In the dynamic sport of football, central midfielders play a crucial role in orchestrating the flow of the game. Most of the responsibility to get the ball into the attacking third lies with them, while they cannot afford to lose the ball in dangerous areas.
An aspect that sets central midfielders apart is their passing style, which can vary significantly from player to player. In this article, we use a soft clustering model to gain deeper insights into the distinct passing styles exhibited by central midfielders in the Premier League 2022-2023. We will focus on passing metrics of central midfielders who played at least 450 minutes, a total of 70 players.
To determine the optimal number of clusters for our analysis, we apply a clustering algorithm to 25 passing metrics, including several unique to PFF, for an increasing number of clusters. At five clusters, the smallest cluster becomes too small to provide meaningful insights, meaning four clusters are suitable for categorising the distinct passing styles of the central midfielders in our dataset.
Before we dive into the different clusters, our soft clustering approach also allows us to find the most versatile passers by looking at the players with the smallest deviation between clusters, which turn out to be Mathias Jensen (Brentford), Kevin De Bruyne (Manchester City), Curtis Jones (Liverpool), James Milner (Liverpool, now Brighton & Hove Albion) and Bernardo Silva (Manchester City). On the other end of the spectrum, we find central midfielders that have a more distinct passing style that is nicely captured by the model. This group is led by Jorginho, who belongs to the first cluster for a staggering 77%.
The first cluster consists of central midfielders who take on a pass-heavy approach, attempting more passes per-90 on average than the other clusters. Notably, they also stand out for attempting many line-breaking passes, demonstrating their ability to create opportunities and break through opposition lines. Compared to other clusters, the degree to which players belong to cluster 1 is strongest, suggesting it represents a distinct passing style. Furthermore, these central midfielders have the highest average PFF passing grade of 81.1, which isn’t part of the clustering model but highlights their exceptional passing skills.
Example players: Jorginho (Chelsea/Arsenal, 77%), Thomas Partey (Arsenal, 73%), Pierre-Emile Höjbjerg (Tottenham Hotspur, 64%), Youri Tielemans (Leicester City, now Aston Villa, 63%), Mateo Kovacic (Chelsea, now Manchester City, 57%).
In contrast to cluster 1, the central midfielders belonging to cluster 2 have a significantly lower average number of passes per 90. However, they still maintain a comparable relative frequency of line-breaking pass attempts as seen in cluster 1, but with a lower accuracy. What sets this cluster apart is their inclination towards passes that aim to create a contest, which are more speculative in nature. Compared to other clusters, these players take less touches per possession. Unfortunately for the players in this cluster, the quality of their passes turns out to be lower with an average PFF passing grade of 67.4.
Example players: Boubacar Kamara (Aston Villa, 50%), Jefferson Lerma (Bournemouth, now Crystal Palace, 50%), Lewis Cook (Bournemouth, 50%), Cheick Doucoure (Crystal Palace, 47%) and João Palhinha (Fulham, 47%).
The central midfielders in this cluster have the highest pass accuracy, showcasing their ability to consistently deliver precise passes on the field. While they attempt fewer line-breaking passes in general, they try many defensive line-breaking passes, which is demonstrated by their relatively high number of through ball attempts. Despite a slightly lower average PFF passing grade of 78.8 compared to cluster 1, these midfielders exhibit a high level of passing proficiency, further solidifying their impact in dictating play and creating scoring opportunities for their teams.
Example players: Rodrigo Bentancur (Tottenham Hotspur, 36%), Fred (Manchester United, 36%), Douglas Luiz (Aston Villa, 36%), Declan Rice (West Ham, now Arsenal, 36%) and João Moutinho (Wolves, now free agent, 36%).
Cluster 4 represents central midfielders who exhibit the lowest pass accuracy among the clusters analyzed. While they share a similar frequency of defensive line-breaking passes and through balls with cluster 3, their execution of these passes is less effective. Notably, players in this cluster stand out for attempting more flick-on passes, showcasing their willingness to quickly distribute the ball. Together with cluster 3, these midfielders also have the highest number of possessions in the opposition box. Their average PFF passing grade of 72.9 indicates a lower overall passing performance compared to clusters 1 and 3.
Example players: Sean Longstaff (Newcastle United, 51%), Joe Willock (Newcastle United, 48%), Josh Dasilva (Brentford, 45%), Conor Gallagher (Chelsea, 45%) and Joelinton (Newcastle United, 45%).