Liverpool need to get back on track after Atalanta humiliation – and AI can help

Liverpool need to get back on track after Atalanta humiliation – and AI can help
Liverpool suggested Google DeepMind focus on corner kick analysis - Getty Images/Daniel Chesterton

Liverpool bid farewell to a legendary manager this summer, but they may have already found the next super-coach; the Premier League’s artificial intelligence revolution is here.

A pioneering five-year collaboration between Liverpool and researchers from Google DeepMind began in 2019. The result is TacticAl – an AI system predicting and generating practical coaching advice, which also accelerates access to the most relevant information for data analysts scouting opponents.

The academics responsible for the tool say their work could have a profound impact on football and other sports as clubs and managers pursue the ‘marginal gain’ that can define a championship season, with Liverpool bidding to keep their title challenge on track when they take on Crystal Palace this Sunday.

TacticAI was unveiled in March, with the immediate focus on corner kicks. Over the course of two seasons, 7,176 Premier League corners were studied in forensic detail, establishing the variations of attacking and defensive strategies, and which were most likely to succeed.

“It was Liverpool who suggested we focus on corner kick analysis – it is rigid, happens frequently and gives you goalscoring opportunities,” explains Petar Velickovic, a staff research scientist at Google DeepMind and affiliated Cambridge University lecturer.

“You work out strategies well in advance so it is easier to make concrete actions. TacticAl is a collection of three different systems. There is the predictive – where the tool can tell you what is most likely to happen, for example who will receive the ball and have a shot.

“Then there is retrieval, so video analysts re-watching corners from a particular team can prepare for an upcoming match based on what the opponent has done before. It can be a painstaking process watching a lot of videos, so a system streamlining the most relevant will speed up the time needed to do so.

“Most exciting is the generating model, which can adjust players’ position and velocities to make the probability of a certain outcome go up and down. They can help in defending a corner, or when attacking.”

Once the tool was fully functional, in 90 per cent of cases Jürgen Klopp’s assistants could not differentiate which suggestions or formations were a result of human or AI input.

Liverpool are coy as to whether they put TacticAI into use during this season. Their affiliation with Google DeepMind followed the famous Trent Alexander-Arnold corner against Barcelona in the 2019 Champions League semi-final (watch below), but they have already won one trophy in 2024 thanks to a set-piece. Suffice to say, Virgil van Dijk’s 119th minute header from a corner against Chelsea in the Carabao Cup final will have been heartily celebrated by the research team, wherever the sources of input for Klopp’s best point of attack came from.

In other industries, such AI influence has not been without controversy, with many speculating if supercomputers will be exploited by cash-conscious executives to reduce staff numbers.

The scientists are adamant their tool is no threat to up-and-coming coaches.

“Liverpool’s set-piece coaches were excited. It makes their job easier, not harder,” Velickovic is keen to stress.

“We did not build an AI corner coach. Our aim was always to build an assistant to the coach. It is a system to help the coach to do their job better. It will still be up to the coach to make individual decisions about how to use the information and prepare the players.

“TacticAI probably will not discover novel tactics, but it will identify deviations in existing situations. It can show if a player is asleep at the wheel when defending a corner, for example.”

Velickovic and co-leader on the TacticAI project, Zhe Wang, says it is no surprise that Liverpool have been so proactive in the AI field. Liverpool’s former and current head of data, Ian Graham and Will Spearman, are listed as co-authors in the corner kick study.

“When Ian Graham and Will Spearman launched the football data department at Liverpool they became heroes of the data driven community in sport,” says Wang, a research engineer at Google DeepMind.

“Liverpool’s data team is very technically orientated – they have people with PhDs in physics and astrophysics. We speak the same language so collaboration was easy. Football is not only art. It is also embracing science and technology.”

However, there was a stroke of fortune – or as Wang puts it ‘serendipity’ – about how the partnership began.

“Our former team leader, Karl Tuyls, was a professor at the University of Liverpool and was a neighbour of someone who worked at the club, and our chief executive Demis Hassabis is also a big Liverpool fan, so that accelerated the collaboration.”

‘There are moments where you can give yourself an extra one per cent’

Velickovic says individual sporting genius and spontaneity means such tools will always be an enhancement for the most curious footballing thinkers rather than a substitute.

“I fell in love with football because it is art, and I still think it is art,” he says.

“Truly magical things happen in situations that cannot be predicted or figured out. The Trent Alexander-Arnold free-kick versus Barcelona is a perfect example of that. But when optimising a club’s chances of winning over a season there are certain moments where you can give yourself an extra one per cent percentage point of stopping or scoring a goal. This accumulates over the course of a season.”

So where next? The exclusivity arrangement with Liverpool expires this year.

“We anticipate there will be interest across football and in other sports,” says Velickovic, who believes AI will enable clubs to profile players suited to their systems, and can offer insights into the most successful penalty and free-kick techniques.

“Building an agent that models open play is tricky. The horizons are broad,” he says.

“But I will not be surprised if TacticAI is used by a lot of teams, not just with corners. Now the research paper is out in the open anybody interested in this technology can take it. We anticipate there will be interest across football and in other sports.

“Generally, the more set-pieces there are in a sport, the more meaningful the actions can be.

“Rugby is one good candidate, so is American football and basketball. In baseball, every action is a set-piece. The more frequent the set-pieces, the more immediately actionable the system. A freeze in game play makes it more practical.”


AI scouting is changing football at elite and amateur level

Read more

Broaden your horizons with award-winning British journalism. Try The Telegraph free for 3 months with unlimited access to our award-winning website, exclusive app, money-saving offers and more.