Recognising Group-Tactical Manoeuvres from Player Positions

Abstract

This thesis presents a novel methodology for recognising group-tactical manoeuvres in soccer by analysing player positions using advanced computational techniques. Leveraging tracking data from Bundesliga matches, the study focuses on detecting key tactical behaviours such as dropping midfielders, inverted players, and over- and underlap manoeuvres. Team formations are represented as shape graphs in a goal-aligned coordinate system with player positions derived from the shape graphs. The thesis introduces rule-based algorithms to classify player movements and detect specific tactical patterns, offering insights into how these manoeuvres influence game dynamics. The effectiveness of the proposed methods is showcased through a discussion of the detected manoeuvre instances, and overlap runs are compared against established approaches.