There has been a rise in available data within the sports industry. Soccer is the most popular sport in the world there is high interest in exploiting this data for a wide variety of applications. Machine learning tools seem to be ideal for doing this in an array of subfields within the sport. With this in mind, the current study investigated the KPIs distinguishing elite defenders (CL) from their sub-elite counterparts(N-CL). The study analyzed 1661 defenders with 63476 individual match performances from the top 5 European soccer leagues on a match performance level. Three machine learning algorithms, viz logistic regression, random forest classifier and linear support vector classifier were used to build three binary models which classified the defenders in CL and NCL categories based on 20 performance statistics selected on the k-best feature selection algorithm. The results suggested that elite defenders play the short passing game in high areas of the pitch and keep a high number of clean sheets. Sub-elite defenders on the other hand have a tendency of going long and having a high total volume of passing. These findings might be a result of the stylistic preferences of the elite teams and a reflection of their domination of ball possession. This study provides the first step towards automated talent identification and recruiting but further research on an event level seems to be necessary before translating the findings into the industry.
© Copyright 2023 13th World Congress of Performance Analysis of Sport and 13th International Symposium on Computer Science in Sport. IACSS&ISPAS 2022. Veröffentlicht von Springer. Alle Rechte vorbehalten.