The detailed method is outlined here.

In brief, every pass is assigned to a cluster by k-means clustering. For each cluster of every team, a z-score is computed using the # of passes in that cluster divided by the total # of passes. The highest and lowest-ranked clusters by z-score are plotted.