In football, corner kicks are important moments because they offer an immediate opportunity to score a goal and provide coaches with the most direct opportunities to intervene and improve. In practice, routines for these corner kicks are determined before each game, so a system that would help analyze and improve the chances of scoring would be beneficial for human experts.
TacticAI can predict the outcome of corner kicks and offer realistic and accurate tactical suggestions in football matches. Now, researchers from Google DeepMind present in the journal Nature Communications an artificial intelligence (AI) system, called TacticAI, that can predict the outcome of corner kicks and offer realistic and accurate tactical suggestions in football matches. According to the authors, this tool “incorporates a predictive component and a generative one, allowing coaches to efficiently sample and explore alternative player configurations for each corner kick routine and select those with the highest probability of success.”
In the tests conducted, TacticAI was able to accurately predict the first ball receiver after the corner kick and the direct result of it. The system was also able to assess possible outcomes with various player configurations and generate tactical variations that improved match results.
Data and experts from Liverpool FC
The authors trained and tested TacticAI with a dataset of 7,176 corner kicks from past seasons of the Premier League provided by Liverpool FC. They then used deep geometric learning to identify key strategic patterns in order to produce predictive and generative results. They demonstrated the advantages of this tool and found that the tactical configurations it generated were realistic and indistinguishable from real-world scenarios evaluated by a group of five football experts: three data scientists, a video analyst, and an assistant coach from Liverpool FC.
TacticAI’s model suggestions are indistinguishable from real tactics and, moreover, are favored over existing tactics 90% of the time by several experts. A survey conducted with these experts from the English club shows, according to the study, “that TacticAI’s model suggestions are not only indistinguishable from real tactics, but are also favored over existing tactics 90% of the time, in addition to offering an effective system for corner kick recovery.”
The authors suggest that this method could be generalized to other set plays, such as throw-ins, and also to other team sports with set-piece situations. This research from Google DeepMind – the same company that developed a program that beat humans in the game Go – could lay the foundation for the next generation of AI assistants to help coaches determine optimal player configurations and develop counter tactics that maximize their chances of winning.
via: MiMub in Spanish
