08 Jun From Movement to Advantage: How AI Can Help Us Understand Player Behaviour in Football
There is a part of football that still escapes most data.
Not the pass.
Not the shot.
Not the sprint.
Not the tackle.
The movement before all that.
The adjustment before receiving.
The run that opens a passing lane.
The pause that holds an opponent.
The step away from the ball that creates space.
The body position that prepares the next action.
These moments often explain the game, but they rarely appear in traditional event data.
Coaches see them. Analysts see them. Players feel them.
The challenge is turning them into evidence.
At LALIGA’s Football Intelligence & Performance Department, we are working on the definition and implementation of an AI-based framework designed to analyse individual player behaviour through tracking data.
The goal is not to replace tactical analysis.
The goal is to make movement more observable, explainable and coachable.
The real question: does movement create advantage?
For years, tracking data has helped football measure physical output: distance, speed, accelerations, high-intensity actions.
That remains useful.
But movement is not valuable only because it is fast or intense.
Movement is valuable when it changes the game: when it helps a player receive, creates time, opens space, protects the team, reduces an opponent’s option or prepares the next action.
So the key question is not:
How much did the player move?
It is:
Did that movement help create, maintain or prevent advantage?
That question is the centre of the framework.
Player-ball harmony
The first step is to understand the relationship between the player and the ball.
We call this player-ball harmony.
It is not simply proximity to the ball.
Sometimes the best movement is towards the ball. Sometimes it is away from it. Sometimes the most valuable movement is not to receive, but to make another player available.
Player-ball harmony looks at whether the player’s movement is connected with what the ball is generating.
Is he adjusting before the ball arrives?
Is he late to the new ball location?
Is he moving towards useful space?
Is his trajectory aligned with the next possible action?
Is he reacting to where the ball is, or anticipating where advantage may appear?
Player-ball harmony is our first measurable proxy for whether individual movement may be contributing to advantage.
This is where raw tracking can start becoming behavioural information.
What AI adds
AI is not being used here as a black box that “understands football”.
Its role is more precise.
The framework uses tracking sequences, temporal windows, role-conditioned patterns and contextual labelling to identify repeated spatio-temporal behaviours.
In simpler terms, it searches for movement patterns that happen again and again in similar football contexts.
But the output must remain honest.
In this first phase, the model should not say:
“The player moved to attract a defender.”
That would require relational validation.
A more rigorous output would be:
“This movement pattern is compatible with a behaviour that may create space or fix an opponent, pending teammate-opponent context.”
That difference matters.
It separates useful AI from football-shaped speculation.
Why ball height matters
The ball is tracked 50 times per second, including its height through the Z coordinate.
This is not a technical decoration.
Ball height changes the behavioural demand.
A ground pass, a bouncing ball, a clearance, a cross, a second ball or an aerial switch require different timing, body orientation and movement preparation.
For a coach, this is obvious on video.
For an AI framework, it must be measured.
The Z coordinate helps avoid treating all ball movements as if they were flat. Football is not played only on the grass.
Making invisible behaviour visible
The greatest opportunity is off-ball behaviour.
Traditional data mainly rewards the player who touches the ball. But many decisive actions happen without contact.
A midfielder opens his body before receiving.
A winger holds width to stretch the opponent.
A forward delays his run to stay connected with the passer.
A full-back closes a passing lane without making a tackle.
A centre-back drops two metres before danger becomes obvious.
These behaviours are often decisive.
The framework aims to help coaches and analysts find them faster, review them better and train them with more precision.
Not by replacing the expert eye.
By directing it towards better evidence.
The player’s behavioural fingerprint
Every player has a behavioural signature.
How he reacts to ball movement.
How quickly he adjusts.
When he accelerates.
Where he tends to offer support.
How he behaves in different phases.
How his role changes inside the tactical system.
The framework aims to build this individual fingerprint.
But always with context.
A full-back, winger, centre-back and attacking midfielder cannot be evaluated with the same behavioural expectations.
Role matters.
Phase matters.
System matters.
Game context matters.
Without context, tracking data can easily become misleading.
The next layer: skeleton data
The framework is also prepared to incorporate skeleton data, based on 29 body points per player.
This could add a deeper reading of body orientation, reaction time, turning behaviour, movement preparation and readiness to act.
Because football decisions often appear in the body before they appear in the run.
Tracking tells us how and where the player’s mass moved.
Skeleton data can help us understand how his body structure was oriented to act.
That distinction is important.
One layer explains displacement.
The other may explain preparation.
An honest limitation
This first phase analyses each player as an individual unit.
In simple terms, it is like watching the match from 22 individual perspectives.
That is powerful.
But it is not the whole game.
Football is not a collection of isolated behaviours. A player’s movement only has full tactical meaning inside a network: teammates, opponents, ball, space, phase and team intention.
A player may move away from the ball to create space.
He may not press because he is protecting a passing lane.
He may not receive because his movement helped free a teammate.
That is why this framework should be understood as a first layer, not a final answer.
The current focus is the individual behavioural fingerprint.
The next step is the relational layer: how the player interacts with teammates and opponents.
The final step is the collective layer: how individual movement contributes to creating, maintaining or preventing advantage.
Current stage
The framework is currently in a theoretical and implementation phase.
Initial visual tests and applied explorations have already been carried out in real tracking environments, but this should not be presented as a fully deployed production tool.
Initial visual explorations already suggest that the framework can highlight behaviours that are difficult to capture through event data alone.
The priority now is to define the right football questions.
Because the future of AI in football should not be about producing more metrics.
It should be about producing better evidence for better decisions.
Why this matters for coaching staffs
For coaches and analysts, this framework could help answer practical questions:
Is the player supporting the ball at the right time?
Is he moving towards useful space?
Is he synchronised with the passer?
Does he react quickly to changes in ball trajectory?
Does his movement fit the role required by the system?
Is his off-ball behaviour helping the team create or prevent advantage?
That is the practical value.
AI should not make football analysis more abstract.
It should make behaviour more visible, more explainable and more trainable.
Final thought
The future of tracking analysis is not to count movement better.
It is to understand when movement changes the game.
Not movement as output.
Movement as advantage.