AI in football: How new tech can help teams take more effective corner kicks

This is where artificial intelligence (AI) — which is generally seen as having limited impact in sports other than perhaps board games such as chess or individual track and field events — could potentially play a role.

DeepMind’s TacticAI

Google’s DeepMind has introduced TacticAI: an AI system that can provide experts with tactical insights, particularly on corner kicks, through predictive and generative AI.

A corner is awarded when the ball passes over the goal line after touching a player of the defending team. The average Champions League game sees about 10 corner kicks, and predicting their outcomes is complex, given the randomness in gameplay from individual players, and the dynamics among them.

Despite the limited availability of gold-standard data on corner kicks however, TacticAI says it has achieved encouraging results by using “a geometric deep learning approach” that helps create more generalisable models.

Predictive and generative

DeepMind developed and evaluated TacticAI in a multi-year research collaboration with experts from Liverpool Football Club.

It says that human expert raters preferred TacticAI’s suggestions over tactical setups seen in practice an astounding 90% of the time. Petar Velickovic, who led the team that worked on TacticAI, has pitched for the tool to be used by team managers as an assistant for tactics.

Identifying key patterns of tactics implemented by rival teams, and developing effective responses, lies at the heart of modern football — and TacticAI, according to Velickovic, has incorporated both a predictive and generative component to allow coaches to effectively sample and explore alternative player setups for each corner kick, and to select those with the highest predicted likelihood of success.

Sports such as football are a dynamic domain for developing AI, as they feature real-world, multi-agent interactions, with multimodal data. While TacticAI essentially demonstrates the potential of assistive AI techniques to revolutionise sports for players, coaches, and fans, advancing AI for sports could potentially translate into actionable models in many areas off the field as well — from computer games and robotics to traffic coordination.

Using AI in corner kicks

Velickovic’s team at DeepMind first published a paper called Game Plan, which looked at why AI should be used in assisting football tactics, highlighting examples such as analysing penalty kicks. Then, in 2022, it developed ‘Graph Imputer’, which showed how AI can be used with a prototype of a predictive system for downstream tasks in football analytics.

(A) How corner kick situations are converted to a graph representation. Each player is treated as a node in a graph, and their relationships as edges. A neural network operates over this graph, updating each node’s representation using message passing.
(B) How TacticAI processes a given corner kick. All four possible combinations of reflections are applied to the corner, and fed to the core TacticAI model. They interact to compute the final player representations, which can be used to predict outcomes.
(Source: DeepMind)

The latest iteration, TacticAI, is being pitched as a “full AI system”. According to the DeepMind team, it is built to address three core questions:

• For a given corner kick tactical setup, what will happen? E.g., which player is most likely to receive the ball, and will there be a shot attempt?
• Once a setup has been played, can we understand what happened? E.g., have similar tactics worked well in the past?
• How can we adjust the tactics to make a particular outcome happen? E.g., how should the defending players be repositioned to decrease the probability of shot attempts?

The tool represents corner kick situations as graphs, with players depicted as nodes and their relationships as edges. A neural network then operates over this graph, updating each node’s representation. This enables modelling the interactions between players, which may be more important than the absolute distances between them.

Why specifically corners

The focus on corner kicks is somewhat strategic as well. The models began with trying to predict aspects of open play in a football match, but there are pitfalls: even if a model were to give a suggestion about the current state of open play, a coach cannot always meaningfully act on it in that moment. Shouted instructions during open play may confuse players, or let the other side in on team tactics.

Corner kicks are apt for strategising by leveraging AI tools, primarily because they are moments when the game is effectively frozen — and always starts from the same kind of position at the corner of the pitch while giving players an immediate opportunity to score. Strategies for corners are usually also decided long before the players actually go on to the pitch, so that there is no confusion on match day.