“Back in November we applied a clustering algorithm to find out which Premier League clubs had similar attacking styles. We wanted to see what we could find using match summary stats that anyone with an internet connection could get hold of. Our main rule was that we wanted to avoid using pure outcome stats, e.g. shots on target, completed passes, completed crosses, goals, assists etc. We thought we’d run the risk of just clustering teams together on how good/lucky they’d been so far. We didn’t use anything too fancy, just per game stats based on the way teams attempt to attack; shots from outside the box, inside the box, open play, set pieces, short passes, long passes, dribbles, crosses and how much they use the wide areas when they attack. …” StatsBomb
