Find the leak! Analyzing PSV’s midfield…

Three men in suits, one interviewer, Sky Sports One, ninety minutes, pure inspiration. Sometimes you just need a little spark to get your writing going again, and for me, this show worked. In case you missed it, you can listen to the podcast version of this interview with Sam Allardyce, Damian Comolli and Chris Anderson here. Three men with different backgrounds, yet one conviction: the use of data provides you an advantage in the playing field that is professional football.


West Ham manager and football stats pioneer Sam Allardyce

Ironically, when I listen to something I am immensely grasped by, instead of concentrating on the presentation at hand, my mind goes off wandering. This also happened during the show. There were so many great talking points that I could hardly resist chewing on several concepts, at the risk of missing content.

Do these men on TV really focus on shots on target, ignoring the heterogeneity of that parameter? Are clean sheets really that important that focusing on defense over offense pays off? And who should be the Dutch equivalent of Allardyce, a manager often portrayed less sensible than he is, convinced of the use of data? Which two other guys would be able to fill in an talk sensible analytics on a Dutch TV broadcast?


The bucket game

But the main thought that crept in my head was of an old game we used to play on summer days when I was a little kid. You were handed a pretty large bucket of water, with the sole and simple goal of bringing this bucket over the finish line, with as much water as possible still left in it. This was complicated by the fact that the bucket contained some holes, small and larger ones, which allowed the water to leak out freely, if you didn’t control the holes with your fingers.


But controlling the waterholes required the use of at least one hand, which significantly slowed down the process of walking the bucket towards the finish line. Most kids started out at a frantic pace, allowing the water to leak freely, but rushing the bucket to the finish line, only to find out that this was not the optimal strategy. On the other hand, concentrating too much on keeping the holes as tight as you could did not win you the game either, as the bucket was created to prevent you from stopping all the leaks simultaneously. Therefore, a balancing act between stopping leakage and keeping the pace going was the best.

So, why am I telling this long introduction story? Because it makes up a perfect metaphor for a football match, and it that sense it kept creeping up my mind during the footy stats debate on Sky. Football teams are just like the water bucket. The water is their possession of the ball. Some teams focus too much on preventing the water from leaking out and fail to create enough output, while others rush to the finish line, losing a lot of water along the way.



This theoretical metaphorical talk will probably work better with a real life example at hand. Let’s focus on what’s presently my favorite study object, PSV, and try to implement some of the detailed performance statistics that we can gather from various sources around.

PSV are an interesting case study, because, as you’ve probably noticed, things are not entirely going according to plan. Instead of competing to the title, they have hit a two-wins-from-twelve streak and are currently back in 10th position. Pressure on young manager Phillip Cocu is building and it remains to be seen if Cocu is allowed to even finish his debut season.

As I wrote for De Volkskrant, PSV have a huge offensive problem, and that is part of the reason I’ve picked them for this case study. After all, defensive stuff is still next to impossible to analyze with public data focusing on on-the-ball events. PSV create the 3rd most shots of the Eredivisie, yet their horrific shot quality makes them only rank 11th in terms of Expected Goals scored.

Let’s assume PSV starts with a full bucket of water – there’s more in this metaphor as the point of regaining possession may represent the amount of water to start with – and conclude that they don’t carry enough of it over the finish line, or in other words, they don’t create enough shots of decent quality.


Now, where do PSV leak?

At around 380 completed passes per match, PSV are well behind Ajax (530), Twente (450) and Vitesse (410).  In the reverse stat, passes completed by the opposing team, PSV (270) is at a level playing field with Vitesse (260) and Ajax (270) and slightly ahead of Twente (300). So, PSV is less of a passing side, but out of possession it keeps the pressure on, like the other top teams. I’ve deliberately left Feyenoord out for now, not because I don’t rate them at the same level as said teams, but because their profile is quite different, which makes them a nice case study for later.

Another interesting way to look at passes is the amount of incomplete passes. Variation here is self limiting as you can’t keep losing huge amounts of passes, since in between each lost pass, you’ll need to win the ball back first. So, less spread among the teams, and PSV comes in nicely at 86 incomplete passes per match, ahead of Ajax (88), Twente (90) and Vitesse (94).


Where is the leak?

Now, where do teams leak the water? Which players are responsible for the mainstay of misplaced passes? Naturally, going from front to back, outfield players should have a decreasing amount of misplaced passes, as defenders have less opponent pressure to worry about when passing and the consequences of their misplaced passes can be more severe.

PSV’s top players in terms if misplaced passes are Stijn Schaars (11.3 per match), Jetro Willems (9.7) and Adam Maher (9.2). Yes, you read that correctly, two of PSV’s midfielders are responsible for 20 misplaced passes per match.

The context for this number consists of Ajax (Poulsen 4.9, Blind 7.0), Twente (Gutierrez 8.3, Eghan 6.4, Ebecilio 5.2) and Vitesse (Atsu 6.1, Vejinovic 4.8). PSV ‘s rivals lose around 11 to 12 passes in central and defensive midfield, PSV around 20.

So, the problem is not the amount of passes going astray, but probably the area where they do.

This would be okay-ish if the likes of Maher, Toivonen and Schaars had enough offensive output going to make up for it, but that is definitely not the case. PSV’s strikers get a horribly low amount of touches on the ball, which may be partly their own fault, but if I were to manage a football team, the midfielder should know it’s their job the make sure the striker sees enough of the ball.

Matavz (12.4 passes per match) and Locadia (13.8) are bystanders in comparison with Sigthorsson (26), Havenaar (25) and Castaignos (23). If all your rival teams note performances that are this close in the same range, you’re probably the outliers, and not in a positive sense.


In the end

Overall, it seems PSV are leaking to much water to get a top level offense going. The paper I wrote for De Volkskrant already diagnosed the horrific shot quality, with main contributors Depay, Bakkali and Maher. Today’s article on 11tegen11 goes into more detail and highlights their water leak in midfield, with Schaars and Maher misplacing nearly twice as many passes as rival midfielders do. As long as Cocu does not fix these type of issues, the leak will continue and results will continue to surprise in a negative sense. Performance data are here to help, and this type of analysis is an example of what quite superficial passing information can already tell you.

3 thoughts on “Find the leak! Analyzing PSV’s midfield…

  1. Mo

    Good stuff. I am curious to know how this divides into different types of passing in midfield (penetrating, backwards, sideways etc.). Not all midfield passes are created equally. How is the movement off the ball? Is there a clear pattern of play/philosophy? The understanding would be greater if there are more supporting stats which puts the high number of misplaced passes into a greater context. Besides, isnt misplaced passes as a percentage of total passes a better comparable than sheer number of misplaced passes? If i misplace 5 of 10 passes (50%) then that i surely worse than 7 of 50 passes (14%), no?

    1. 11tegen11 Post author

      Thanks for the comment, Mo.

      You’re absolutely spot on with the remark about context. In my view, this type of broad scanning analysis only serves to identify issues for further examination. Once we’ve now seen that Schaars and Maher leak passes at almost a double rate compared to the rival teams, we need to study those passes in more detail. Things like direction, length, match situation, score line, etc. All of this is needed to check if this is a real issue, or may be something that can be explained in terms of context. It’s not like those two players are technically not capable of completing passes, it’s something tactical at hand.

      On the other hand, I don’t agree with your idea about using relative passing stats. This would not identify Schaars, for example, as he has 60 pass attempts per game, of which 11.3 are incomplete, for a completion rate of 81.2%.
      In my view, his complete passes don’t negate the fact that his absolute incomplete pass number is too high. Making more safe passes does not compensate for incomplete passes.
      Of course, if Schaars had 20 passes per game at this rate he would have only 4 incomplete passes, but that would signify a complete lack of involvement from a defensive midfielder.
      The only way to compensate for more than 10 incomplete passes per match is by way of key passes and passes into danger zones like the opposition penalty area. Schaars has 2.8 of those combined per match, with 35 Eredivisie players (> 270 mins) doing better than that. And only 6 of those have more incomplete passes…

      1. Mo

        Thank you for your reply. Born and raised in Norway but having spent 15 out of my past 21 years abroad (all but one in the US) I have a deep appreciation for statistics when trying to understand sports. Unfortunately, a lot of football people (fans, writers, players, coaches, administrators) look at statistics in a very black and white way. Either they dismiss it or they believe in it. Instead they should merge the two views. Football is a complex picture where statistics (along with other traditional and non-traditional parameters) help give us a better underanding of why things are as they are. Twitter has introduced me to a group of people (like yourself) who have an amazing curiousity about what the numbers can tell us and luckily you want to share your work with the rest of us. Thank you.

        You are right that looking strictly at the completion percentage is not ideal either. Instead, we need to look at a number of variations of the same number to get a better understanding of what that number is telling us. In Schaars’s case, total number of passes, misplaced passes, the percentages and so on. We need to be able to juggle several different nubmers, giving us several different angles to the same question before we pass judgment. It is not meant as criticism of this piece (your work is great and I hope you continue to publish it) but rather a personal realization of the difficult balancing act you, as a writer, has to master when writing about complex matters in a less complex way so that the reader can better understand what it is that you are trying to convey without oversimplifying it and losing the essence of what it was that you want to say. Finding the right balance is very, very difficult. The topic is incredibly interesting and a thousand times more interesting that the regular football discussions seen in mainstream media.

        I didnt realize until reading your reply to my comment that Schaars was a defensive midfielder. That tidbit of knowledge obviously changes how I view him and how I interpret his numbers. Generally speaking, you expect to see a higher completion percentage and a lower number of misplaced passes from a def mid compared to an attacking midfielder, at least within the same team. For different players, from different teams, variables such as playing style, away/home game etc matter as well. Knowing these variables help paint a broader picture and help me better understand the issue discussed. These variables challenges me as a reader and you as a writer. Football statistics and writing about football statistics is still in its infancy and i think both writers and readers are searching for a basic, common understanding/platform on which to share/discuss/better understand the game we love. Personally, I cant wait to see what the future brings. It is by far the most exciting development in football today and it is great to see that people like yourself are willing to lead the charge. It isnt easy but you are doing a tremendous job.


Leave a Reply