Google DeepMind has developed a soccer tactical AI called 'TacticAI' in collaboration with Premier League powerhouse Liverpool, which can increase the chances of scoring in corner kicks and reduce the probability of conceding a goal.
Liverpool, the prestigious English soccer club that Japan national team player Wataru Endo plays for, is known for its proactive
TacticAI: an AI assistant for football tactics - Google DeepMind
https://deepmind.google/discover/blog/tacticai-ai-assistant-for-football-tactics/
DeepMind and Liverpool FC develop AI to advise on football tactics | New Scientist
https://www.newscientist.com/article/2422562-deepmind-and-liverpool-fc-develop-ai-to-advise-on-football-tactics/
Google's AI is now suggesting new football tactics
https://www.zmescience.com/future/googles-ai-is-now-suggesting-new-football-tactics/
Google DeepMind unveils AI football tactics coach honed with Liverpool
https://www.ft.com/content/e5a64dd3-7fe0-4db4-9f65-6f7517c2c573
Liverpool, a powerhouse English soccer club, has seen many miraculous comebacks. One of Liverpool's most notable upsets was their comeback against Barcelona in the Champions League semi-finals of the 2018-2019 season, also known as the ' Miracle at Anfield .' In the final tournament, which consisted of home and away matches, Liverpool lost the first away match 0-3. This meant that to advance to the semi-finals, they needed to score at least three goals in the second home match against Barcelona, home to Lionel Messi, arguably the best soccer player in the world. However, Liverpool went on the offensive from the start of the match, eventually scoring the fourth goal from a quick corner kick by local player Trent Alexander-Arnold, securing their place in the semi-finals and achieving a comeback victory against one of the world's leading clubs, Barcelona.
Reds' Miracle Comeback Against Barcelona: Liverpool 4-0 Barcelona | Champions League - YouTube
Corner kicks are one of the plays that have a high chance of leading to a goal, as seen in this dramatic goal. To generate more goals from corner kicks, you need to identify the characteristics of the players in the game and the opponent's tactics, and respond flexibly.
To help Liverpool score more goals from corner kicks, researchers at Google DeepMind collaborated with the team to develop a soccer tactical AI called 'TacticAI.' The AI uses predictive and generative AI to provide tactical insights into corner kicks, and its findings were published in the academic journal Nature Communications on March 19, 2024 (local time).
TacticAI: an AI assistant for football tactics | Nature Communications
https://www.nature.com/articles/s41467-024-45965-x
Google DeepMind and Liverpool have been collaborating since 2021 to advance sports analytical AI, and have published several papers as research results. The first paper published was on ' game plans ,' focusing on penalty kick analysis and examining the usefulness of using AI to support soccer tactics. This was followed in 2022 by the development of ' Graph Imputer ,' a prototype of a predictive system for analyzing soccer data. Following these, Google DeepMind and Liverpool developed TacticAI as a new AI.
In machine learning, input data for which prediction results are known is called 'gold standard data.' Obtaining such gold standard data is extremely difficult in soccer, where play flows constantly. However, TacticAI has succeeded in achieving superior performance by adopting a geometric deep learning approach that supports the creation of generalizable AI models.
Researchers from Google DeepMind and Liverpool experts worked together to evaluate TacticAI's performance, and found that experts preferred the tactics suggested by TacticAI 90% of the time over those actually implemented in football matches.
TacticAI is a complete AI system that combines predictive and generative models. First, the predictive model predicts what will happen, then it searches a dataset for 'what happened' in past plays, and then it suggests 'how to achieve a specific result.'
TacticAI allows you to answer three key questions:
1: What happens in a specific corner kick tactical set-up?
Which player is most likely to receive the ball, which player is most likely to attempt a shot, etc.
2: What happens when you play a setup (a pre-planned attacking strategy)?
Have similar tactics worked well in the past?
3: How should you adjust your tactics to achieve specific results?
For example, how should the defensive players be positioned to reduce the chances of a shot being taken?
TacticAI converts corner kick situations into a graph representation by treating each player as a node in the graph. 'By representing corner kick setups as a graph, we can directly model the implicit relationships between players,' Google DeepMind explained. Each node contains characteristics such as a player's position, speed, and height. The neural network uses message passing to keep each node updated.
The diagram below summarizes how TacticAI handles a corner kick. Four possible plays are generated and fed into the TacticAI model, the core of the AI, as predictive information. It then outputs information such as 'Who will touch the ball first?', 'Who is likely to take the shot?', and 'How can we increase the chances of scoring or decrease the chances of conceding a goal?', which can then contribute to the team's tactical proposals.
TacticAI leverages predictive and generative models to assist coaches by searching for similar corner kick situations and simulating different tactics. Traditionally, to develop tactics, analysts would replay many game videos repeatedly to look for similar situations or study rival teams.
Meanwhile, TacticAI can automatically calculate numerical representations of players, allowing experts to easily and efficiently search for similar situations from the past. Google DeepMind further validated this intuitive observation through extensive qualitative research with soccer experts, revealing that TacticAI's top 1 search successfully found relevance 63% of the time. This is significantly higher than the 33% success rate seen with existing approaches based on directly analyzing the similarities of players' positions.
TacticAI can be used to modify corner kick tactics developed by human coaches, reducing the chances of conceding shots during defense. It can also provide tactical recommendations, such as adjusting the positioning of all players on a specific team. Google DeepMind claims these recommendations 'allow coaches to more quickly identify key patterns as well as players who are key to the success or failure of a tactic.'
The graphs below show examples of the data TacticAI can actually output: (A) shows an example of a corner kick where a shot was attempted, (B) shows a suggestion to reduce the probability of a shot by adjusting the defender's positioning and movement speed based on the data in (A), (C) shows how the suggestion in (B) reduces the probability of the attacking player receiving the ball, and (D) shows a series of suggestions generated so far, allowing the coach to propose various tactical options.
Google DeepMind said about TacticAI, 'Our quantitative analysis shows that TacticAI accurately predicts the ball receiver and shooting situations in corner kicks, showing that player positioning is just as important as the actual play development. We also qualitatively evaluated TacticAI's suggestions in a blind case study in which evaluators evaluated which tactics were proposed by whom. Liverpool experts found that TacticAI's suggestions were indistinguishable from actual corner kick tactics and were supported over the original tactic 90% of the time. This shows that TacticAI's predictions are not only accurate, but also useful.'
Sports like soccer are also a dynamic area for AI development because they feature real-world, multi-agent interactions using multi-modal data, meaning advances in sports AI 'have the potential to be applied to a variety of fields, from computer games to robotics,' according to Google DeepMind.
Related Posts:
