Columbus – Dallas, and the question of lineup stability


Columbus Crew SC will play host to FC Dallas this weekend in a game that will be nationally televised on Fox Sports 1. While the game is still several days away, and will feature two clubs in third place in their conferences and fighting for playoff seeding, much of the conversation ahead of the game has instead focused on absences.

FIFA has declared the period from August 31 – September 8 as an international game window, allowing confederations and national teams to schedule qualifying and friendly games. Most top-flight leagues go on hiatus during these periods because of the number of players who must be released for international duty.

Major League Soccer does not suspend play, and so Columbus and Dallas are each preparing to enter Sunday’s game significantly short-handed. A total of 14 players between the two teams could be missing.

A closer look at the issue

To help put the scale of this issue in context, I’ve put together two plots that focus on the distribution of playing time. Here is the first:

playing time - dallas - missing players


The structure of the plot is relatively straightforward, although its format is one I’ve not seen published anywhere. Each line on the plot represents a player on the roster (in this case FC Dallas), and the % of the season in which that player has seen minutes. Each game of the year is laid out from left to right, so you can track how a player’s playing time changes through the course of a season.

Here is the second plot, for Columbus Crew SC:

playing time - columbus - missing players

If we examine the plot for Columbus, and look for the line depicting Tyson Wahl (in purple) we recall that Wahl did not appear in the first three games – and thus he is plotted at 0% of the season through those games. From the fourth through 21st games of the season he appeared regularly, but not permanently – and so his line wavers between 20-30% of the season. Starting around game 21, however, he supplanted Emanuel Pogatetz at center back and has played consistently since. This change in fortune is visible as his line soars from ~28% up to now ~48% of the season.

At the upper end of the plot, we can see Emanuel Pogatetz suffer the inverse fortune. After playing the entirety of the first three games, his percentage of minutes played starts to slip – not drastically, but he gets substituted here and there. Then, around game 21, he gets benched and has not appeared since – so his line plummets and is currently sitting around 60%.

There is another layer to the plot above, however. A number of players appear in gray, with their lines dotted. These are the unavailable players this weekend – either because they have been called to international duty, or because they’ve left the team permanently.

A comparison of these two plots, side by side, reveals some interesting patterns:

playing time evolution - combined - missing players

It is immediately apparent that while Columbus has a significant number of absences, their players still available have appeared for a greater portion of the season than have those from FC Dallas. Six Columbus players have played more than 80% of the season thus far:

  • Steve Clark
  • Tony Tchani
  • Kei Kamara
  • Ethan Finlay
  • Michael Parkhurst
  • Federico Higuain

Of these, only Kamara is unavailable on Sunday.

Compare this core with the experience of FC Dallas. Only two players on the Hoops’ roster have appeared in more than 80% of the season – Victor Ulloa and Fabian Castillo. Now recall that Castillo has been called up by Colombia, and will not be available on Sunday.

Without this stable core of players, the Dallas roster has experienced significant churn, and as a result the middle of this plot is much more congested. While the Columbus plot is sparse in the middle (although players like Wahl, Pogatetz, Saeid and Trapp are threatening to bridge the gap), a total of 13 players for FC Dallas have percentages of playing time in the middle third of the plot.

What could this mean?

There are a few ways to interpret these data points, and it isn’t clear to me which is the most appropriate. On one hand, the fact that Columbus has so many core players still available could indicate that they can weather the loss of several players without too much trouble. Their foundation is still intact. Indeed, several of the missing players – such as Cedrick Mabwati, Harrison Afful, and even Wil Trapp – are relatively recent additions to the team. Afful recently replaced Hector Jimenez at right back, while Trapp was replaced adequately by Mohammed Saeid while he recovered from concussion symptoms. Even the double loss of Cedrick and Justin Meram on the left wing reveals Kristinn Steindorsson waiting in the wings, an Icelandic international who has himself played 30% of the season.

On the other hand, a Columbus team that – despite some fluctuation – has remained so constant is still facing an unprecedented number of changes to its roster. Their very stability could prove to be detrimental if they turn out to be unable to handle this, their most significant shock.

The Dallas roster, meanwhile, has not had a chance to solidify around a core group of players. Their lineup has changed much more regularly, allowing a greater number of players to see more significant playing time. This churn, combined with the fact that Dallas is only missing five players to Columbus’ nine, could indicate that Dallas will weather their absences with greater level-headedness.

This second interpretation is worth considering when we look at three final plots.

A final measure

In order to wrap up the question of roster variability, I calculated a Herfindahl-Hirschman Index (HHI) for the team rosters after every game. In economics, this measures the concentration of market share in a sector of the economy. On a soccer team this is a way of illustrating, in a single data point, how variable your lineup has been. Here is a game-by-game comparison of the HHI score for Columbus and Dallas this year:


Sure enough, while Columbus experienced some upheaval over the first few games, after the first two months of the season their lineup has been much more stable than FC Dallas. And yet, Dallas’ performance has not appeared to suffer:points ppg

(The code to generate last three plots has been posted to GitHub as an R script)

Using either points earned or points-per-game, Dallas has consistently led Columbus in one of the only metrics that will be remembered – game results.

This realization, that Dallas has outperformed Columbus over the course of the season while simultaneously dealing with a greater degree of turnover to their lineup – of exactly the sort now faced by both teams – gives me significant pause heading into this weekend. This game appears to be a far stiffer test than the team has faced recently.

While nothing in this analysis gives Columbus no hope of winning, this setup for Sunday’s game does give observers a chance to examine the Crew SC performance closely. Will the team rise to the challenge and defeat a more practiced, accomplished opponent? Doing so could catapult the team into contention for the Eastern Conference championship, while a slip-up threatens to undo the positive momentum that has been building around the team of late.

1 comment

Leave a comment

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