Every football/soccer team fields 11 players at a time. A very basic grouping of these 11 players would be to break them down into goalkeepers, defenders, midfielders, and forwards. To gain a slightly better understanding of how these players are utilized, you could break them down into more specific positions: defenders into FBs and CBs, midfielders into DMs, AMs, CMs, and WMs, and forwards into RWs, LWs, and CF. Even after doing so, you still wouldn’t have a thorough understanding of their stylistic differences, how these players are utilized and how they complement each other. In this series of daily (weekday) articles, I’ve set out to do just that using fbref data from the 2020/21 season. This first article serves as an introduction to the project and an outline of the series.
The Premise
Partly inspired by both Aneesh Namburi’s “Blueprint” and Ben Starks’ “Position-less Basketball”, the idea behind this series was to sort players by their style, find out how the best teams are built, and then understand why it is that they are successful. To do this, I used Orange Data Mining’s clustering tools to cluster/group the players into new “positions”. I did this because current positional labels can be misleading. For example, Scott McTominay and Paul Pogba are both technically midfielders, but they have vastly different skillsets and roles. Once the clustering was complete, I was left with 18 clusters of players, each with their own defining style and skillset. This series will cover each cluster individually before the final piece analyzing and understanding how players from different clusters work with each other on a winning team. Please keep in mind that the clustering model is not perfect and that some players will not be a perfect fit within a certain cluster. As always, feel free to comment below or contact me on Twitter if you have any questions, subscribe to the newsletter if you haven’t already (it’s free!), and make sure to return tomorrow for “Cluster 1- Limited Creation Workhorses”.