Updating the Evolution of Playing Time

With the final weekend of the season coming up – dramatically branded Decision Day by the league office – this seems an appropriate time to update the Playing Time Evolution plots that I first published about a month ago.

As a refresher, these plots illustrate how a team’s lineup changes over the course of a season. Every player to have appeared for a team is represented by a line. The progression of the season reads from left to right. The vertical axis presents the cumulative percentage of minutes played through each game of the season. The link above has a more in-depth explanation of the process.

More detailed notes are below if you’re really interested – but lets start with the plots themselves:

The R script and data files necessary to create these plots can be found on GitHub.

One phenomenon that always strikes me when looking at these plots is the difference between teams like Orlando and Portland. Some teams, like the expansion Lions, shuffle their lineup so much that the most used player on the roster only plays 70-80% of the season. Others quickly identify a core group of players whose performance satisfies the coaching staff, and they ride that group the entire season.

It is tempting to see these two types of teams as “good” and “bad”, but that misses some details that muddy the waters. Seasons are long, and there are a myriad of reasons why a team may be forced to replace even stalwart players. Injuries, trades, new signings, and international absences can all force a re-shuffling for reasons that don’t necessarily reflect a lack of playing success. FC Dallas is the poster child for that in 2015, and their position atop the Western Conference standings argues strongly that it is possible to succeed with a squad rotation system.

Another detail that emerges from some charts is the existence of in-groups and out-groups. Columbus exhibited this briefly this season, while New York Red Bulls and perhaps Portland are the clearest examples currently. Teams that follow this pattern have a group of players in the upper half of the plot, and a group of players in the lower half. Players never (or only rarely) cross the divide.

The opposite condition can be seen with teams like Real Salt Lake, or FC Dallas. These teams are very permeable in their evolution, with some players rising and falling constantly. Teams may still have stalwart players – like Tony Beltran in Salt Lake City or Victor Ulloa in Dallas – but there isn’t a strong cluster of them. At any point in the season, at any position along the vertical axis, you’re likely to find a player transitioning either in or out of the playing squad.

I’ll post a final version of these plots once the regular season is over.

Leave a Reply

Your email address will not be published. Required fields are marked *