Diksia.com - Google DeepMind is testing the use of artificial intelligence (AI) to create tactics in football.
They unveiled the TacticAI prototype last March, developed with Liverpool FC as part of their three-year collaboration with the team that competes in the English Premier League (EPL).
The results of this research were recently published by DeepMind researchers in the journal Nature Communications, as quoted by detikINET by Business Insider on Monday (4/15/2024).
According to football experts involved in the project, the AI system used is better than humans at developing corner attack and defense tactics after the AI was trained on a dataset of 7,000 corners from EPL matches.
“The most interesting thing (about football) to me is that it is a game that sits somewhere between art and science. There is a lot of uncertainty in the game, but we can still use data to make better decisions,” said Zhe Wang, one of DeepMind researchers on the TacticAI project.
Meanwhile, Petar Velickovic, one of the researchers of the DeepMind project, also said that it is logical to train TacticAI using data from corners as this usually leads to a more structured part of the game.
“You can never really predict what will happen. But what you can do is recognize patterns in tactics. Corners are a good example of this because they are cumbersome, frequent and lead to scoring chances,” he said.
TacticAI can advise coaches on optimal player positions when attacking and defending corners and has the potential to give teams an advantage to make a difference in this competitive sport.
DeepMind and Liverpool’s development of TacticAI was praised by legendary “Moneyball” baseball manager Billy Beane. Since its takeover by Fenway Sports Group – owner of the Boston Red Sox – in 2010, Liverpool has employed a number of employees with backgrounds unusual for an EPL soccer team.
For example, Ian Graham, a physicist from Cambridge, who will lead the Liverpool research team until 2023.
“For 10 years there have been people in Liverpool with PhDs from Cambridge and Harvard – these are people who rely on engineering, data and research and have been working with statistical analysis for a long time. So you already have a good and solid foundation for development with deep learning,” Velickovic said.
