By Jared Young (@JaredEYoung)
A few miles south on the list of soccer statistics is a little number called minutes played. It doesn’t stand out much and feels more obligatory than anything. Of course we need to know how much a player plays. Then we go ahead and divide it by 90 because “per 90” puts all the players on the same scale. That’s better. Minutes played is used and brushed aside just as quickly. But I came to find out that minutes played might tell us something we didn’t know about the good teams and the bad.
This story begins not with me trying to resurrect minutes played, but with me wondering how to understand the importance of a team’s depth. The first thing I wanted to do was to look at the distribution of minutes across a team. Is there a healthy level of minutes played for starters and for reserves?
To simplify the calculation I added up the total minutes played of the top eleven players on every MLS roster and divided that number by total minutes. It turns out about 75 percent of the total minutes played are accumulated by the eleven players on a team who played the most. The highest percentage in 2014 belonged to DC United (79 percent) while the lowest percentage belonged to the Colorado Rapids (68.5 percent).
It was natural to wonder, given DC United’s success and Colorado’s up and mostly down season, if this simple metric would at all predict, well, anything. The best thing to do when one wonders is look at the data.
While there is a fair amount of variability in the plot, there is a significant Rsquared for a metric that is pretty far from having anything to do with goals. When you draw the likely playoff line of 50 points and cut it off at the average of 75 percent an interesting thing occurs. Just 1 of 7 teams with a ratio lower than 75 percent scored 50 points or more. On the flip side, 8 of 10 teams scored 50 points or more when having a ratio of 75 percent or higher. It looks like you want to get your best players as much of the playing time as possible. The teams that simply do that have a higher percentage of success.
Of course there are other factors to consider. For example, a team with a better starting XI will be more likely to limit reserve players’ time, as the quality of the starting XI demands they play. A team with a weaker starting XI would be more prone to test reserve players as the marginal difference between players may not be that great. So the minutes played distribution could definitely be a proxy for player quality. Injuries are another factor that would drive a lower percentage and cause a team to perform more poorly.
Nevertheless the 75% heuristic can be used to determine if a team is distributing the minutes in a productive way or if the reserve players are getting too much time (which may just be an indication that the starting players aren't all that good or are hurt more than usual).
For readers wondering if adding up eleven players is the right choice, as it is purely chosen for convenience and a desire to use roman numerals, the answer is no. It looks like how many minutes the first fifteen players play is actually most predictive of a team’s points.
Still, I like the XI player and 75 percent target as it’s easy to remember and gets most of the predictive quality. And if you didn't think there was thick cream on this milk quite yet, I’m not done. This minutes played distribution metric remains statistically significant when used in a multivariate regression with expected goals. You might think that expected goals would erase its predictive value as the variables would be correlated, but the Rsquared increases to 38 percent and the minutes played distribution has a slightly lower p-score. This metric holds up even when use in conjunction with a metric that’s closer to measuring goals which is closer to actually winning.
So minutes played has some use after all. An application for its use would be to look at last season’s minutes for a team’s starting XI and analyze how well-suited the team is to achieve at least 75 percent of the team’s total minutes (that’s 25,425 minutes in an MLS season). You could also look at a team’s reserves and see if they have enough quality to play the over 8,000 minutes they’ll likely need to play in a season.
But even if you don’t take it that far, perhaps the next time you look at a soccer boxscore you won’t be so quick to glance past that first lonely and misunderstood statistic, minutes played.