Unsupervised Method to Analyze Playing Styles of EPL Teams using Ball Possession-position Data

In the English Premier League (EPL) matches, a network of advanced systems gathers sports data in real-time to build a possession-position dataset. In this work, data fields from the sophisticated raw possession-position dataset were extracted and processed to build a transformed version of the raw dataset. This transformed version contains ball possession data from 3 areas and 9 zones of the pitch. Two experiments were run based on this transformed dataset, aiming to understand and analyze the playing styles of EPL teams. The analysis answers multiple questions such as, is the playing style of the top 3 teams (Manchester City, Liverpool, and Chelsea) same in both home and away matches, do away match conditions affect the playing style of teams, etc. Existing studies use multiple parameters such as goal scoring patterns, player performances, team performance, etc. to understand and analyze the playing style of teams. In this work, using just the ball possession-position data, the playing styles of teams were able to be derived. This reduces the usage of such multiple parameters to perform the same task, which is to understand and analyze the playing styles of teams.