Back of the net? Google's DeepMind is coming for football tactics
(2024/03/20)
- Reference: 1710932533
- News link: https://www.theregister.co.uk/2024/03/20/googles_deepmind_football/
- Source link:
A team at Google's DeepMind claim to have demonstrated the efficacy of an AI model in predicting outcomes in soccer football game set pieces, as well as generating on-field tactics.
Football - referred to as soccer stateside - is the most popular sport on the planet by a long way. It has been estimated that around 1.5 billion people tuned in to watch the 2022 World Cup, which took place in Qatar.
Using a graph machine learning model and data from 7,176 corner kicks, the team at Deep Mind were able to build a tool called TacticAI which accurately predicted the first receiver of the ball and the direct result of the kick. It could also predict the possible outcomes with different set-up positions and create tactical variations that could help improve the game outcomes, according to a [1]paper published in Nature Communications.
[2]
Zhe Wang and his team of Deepmind researchers worked with Liverpool FC, six-time winner of the UEFA Champions League, Europe's most prestigious competition, to collect the data and evaluate the model with five football experts.
[3]
[4]
The model was built to analyze corner kicks, a set-play in which a team has the opportunity to launch the spherical ball in front of the goal which they are attacking. The developers said it could also be adapted for other set plays including free kicks and throw-ins.
"Identifying key patterns of tactics implemented by rival teams, and developing effective responses, lies at the heart of modern football. However, doing so algorithmically remains an open research challenge," the paper says.
[5]
TacticAI involves both a predictive and a generative component, DeepMind adds, "allowing the coaches to effectively sample and explore alternative player setups for each corner kick routine and to select those with the highest predicted likelihood of success."
A survey of football experts showed the model's suggestions are favored over existing tactics 90 percent of the time. "TacticAI achieves these results despite the limited availability of gold standard data," the team says in the research paper.
[6]Whizkids jimmy OpenAI, Google's closed models
[7]Europe benched in high tech 'Champions League' says ASML
[8]Hundreds of websites cloned to run ads for Chinese football gambling outfits
[9]Hacking a Foosball table scored an own goal for naughty engineers
"We have demonstrated an AI assistant for football tactics and provided statistical evidence of its efficacy through a comprehensive case study with expert human raters from Liverpool FC," the paper concludes. "The system's latent player representations are a powerful means to retrieve similar set-piece tactics, allowing coaches to analyze relevant tactics and counter-tactics that have been successful in the past."
As the USA looks forward to hosting the next FIFA World Cup football tournament in 2026, it seems reasonable to speculate it will be the first when teams widely employ AI as a tactical assistant.
Algorithms or not, it'll take more than AI to help England win the tournament, especially if penalties are involved. ®
Get our [10]Tech Resources
[1] https://nlcontent.springernature.com/d-redirect/TIDP2784183X9A94A61AC3764D7B9CCF883BE22E58A3YI4/
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZfsWKkQwggdJBRC2hUDMwgAAAEY&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZfsWKkQwggdJBRC2hUDMwgAAAEY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZfsWKkQwggdJBRC2hUDMwgAAAEY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZfsWKkQwggdJBRC2hUDMwgAAAEY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2024/03/13/researchers_pry_open_closed_models/
[7] https://www.theregister.com/2024/01/16/eu_champions_league_asml/
[8] https://www.theregister.com/2023/11/16/qurium_cloned_websites_chinese_gambling/
[9] https://www.theregister.com/2023/07/03/who_me/
[10] https://whitepapers.theregister.com/
Football - referred to as soccer stateside - is the most popular sport on the planet by a long way. It has been estimated that around 1.5 billion people tuned in to watch the 2022 World Cup, which took place in Qatar.
Using a graph machine learning model and data from 7,176 corner kicks, the team at Deep Mind were able to build a tool called TacticAI which accurately predicted the first receiver of the ball and the direct result of the kick. It could also predict the possible outcomes with different set-up positions and create tactical variations that could help improve the game outcomes, according to a [1]paper published in Nature Communications.
[2]
Zhe Wang and his team of Deepmind researchers worked with Liverpool FC, six-time winner of the UEFA Champions League, Europe's most prestigious competition, to collect the data and evaluate the model with five football experts.
[3]
[4]
The model was built to analyze corner kicks, a set-play in which a team has the opportunity to launch the spherical ball in front of the goal which they are attacking. The developers said it could also be adapted for other set plays including free kicks and throw-ins.
"Identifying key patterns of tactics implemented by rival teams, and developing effective responses, lies at the heart of modern football. However, doing so algorithmically remains an open research challenge," the paper says.
[5]
TacticAI involves both a predictive and a generative component, DeepMind adds, "allowing the coaches to effectively sample and explore alternative player setups for each corner kick routine and to select those with the highest predicted likelihood of success."
A survey of football experts showed the model's suggestions are favored over existing tactics 90 percent of the time. "TacticAI achieves these results despite the limited availability of gold standard data," the team says in the research paper.
[6]Whizkids jimmy OpenAI, Google's closed models
[7]Europe benched in high tech 'Champions League' says ASML
[8]Hundreds of websites cloned to run ads for Chinese football gambling outfits
[9]Hacking a Foosball table scored an own goal for naughty engineers
"We have demonstrated an AI assistant for football tactics and provided statistical evidence of its efficacy through a comprehensive case study with expert human raters from Liverpool FC," the paper concludes. "The system's latent player representations are a powerful means to retrieve similar set-piece tactics, allowing coaches to analyze relevant tactics and counter-tactics that have been successful in the past."
As the USA looks forward to hosting the next FIFA World Cup football tournament in 2026, it seems reasonable to speculate it will be the first when teams widely employ AI as a tactical assistant.
Algorithms or not, it'll take more than AI to help England win the tournament, especially if penalties are involved. ®
Get our [10]Tech Resources
[1] https://nlcontent.springernature.com/d-redirect/TIDP2784183X9A94A61AC3764D7B9CCF883BE22E58A3YI4/
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZfsWKkQwggdJBRC2hUDMwgAAAEY&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZfsWKkQwggdJBRC2hUDMwgAAAEY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZfsWKkQwggdJBRC2hUDMwgAAAEY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZfsWKkQwggdJBRC2hUDMwgAAAEY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2024/03/13/researchers_pry_open_closed_models/
[7] https://www.theregister.com/2024/01/16/eu_champions_league_asml/
[8] https://www.theregister.com/2023/11/16/qurium_cloned_websites_chinese_gambling/
[9] https://www.theregister.com/2023/07/03/who_me/
[10] https://whitepapers.theregister.com/
Re: Based on Premier League training data
Nuff Said
Then jump up and run away like nothing had happened once the ref has awarded the free kick.
I ain't Spartacus
Is this going to be like the laughably crap Amazon AI introduced in Formula 1 a few years ago?
And is their metric really "a survey of experts"? Let me guess: rather than actually testing whether it works, what they've done is slapped an LLM on a bunch of football wirting from pundits - got it to spit out results based on what they write is the best idea - then asked them if what it outputs is the best idea?
On the other hand, they may have a point with corner kicks. There's an obsession in the game with fancy short corners and clever tactics that rarely work. Rather than banging it in the box at head height - everybody jumps in the air while pulling each others' shirts and random chance means you score a decent number of times - even if it is accidentally, off the strikers arse. Any goal's a goal.
Based on Premier League training data
New strategy is for attacker to dive in penalty area and roll around clutching leg