By A.J. Barnold (@ajbarnold)
My opening post on my blog “came in stats up” was on the issue of substitutions and season length in college soccer, which I analyzed through a breakdown of passing styles in MLS and the ACC. If you haven’t already read that article, please do – it is an important primer for what you’re about to read. Here’s a brief summary for those of you who choose not to, centered around the chart that got everyone talking:
In a brief follow-up, I found that although college games sometimes feel more hectic in the second half, stats such as interceptions, fouls, and cards do not increase or differ much from MLS games, leading to the idea that maybe college players resort to dumping more long balls as the game progresses. Now that we’re all up to speed, let’s explore that idea using actual numbers.
Long Passes Attempted – MLS vs ACC
We’ll use the same datasets that went into the original post – Prozone data for the 48 ACC regular season matches, and WhoScored.com’s MLS data. I previously established the fact that MLS teams attempt around 60 more passes per game than ACC teams. To get more specific, what about long passes – balls played over 25 yards? Do college games involve more long balls than MLS games?
In short, yes:
Teams attempt about 16 more long balls, per team, in each college game. What grabs my attention here is that the long ball trend looks like a mirror image of the total pass trend. The gap between the ACC and MLS starts off large – four to five more long balls in each segment of the first half. It then dwindles to around one more in each portion of the second half. Contrast this with the gap in total passes, which was relatively close for the first half, then widened massively in the second.
There is a simple explanation for this – it’s not that ACC teams forgo passes over 25 yards in the second half. They just play significantly fewer passes overall, which naturally brings down the number of long passes as well.
Fair enough, but this doesn't fully answer our question. College teams do not attempt more long balls in the second half than in the first – but that can be explained by the lower pass total. What about the proportion of total passes played long?
Overall, 7-8% more of the passes go long in a college game than in an MLS game. This makes sense given what we just learned about long ball numbers. Additionally, neither league sees any huge variation in the proportion of long balls over the course of a game. In fact, the proportion goes down in the college game ever so slightly in the second half, with the biggest change (2.5%) coming in the last 15 minutes. This corresponds with the overall passing numbers as well, which increase as the game comes to an end and teams look to secure a result.
This spike at the end of games is interesting, if not somewhat expected. For college games, it matches closely with the average length of a pass over the course of the game as well. In the first 75 minutes of the game, the average pass length in ACC games remains relatively constant, somewhere between 18.5 and 19.5 yards. At the end of the game, this distance surges to over 26 yards.
In some sense, this finding provides an explanation for why the late stages of a college soccer game feel hectic and crazed. The final 15 minutes feature the largest proportion of long passes, as well as the longest average pass distance. This fits into the (completely unscientific) notion I share with many of my colleagues – players go mad in the final few minutes of a game because they can see the clock counting down to 00:00. They eventually hear the announcer counting down the final 10 seconds. This all seems to bring a sense of impending doom that leads players to welly the ball as far as they can – whether to get it forward in hopes of scoring, or in the “anywhere will do” mode of defending.
This frenetic feeling is only exacerbated by the fact that this period of long ball comes after the slowest 30 minutes of passing in the entire game. While the pro game also has its highest proportion of long passes in the final segment, it does not come after such wild variation in pace of play (or percentage of long balls) over the previous 75 minutes.
So… perhaps we have achieved some sense of why the college game feels crazy in the second half, even though the number of passes attempted drops significantly. But this “long ball theory” hasn’t really helped us answer the question we started with. There aren’t more long balls over the course of the second half than in the first, so the idea of more time being wasted by longer passes doesn’t pan out. Why do the college passing numbers drop off so drastically in the second half?
Ball In vs Out of Play
All of this digging through the ACC data led me down another rabbit hole. For each game, Prozone provides the possession percentages as well as the amount of “effective playing time” – how many of the 90 minutes the ball was actually in play. One can also tease out the percentage of time that the ball was out of play. Here’s what that looks like over the course of a game:
To add some context to this chart, a 2012 study of the German Bundesliga found that the average match stopped for 38% of the total time – 2.5% lower than the ACC average of 40.5%. More to the point of the second half passing drop off, ACC games jump from 38.6% ball out of play in the first half to 42.4% in the second. The final stage of the game shows the highest percentage of ball out of play at 44.5%.
Finally a bit of a clue – college games slow down in the second half with more stoppages than the first. This would obviously slow down the rate of passing, but to the extent that we've seen? And is this slow down really any greater than seen in the pro game? We need to go a step further and look at the actual timing of matches, rather than simply percentages.
This is where the college game would presumably get a bit messy – as suggested by a comment on the original post, we have the issue of clock stoppages. The college clock stops almost solely at the referee’s discretion – always for bookings, occasionally for time wasting tactics, sometimes for substitutions (in the last five minutes of the match), and perhaps most frequently for injuries. All of this occurs without any time added on at the end – when the clock hits 00:00, the match is over. This would complicate any comparisons of time spent on stoppages between college and pro games. Fortunately, Prozone’s data is helpful in this sense once again.
Since it originated in the professional game, Prozone (and most other companies) time the college game as though it were on the same clock as a pro game. The clock starts at kickoff and runs without stopping until the half ends. This helps our investigation in two ways – one being that it puts all of the timing comparisons on a consistent system, and the second being that we can essentially see how much stoppage time occurs in a college game. Anything over 45 minutes would effectively be the same as time added onto the half in a pro game.
Combined with the ball out of play percentages from the previous graph, this brings a great deal more insight and context to the numbers. Here is a graph of effective playing time for each portion of an ACC match:
Looking at the first 30 minutes of each half, the effective playing time (green bar) steadily decreases over the course of the match. It only appears to increase in the last portion of each half due to the longer total length created by the time added on. Look at the ball out of play percentages though – the proportion of effective playing time never comes close to getting back to the highest level seen in the first 15 minutes of the match.
The amount of time added onto each half jumps out as well. The average ACC match in 2015 lasted 98 minutes 51 seconds. I think you may be hard-pressed to find a professional league where the average match goes for nearly 99 minutes, including first half added time. FiveThirtyEight’s investigation of stoppage time suggests that the average amount of time added to the second half in the major professional leagues is around four minutes. Specifically, MLS appears to be higher – around four-and-a-half minutes. Even the highest average amount of stoppage time at any one Premier League stadium in 2014-15 only reached five minutes 16 seconds – still short of the five minutes 33 seconds added to the second half of ACC games. College soccer games – particularly in the second half – are longer, slower, and more drawn out than the professional versions.
Ostensibly, a slower game and more stoppages will promote a decrease in the number of passes attempted in a match. So… why so much time with the ball out of play? For me, this brings us back to the main structural difference in a college soccer game – substitutions. In the first post, I discussed Conte et al.’s research on Division I substitution patterns – NCAA teams averaged 11 substitutions per game, with nearly 60 percent of those being re-entries. The researchers broke these numbers down further into similar time periods as our passing analysis
Almost 90 percent of substitutions occurred after the 30th minute – the exact same portions of the match in which we see a decrease in the proportion of effective playing time. More substitutions take more time, slowing down the game and leaving less time to play… and therefore less time to accumulate passes. Finally – some insight into college soccer’s passing drop off!
Further complicating matters, many of these additional substitutions occur during bookings and injuries (according to the NCAA, significantly more injuries occur in the second half than in the first), or during the last five minutes of the match (when the clock stops for a sub by the team with the lead). These all represent events that help contribute to the large amount of “stoppage time,” and generally stretch the game out in the second half. Although professional games likely see the majority of substitutions in the same stages as the college game, the maximum number can only be six – college games average somewhere over 20.
Conte et al.’s findings further suggest that increased substitution numbers may also relate to a decrease in quality – ranked teams utilized around 10 percent fewer substitutions and re-entries than unranked teams. Furthermore, games between two unranked teams featured around 10 percent more substitutions than games including at least one ranked team. All of these findings mesh well with the ACC data’s indication of a decrease in pace of play and number of passes in the time periods when more substitutions occur.
It seems clear at this point that college soccer is simply s l o w e r than pro soccer in some sense, and that the massive drop off in the number of passes attempted in the second half has something to do with the length of time taken up by the increased number of substitutions. By decreasing the amount of time the ball is actually in play, and breaking up the spells of play, substitutions potentially act as a restricting factor on the quantity and style of passing in college soccer. I say “potentially” because my psych/stats professors burned into my head that correlation does not prove causation, but at no point will I be able to run a lab controlled experiment on college soccer games to prove these ideas. So I’ll put the suggestions out there for you, the reader, to think of them as you will. I think it’s a pretty reasonable conclusion.
Recent work by Dr. Hector Ruiz at Prozone came up with conclusions similar to mine as well, in terms of the stylistic differences between MLS and college:
In comparison with professional games, the level and style of passing – and many would say the overall level of play – in college soccer clearly drops off as the game progresses. The findings presented here suggest that more stoppages and a decrease in effective playing time could be to blame. Addressing this issue through changes in the substitution and re-entry rules could potentially have significant impacts on the style of the college game. Would this improve the level and development of our players? We may only find out by trying.
A.J. Barnold works as the Performance Analyst for the men’s program at the University of Virginia. Check out his blog Coming In Stats Up for more analysis, and follow his commentary on Twitter @ajbarnold.