DeepMind teamed up with Liverpool FC to create TacticAI, an AI system set to remodel how football teams approach tactics, starting with corner kicks.
Remember the unforgettable moment within the 2019 UEFA Champions League semi-finals, with the fast corner that led to Origi’s legendary goal?
You will for those who’re a Liverpool or Barcelona fan – and that’s the form of play TacticAI goals to create more of, mixing the art of human intuition with AI optimization.
TacticAI, detailed in a study published in Nature Communications, uses geometric deep learning to provide models able to predicting corner outcomes and analyzing tactics.
A: For graph representation of corner kicks, each player is a node in a graph with detailed features. Graph neural networks update these node features through message passing based on interactions with neighboring nodes. B: TacticAI analyzes corner kicks by applying all possible pitch reflections to take care of robustness against orientation changes. This creates 4 versions of the identical scenario, processed by TacticAI to derive player interactions. The final result predicts the receiver and shot attempts and suggests position or velocity adjustments for players. Source: Nature Communications (open access).
TacticAI actually began with an earlier project named Game Plan back in 2020, which sought to harness AI for football evaluation and strategy, evolving into sophisticated predictive models able to anticipating off-camera player movements.
For this recent study, researchers collected data on 7,176 corner kicks from the Premier League’s 2020 to 2021 seasons. Each corner kick was transformed right into a graph representation, where players were nodes featuring data on their positions, movements, and physical attributes on the time the corner was executed.
Researchers then applied geometric deep learning through graph neural networks (GNNs) to research spatial interactions and methods during corner kicks.
Compared to traditional approaches, TacticAI’s predictions were the popular alternative of expert evaluators 90% of the time, and it could also predict each receivers and shot-takers.
At its core, TacticAI serves three primary functions:
- Predicting the likely outcomes of a given corner kick setup, resembling identifying potential receivers and estimating shot attempt possibilities.
- Analyzing the effectiveness of past tactics in similar setups, providing insights into what worked or didn’t.
- Offering tactical adjustments to influence future outcomes, for instance, suggesting defender repositions to lower the opposition’s shot attempt possibilities.
From there, TacticAI was measured against three benchmark tasks:
- Shot prediction: TacticAI reached an F1 rating of 0.71 in predicting shot attempts from corner kicks. An F1 rating closer to 1 indicates higher accuracy – so 0.71 is great.
- Receiver prediction: This involved determining the player likely first to receive the ball post-corner. TacticAI demonstrated a top-3 test accuracy of 0.782.
- Tactic suggestion: Proposing adjustments in player positions and movements to optimize tactical setups. This is where qualitative expert evaluators generally preferred TacticAI.
TacticAI enables coaches to refine corner kick strategies, enhancing outcomes for each offense and defense. For instance, in a scenario where a shot was initially likely (A & B), TacticAI proposes defensive repositioning to lower shot possibilities (D). This decreases the likelihood of attackers receiving the ball, aside from one positioned removed from the goal (C), illustrating TacticAI’s ability to craft various tactical scenarios for in-depth evaluation and selection. Source: Nature Communications (open access).
Are corners one of the unpredictable sports set pieces? Perhaps not anymore.
TacticAI might end in more injury-time winners – and straight away, Liverpool FC is ready to make the most.
Liverpool fans – you lost Klopp, but a minimum of you’ve TacticAI.
