How do team rosters change over time? Do players of a certain age tend to play more or less often? These are some of the questions I’ve pondered over the years. And they are the subject of a brief experiment I’ve worked on the last few days.
The chart below illustrates the how playing time has been distributed by the Columbus Crew over their history. The default view is for this season (2014), but you can switch to other season by pressing the left or right arrow on your keyboard.
All official games are included in these totals – not only league games but also playoffs, US Open Cup games, and international competitions like the Champions League. Exhibitions and friendlies are excluded.
There are some interesting aspects of this chart, but there are also some challenges. For one thing, it is possible to watch groups of players age. In the team’s first season there were a group of players in the early 20s that would form the heart of the team for almost a decade: Mike Clark, Brian Maisonneuve, Brian McBride, etc. As you step through each successive season you see the columns for those players – born in 1972 and 1973 – progress through the heart of their careers. It isn’t until 2000 that another group of players eclipses that group for playing time – and not until 2002 when a younger group truly takes over.
You can also see some outliers at both age extremes. Elder veterans like Robert Warzycha and Thomas Dooley can be seen marching toward their 40s – Dooley joined the team in 1997 as a 36-year-old, and played the next three years as the oldest player on the team. Warzycha was a few years younger, but can be seen following the same trajectory as a 33-year-old in 1996 until his retirement after the 2002 season.
Of course, grouping all players of the same age together also masks turnover. Danny Szetela joined the team as a 17-year-old in 2004, but his departure after the 2007 season isn’t noticeable because by then he was part of a group that included Robbie Rogers and Tim Ward. Emmanuel Ekpo joined the team the following year – by which point Szetela and Ward were long gone.
Ultimately, I’m still working out how best to use this sort of analysis. The visualization style itself (done with d3.js, modeling code after the Population Pyramid from Mike Bostock) will potentially surface later in some other investigation. The question of how teams build their roster over seasons, however, honoring the competing desires for longevity and new blood, will continue to tug at my mind.