Performance of performance indicators in football

Abstract

Performance indicators (PIs) play a key role in today’s sports analytics. In this paper we examine the performance of performance indicators with respect to (match)day-to-day stability and predictive validity for final ranking. A sample of 279 football matches of the German Bundesliga in the Season 2016/2017 was analyzed. Inter-day correlations were computed per PI using Pearson correlations. The distribution of 450 inter-day correlations per PI was analyzed. Spearman’s rank correlations between season mean values of the PIs and final ranking were calculated as validity check. Also, the Pearson inter-correlations between the PIs were analyzed for exploring their mutual relationships. The results show rather low correlations with the highest median correlation of 0.43 for number of completed passes. The highest predictive values (0.794–0.771) were found for traditional PIs like number of passes as well as for more recently introduced Packing PIs. Inter-correlations reveal that the more recent PIs are very highly correlated with traditional ones. The findings support the view of matches as not repeatable, dynamic interaction processes with emergent behavior rather than being predictable by summative PIs.