In elite football, the aim of a team’s defence is to make the opposition’s play predictable. For example, a high-pressing team will press the opposition with the aim of forcing them to give the ball away in specific areas of the pitch by limiting the number of passing options available in dangerous areas.
So if the art of good defending is to make play predictable, then it should be measurable.
At the 2021 Pro Forum, Stats Perform’s Paul Power, Michael Stöckl and Thomas Seidl delivered a presentation on making offensive play predictable. The presentation applied a Graph Convolutional Neural Network (GNN) to demonstrate how using tracking data, it is possible to accurately model defensive behaviour and measure its effect on an opponent’s attacking behaviour.
Watch the presentation in full below.
In addition to the Forum presentation, a Research Paper on the subject, which has been shortlisted as a finalist in the research track at the 2021 MIT Sloan Sports Analytics Conference, is available to download here.
