Editor’s note: The MLS partnership with Opta Sports has resulted in richer ways to understand the performance of the Philadelphia Union and its players through statistical analysis. PSP’s efforts to collect and share match statistics have been enhanced by the deeper analysis offered by Ford Bohrmann. We are now excited to share the work of Rolando de Aguiar and welcome his contributions to our endeavor to better understand what is happening on the pitch.
Ed Farnsworth wrote an interesting post in early December looking at the Union’s passing numbers. I took this data and broke it down a little more, in order to understand it better.
The relationship between usage and efficiency is an interesting one; i.e. what happens when a player is required to do more of x in a game? Usually, his efficiency decreases in that area. (Exploring the nature of this relationship is a small industry in basketball analytics.) The intuitive reasons for this are straightforward: when he doesn’t have to do much, he does what is easy; when he has to do more, he has to do some things that are harder. In taking on difficult tasks, he has more trouble finding success.
Ed’s data showed passing efficiency (pass completion percentage) and usage (pass attempts), but the latter was not adjusted for playing time. The chart below (for outfield players with > 500 minutes) adjusts for playing time.
That data allows us to plot passing usage (i.e. number of passes per 90 minutes) vs. passing efficiency.
Importantly, note that the players appear in positional bands. The natural shape of this data suggests to me that a huge component of these passing statistics is structural—passing accuracy and pass frequency are largely dependent on the position the player is playing. However, we can see clear differences between players within these bands.
Within these bands, the players at the bottom right (e.g. Sheanon Williams among defenders, Roger Torres among attacking midfielders, Sebastien Le Toux among forwards) are high usage, low efficiency passers, while those at the top left (Danny Califf, Justin Mapp, Danny Mwanga) are low usage, high efficiency. To put that in everyday terms, Torres (and Williams) make a lot of passes, and not very many are successful. They take risks when passing. Califf and Mapp were more conservative.
The following is a dendrogram generated from a cluster analysis of the data above:
Since I included only two variables—passing rate and passing accuracy—in the analysis, it’s essentially a different view of the data above. Cluster analysis becomes much more useful when looking at a greater number of variables. See below, and let us know in the comments if this sparks any ideas.