A historic partnership between Liverpool and DeepMind to harness artificial intelligence skills in the service of football tactics (Getty)

“Do you know who is the most exciting young coach in modern football? He is not at Bayer Leverkusen in Germany or Stade Reims in France, or even Bologna in Italy. He is with Google.” These were the opening words of a report published by the American magazine “Wired” under the title “How Can Google DeepMind's new AI model help football clubs achieve the perfect corner kick?

A few days ago, the English Liverpool team launched a historic partnership with DeepMind, a subsidiary of Google, with the aim of harnessing artificial intelligence skills in the service of football tactics. According to what was published by the “New Scientist” website, the model called “Tactical” was trained on a series of... Huge data, relating to more than 7,176 corner kicks in the Premier League, from 2020 to 2021, which enabled it to analyze patterns and determine the best offensive and defensive options, especially when executing corner kicks.

Tests have shown that he has a high ability to accurately predict the results of corners. He presented tactical suggestions that proved effective in increasing goal scoring chances. Even more dangerous is that there were many football experts who were unable to distinguish between those tactics he suggested and those developed by human coaches. Preferring tactics suggested by AI in 90% of cases.

Therefore, Jurgen Klopp, Liverpool coach, expressed his happiness with this cooperation, stressing that “artificial intelligence can help understand the game better, especially with the advanced capabilities of modern artificial intelligence models, during which the way football is played and analyzed radically changes, which “It will contribute to raising the level of the game and providing an exciting experience for football fans around the world.”

We're announcing TacticAI: an AI assistant capable of offering insights to football experts on corner kicks. ⚽

Developed with @LFC, it can help teams sample alternative player setups to evaluate possible outcomes, and achieves state-of-the-art results. 🧵… pic.twitter.com/8oVYtamGGe

— Google DeepMind (@GoogleDeepMind) March 19, 2024

Tactical

Before this attempt, there were many attempts between Liverpool FC and Google to create an artificial intelligence system that would help coaches make better tactical decisions, but they were somewhat desperate attempts and did not bear fruit quickly, especially with the pressure of matches and intense competition, so experts resorted to codifying conditions. The “Tactical” system in 2021, and working to transform it into a special model for analyzing and devising mechanisms that help players in specific situations, so that the choice falls on corner kicks and set pieces. Although predicting the results of corner kicks is also complex, due to the limited data available, it remains a simpler process in general. From studying open-ended play situations, which are more dynamic in nature.

This AI model is therefore designed to address fundamental questions: 

Who is the most likely player to receive the ball? Will there be any shot attempts? Can we understand what happened? Have similar tactics worked well in the past? How can we modify tactics to achieve a different result? How can we change the players’ position to increase the probability of shooting during attack, and vice versa during defense? This is done by providing constructive suggestions to coaches, based primarily on analyzing previously available shots (historical data), and finding similar tactics and solutions through them, somewhat similar to the expected goals model based on studying thousands of different shots from previous matches.

This is exactly what “Tactical” does in an automated and simple way, as it searches and analyzes the database it has, and then presents the answer in the form of a graphical representation and tactical suggestions. After the experts conducted some extensive qualitative studies, they found that the model’s solutions are often related to the database. It was already available to him, in 63% of cases specifically, meaning that these recommendations were not an innovative solution at all, but rather relied primarily on stealing ideas from the database already available to him.

The study also shows that this model is the first artificial intelligence assistant that “can provide useful and practical suggestions,” and cannot be distinguished in any way from human tactics, as it uses a complex database related to position, speed, weight, and height, which is extremely difficult, because artificial intelligence models usually You are training on billions of data points, which is an element that is not available in corners, because they are inherently rare.

Therefore, it was necessary to find a “magic solution”, which lies in “deep engineering learning”, in a way that enables the machine to work similar to the human brain. According to what Peter Velicovich, the mathematician, said, this innovation works to simplify the process of tactical planning that requires labor. Dense, so it involved club analysts re-watching numerous match videos to study past routines and the opposing teams.

When the IA vient apporter his aid in the tactics of Jurgen Klopp. 🧠⚽️

Google DeepMind has developed a tactical analysis system in collaboration with Liverpool!

L'objectif? Suggérer des tactics à l'traineur sur l'exercice des corners.

« 𝗟𝗲𝘀… pic.twitter.com/isixYLlpYe

— Actu Foot (@ActuFoot_) March 20, 2024

Rich history

This is not the first cooperation between sports clubs and these technology companies, as the use of data analysis to improve results in sports has become more developed in recent years, specifically since writer Michael Lewis drew the world’s attention to this technology, in his book published in 2003 under the title “Moneyball: The Art.” of Winning an Unfair Game.

In his book, Lewis recounted how the Oakland Baseball team used new technology that enabled it to face better-financed opponents, and how the National Football League and Amazon Web Services developed a tool for players known as “Digital Athlete” that can predict... Injuries before they occur, to prevent them.

Then he talked about how Matser Golf worked with IBM Technology Sciences to create an application that can predict a player’s ability and comprehension, in addition to a feature that adds spoken commentary to improve performance, on every shot, and for every player individually. And completely automatically, and as the matter developed, there became special mobile phone applications such as “Home Court,” “ESBN,” “Smart Coach,” and other applications that give players a good and easy way to improve.

Accordingly, the sports artificial intelligence market has witnessed significant growth in recent years, and its value is expected to reach 22 billion US dollars by 2030, and the global sports artificial intelligence market is also expected to reach 19.9 billion dollars by the same year. A clear indication that the use of artificial intelligence in sports has already expanded, and will continue to grow as sports teams, leagues and technology grow.

unknown future

Google Deep Managed researcher Petar Veljkovic explained that the smart model does not aim to replace human technical managers until the present time, but rather it is a step on the part of the Google research sector, to enhance the capabilities of humans and raise the efficiency of their performance, instead of replacing them, but in the same statement, He added, "This does not prevent the system from being expanded to include other parts of football, and even other sports," pointing out that "if we can automate (automate) a game like football, with all the complexities it carries, then we can design several aspects better than human psychology,” he said, noting that his team’s research has broader applications beyond sports.

It is clear that the world is not moving towards getting rid of human efforts at the present time, and that artificial intelligence is unable to hijack the coach’s job, especially since the coach’s personality plays an important role in the educational, social and emotional aspects. All of these human qualities facilitate the process of delivering messages to the players, and enable... The coach is able to know their true feelings regarding training or matches, given that group games are poetic in nature, and depend on the extent of harmony between individuals, and the extent of their connection, division, and harmony as a single bloc, away from dull tactical sentences.

But the terrible development of artificial intelligence technologies over the past years may actually threaten the future of coaches in sports stadiums, especially since artificial intelligence and robots have succeeded in all technical tests. Can they also understand human feelings in the future, and put an end to the human coach forever?

AI will NEVER be superior to human coaching, as it relates to habit formation and affirmation.

All bias aside, Future will serve as the model by which all other successful one-to-one coaching products are built.

AI is meant to augment great Coaches, it will never replace them. https://t.co/BdzimdCVNk

— Joshua Bonhotal (@jbonhotal) June 15, 2023

Source: Al Jazeera