olympic sport," explains Prof. Lames.
However, the training scientist also warned in his presentation against the improper use of technology. For example, the media would exploit statistics to generate entertainment and clicks, in contrast to the scientific standards of precision and validity. Applications for predicting games or match situations are also not in the primary interest of sports science, as the practical relevance of the applications of so-called "benchmark improvements" must be more or less denied. In principle, however, the development is positive: "I'm pleased to see that sports data is increasingly finding its way into basic research in computer science to make progress in the fields of artificial intelligence and machine learning. This development also harbours the potential of new ideas and findings for data analysis, at least for us," adds Prof. Lames.
The "silver bullet" in the utilization of sports data and machine learning ultimately lies in the "right" questions and interdisciplinary collaboration. "We need the skills and expertise from computer science, but also the right input from sport," explains Prof. Lames. The range of scientific topics, from pattern recognition in moves to the sensible application of predictions and the identification of players' behavioural tendencies, particularly relevant for scouting and preparing for matches, illustrate this potential.
Prof. Lames believes there is no danger of sport being appropriated by the data: "The scientific basis in sport is increasing - I don't think that's a bad development. The quality of the information teams can access is also increasing, which means that benefits can be derived from it. However, sports will always remain unpredictable and incalculable. A famous example is the question: where is the ball five seconds after a corner? It can range from a goal kick to a counter-attack. The great thing is that nobody can predict this."
To the homepage of the Chair of Performance Analysis and Sports Informatics
To the homepage of the icSPORTS-Conference 2023
Contact:
Prof. Dr. Martin Lames
Chair of Performance Analysis and Sports Informatics
Georg-Brauchle-Ring 60/62
80992 München
phone: 089 289 24500
e-mail: Martin.Lames(at)tum.de
Text: Bastian Daneyko
Photos: ?