By Kevin Minkus (@kevinminkus)
There’s been a decent amount of discussion this week about how the pace of play in MLS looked quicker in week one than it typically does. Teams like Atlanta, New York Red Bulls, Kansas City, and Houston all came flying out of the gate, with fairly up-tempo styles of play both with the ball and without the ball.
Unfortunately, coming up with a metric for pace is pretty tricky, and it depends specifically on what type of pace you’re talking about. Going all the way back to 2013, Ted Knutson looked at pace as the total number of shots taken in a game. More recently, Thom Lawrence looked at pace as the distance covered over time within a team’s possessions. Both of these definitions speak to a certain amount of directness of play that I don’t think meshes with what people currently mean when they say MLS is playing ‘faster’ so far this year.
Alex Olshansky has done a lot of work to define tempo within MLS based on the number of possessions in a match. This is probably a pretty good stand-in for the type of pace MLS fans and analysts are talking about- a faster pace means the ball is changing between teams more quickly. In a similar vein, Matthew Doyle recently used total passes per game as a shorthand for MLS’s pace. Generally I think the rationale there is that the more passes made, the quicker and more often the ball moves, and the quicker a player makes a decision about how the ball will move, and therefore the faster the pace. This, I’d suggest, is what is currently meant by ‘pace’, and therefore I think the logic behind the metric is sound.
Going by this definition, we do in fact see an uptick in pace in week one of 2017 over week one of the two seasons prior. There were 944 total passes attempted per game in week one this year, versus 906 in 2016 and 895 in 2015 (these are a bit different numbers than what Matt Doyle has, since the data is different).
Here’s where the total passes of the 11 games of this past weekend fall on a density plot of all games played since 2015:
While not completely out of the ordinary, those games are definitely skewed towards the higher end, so that confirms our intuition about the weekend.
The question kind of naturally arises about whether the league gets more or less ‘pacy’ as the season progresses. Looking at the moving averages by game day for this stat over 2015 and 2016, we can see that there’s not any discernible pattern throughout the course of the season. The plots are pretty noisy:
That also probably makes sense, and lines up with what we tend to see on the field. If we look across leagues, too, total pass numbers match expectations. Using WhoScored stats (rather than ASA’s), the EPL and Bundesliga both average about 940 passes per game, to MLS’s 810.
The metric starts to breaks down, though, when we look at it in finer detail. Here’s that stat for each game from last weekend:
|Home Team||Away Team||Total Passes|
The chart doesn’t quite pass the eye-test. Intuitively, I want to say that Atlanta United - New York Red Bulls and D.C. United - Sporting Kansas City were two of the faster games of the weekend, and that Columbus - Chicago definitely shouldn’t be up there. So what’s the deal?
One element that definitely factors in here is the amount of time the ball is actually in play (something Howard Hamilton, among others, has looked at). That number isn’t spread uniformly across games, and, because of that, comparing total passes per game isn’t apples to apples. If the ball is in play more, that game will tend to have more total passes made.
A second factor is the type of passes being made. A team like NYCFC, for example, that knocks the ball around the back line in their own half a lot, will register a high number of passes, even though those passes don’t necessarily lead to a game that we’d say has more pace.
The third thing to point out is that pace is about more than just passes. Houston - Seattle, which included at least one very high-tempo half, featured a lot of Houston carrying the ball up the field at someone’s feet, rather than hitting passes to advance it. That movement and decision-making won’t be captured here.
That probably leaves us, then, with a metric that works well on a macro-scale to define the type of pace we want to talk about, but doesn’t quite hold up in a more granular context. Developing a different metric to address this is something I want to work on over the next few days and weeks.*
*It’s entirely possible someone else has already developed that metric and I just haven’t seen it.