Clustering of football players based on performance data and aggregated clustering validity indexes

Abstract We analyse football (soccer) player performance data with mixed type variables from the 2014-15 season of eight European major leagues. cluster these based on a tailor-made dissimilarity measure. In order to decide between many available clustering methods and choose an appropriate number clusters, we use approach by Akhanli Hennig (2020. “Comparing Clusterings Numbers Clusters Aggregation Calibrated Clustering Validity Indexes.” Statistics Computing 30 (5): 1523–44). This is several validation criteria that refer different desirable characteristics clustering. These are chosen aim clustering, this allows define suitable index as weighted average calibrated individual indexes measuring features. derive two clusterings. The first one partition set into groups essentially players, which can be used for analysis team’s composition. second divides small clusters (with 10 players average), finding very similar profile given player. It discussed in depth what Weighting informed survey experts.