# Predicting the course of the game

Is it possible to predict the trend of a football game?

Can we make specific changes on the course of the game, which may invert this trend? In joint collaboration between TEF and the Department of Fundamental Physics at Universitat de Barcelona tries to tackle this problem.

Here we employ a different approach. Instead of putting the players at the centre, we consider the flow of the game in field positions (nodes in a network).

Players have an important role providing a better strategic flow of the game, but are not centralised in the nodes. Positions in the field are the nodes. Therefore, we consider a flow network for both teams as well as the change of flow from one team to the other.

That is, **Figure A**, shows two field layers networks (one for each team) and interlayer networks corresponding to the change of flow from one team to the other (intersections). The model in this case is a network of networks that can mathematically be represented by the tensor in **Figure C**.

Each game defines a network of networks. Where the nodes are field positions and the edges the probability that a pass goes from a to b in the network. These probabilities change with the progress of the game.

The goal of a given team is to maximise the flow towards the opponent team goal and at the same time minimise the flow into its own goal.

Once analysed a full Spanish Football season, we realise a strong correlation between the difference in goal scores and the projected future flow after mid game term **Figure D**. This allow us to statistically predict if a given team will loose the game at mid term.

Our model may allow us to detect which flows should we change at a mid term game in the field with specific players in order to change the projected trend. And in this case in **figure E**, after some strategic changes we may invert the trend shown in **figure E**.