Artificial Intelligence

Artificial Intelligence, Deep Learning, Machine Learning

The science of the corner… with TacticAI from Google DeepMind

Apr 3, 2024 | Ai | 0 comments

Google is strengthening its position at Liverpool FC. TacticAI, from its DeepMind subsidiary, is bringing its predictive and generative models to the football club. First results around the corner.

Setting up a corner may seem simple at first glance: a player sends the ball into the penalty area in the hope that a team-mate will score a goal. However, behind this apparent ease lies in-depth analysis and meticulous preparation.

Corner tactics are designed to create effective space within the opposition defence and offer a golden opportunity to the attackers.

Football is a constantly evolving sport, and corner strategy is no exception. Whereas in the past it was simply a matter of centring the ball into the ground, today each player has a very specific role and each movement is choreographed. Data analysis plays a key role in the development of innovative tactics. Google has just demonstrated this through its DeepMind subsidiary at Liverpool FC.

Geometric Deep Learning

Recently, Google DeepMind introduced TacticAI, an artificial intelligence-based system that can provide coaches and players with tactical information primarily about corners using predictive analysis and generative AI models. Despite the limited availability of benchmark corner data, TacticAI achieves cutting-edge results by using a geometric Deep Learning approach to help create more generalisable models.

The game situation was not chosen at random: corner kicks always involve an interruption in play, during which players from both teams occupy a fixed position for a short period of time. In other words, conditions that lend themselves to analysis…

The corner is not trivial at Liverpool FC

For example, TacticAI could analyse a defensive corner configuration and recommend repositioning defenders to reduce the chances of the attacking team receiving a shot on goal. This allows coaches to proactively reshuffle formations to tip the balance in their favour. Liverpool have had immense success with corners. Trent Alexander-Arnold’s ingenuous 2019 routine in the historic comeback against Barcelona showed what can happen if you get your corner tactics right. Google DeepMind found that AI suggestions were preferred by human expert assessors 90% of the time over tactical set-ups seen in practice.

A system like TacticAI could therefore revolutionise sports analysis, enabling coaches to spend less time manually examining videos and more time devising effective strategies.

10,000 corner situations as input data

To develop tactics and counter-tactics in corner situations, analysts have previously had to watch numerous match videos to look for similar examples and study opposing teams. This is where TacticAI comes to the rescue with its predictive and generative models.

According to DeepMind, TacticAI has been fed a dataset of almost 10,000 corner situations from the 2020 to 2023 Premier League seasons to answer three central questions using Deep Learning 

  • What will happen for a given tactical formation at a corner (predictive function)? Who is most likely to receive the ball? Will this situation lead to an attempted shot?
  • Is it possible to analyse what happened after the match (recovery function)? And can these findings be compared with tactics used in the past?
  • How can tactics be adapted to achieve a certain result (generative function)? How, for example, should defending players be repositioned to reduce the likelihood of a dangerous situation on their goal?