Clustering of offensive playingstyles in football

Sammanfattning

Data analytics in football is becoming a crucial part of the modern football and machinelearning has become a fundamental part of everyday life for many. Connecting these two seemlike an self-evident choice, this is what this thesis will do with analysing clustering in regards toplaying styles. The clustering used is K-Means and Gaussian Mixture Model combined with PCA-projection. The project dives into how these methods work and applies it to data from Allsvenskan2023. The model showed that the there were four prominent clusters with the most successful interms of output (e.g. xG and number of shots) being high risky pressure, possession based withsafe passing. Another cluster which was quite unsuccessful but still the most original of the fourwas Low pressure, long vertical passing, the other were closer to the most successful cluster, bothbeing possession based.