A new study uses AI to simulate ‘what if’ scenarios in football, revealing how coordinated team behaviour can improve the chances of scoring goals.

Football is a team sport built around coordinated play to score goals – but improving goal-scoring potential lies not just in how how players pass the ball, but also in how they move and coordinate without it.

It is this aspect of the game that a research team led by the Chinese Academy of Sciences in Beijing wanted to analyse. In particular, they were keen to better understand a common defensive tactic called a low block, which is especially challenging to break down. It involves a team defending in a tight, narrow shape very deep in their own half of the pitch, just outside of their penalty area. When the opposition meets this low block, the weight of numbers and lack of space makes it very hard for them to find opportunities to score.

In their study, the researchers developed an AI model that learns from real-world match data to better understand and optimise attacking play against such compact defensive structures. Using data from existing large-scale events and professional matches, the framework treats each attacking player as an individual decision-maker, or AI ‘agent’, and analyses how they all interact together as a coordinated unit.

Yi Pan, corresponding author at the Chinese Academy of Sciences, said: “Our goal was to move beyond analysing isolated actions and instead understand football as a truly collective decision-making process.

“In particular, we wanted to capture how off-ball movements, which are often invisible in traditional statistics, contribute to creating space and breaking defensive lines.”

To reflect real tactical thinking, as well as match data, the AI model also includes knowledge from professional football coaches and analysts. It then predicts and evaluates both on-ball actions, such as passing and carrying, and off-ball movements, such as runs that stretch or disrupt the defensive structure. 

“By learning from historical match data, it can assess the effectiveness of different strategies and even suggest alternative actions that could have led to better outcomes,” said Pan.

One of the main findings from the research is that unlike typical human play, which often avoids risk, the AI model suggests more creative and proactive attacking behaviours while remaining tactically realistic and consistent with real-world matches.

Pan said: “Human players often favour safer, lower-risk decisions, but the model identifies opportunities where more dynamic movements and coordinated positioning could create space and increase scoring potential.”

According to the researchers, this balance between realism and innovation makes the approach particularly valuable because it provides a new way for coaches and analysts to evaluate not just what happened in a match, but also what could have happened.

“By enabling counterfactual analysis of decisions and movements, it supports more informed tactical planning and offers deeper insights into how coordinated team behaviour emerge on the pitch,” said Pan.

The study, ‘Offline multi-agent reinforcement learning for evaluating and optimising football attacking strategies against low-block defences’, has been published in the journal Intelligent Sports and Health.