In an article published last October I took my first look at some defensive metrics. That piece was very much an introductory one as I offered up a few of my initial ideas for consideration. I’m now going to take the opportunity to expand on one of the ideas that I wrote about in that initial article, Passes Allowed Per Defensive Action (PPDA).

In this article, with the use of detailed Opta data, I am going to create a metric that I believe can quantify the extent and aggression of high presses employed by teams, both over a season and in any specific match. I’m going to see if it is possible define the intensity of a press with the use of numbers, more specifically by using some of the events that Opta record.

 

Why would anyone want to do this?

Well, for pretty much the same reason that we undertake any analytics study. If we can develop an objective scale which measures the intensity of a press then coaches can quickly see at a glance the strength, or otherwise, of the high pressing that their opposition has utilised in recent games. Teams or fans can also assess how much pressure their team exerted on the opposition in deep positions, and who knows, perhaps in time we will be able to assess the effectiveness that individual players have on the ability of their team to press. In essence we can take what is otherwise a subjective description and reduce it to one number so that it allows for comparison, analysis and ranking, if so desired.

 

Passes Allowed Per Defensive Action (PPDA Metric)

Although I introduced this concept in my October article, I wasn’t able to analyse the defensive pressure by specific areas of the pitch – I simply had to use whole game stats. However, I can now extract the required information from the detailed Opta files and begin to create a meaningful metric. I have no doubt that this will be an iterative process and that modifications and improvements will be made in due course, but we have to start somewhere……

 

Defensive Actions

The defensive actions that I am using are the following four Opta defined events:

  • Tackles
  • Interceptions
  • Challenges (failed tackles)
  • Fouls

For this analysis I am not concerned about whether the opposition made or completed passes. I want to measure the pressure that the defending team puts on the opposition players when they are in possession of the ball; and I think those four stats suffice in this regard. I am developing a descriptive measure, and at this stage whether the press is successful or not is not the aim of this initial step.

The PPDA metric is calculated by dividing the number of passes allowed by the defending team by the total number of defensive actions; both values will be calculated with reference to a specific area of the pitch.

PPDA = Number of Passes made by Attacking Team / Number of Defensive Actions

A smaller PPDA value signifies a greater level of defensive intensity, as in essence, the defence has allowed a smaller ratio of uncontested passes to be made.

Not every pressing team will press all of the time; perhaps they wait for certain pressing triggers, other than just the position of the ball. However, this metric provides an objective measurement as it is calculated with reference to the number of times that it is possible for the defensive team to press, ie on all opposition passes in the specified area of the pitch.

 

Which area of the Pitch?

The major decision I had to make was in relation to the area of the pitch that I wanted to use to measure the intensity of the pressing. I am conscious that there is probably no right or wrong answer for this, but I tossed it around for a few days and I also asked Rene Maric from the excellent Spielverlagerung tactics focussed site. The short answer is that we both eventually settled on the pitch area that, in Opta parlance, has an x co-ordinate of greater than 40.

Pitch

Pitch image created by Valentin

We are looking at the pitch from the perspective of the team playing left to right, ie defending the goal to the left. We are using the Defensive Actions (tackle, interception, challenge or foul) that occur to the right of the line that is denoted by “x = 40”. So this includes all defensive actions in a team’s attacking half as well as a little bit into their own defensive half.

I outline the complete logic that I used to settle on x = 40 as the boundary at the bottom of this piece.

From now on, all PPDA values refer to PPDA Beyond40 (ie the metric only uses defensive actions and oppositon passes that occurred beyond the x=40 line) as that is the boundary I have decided to use to assess the intensity of the high pressing by the defending team.

 

So what do the PPDA values look like?

Mauricio Pochettino is one of the poster boys for the high press, and this is a great analytical piece on the Argentine manager written by Michael Caley.

The side he managed, Southampton have, over the last two seasons served as a very good example of what the PPDA values can look like.

Note: to smooth out the variations that any metric can show on a game by game basis I have plotted the PPDA values on a 6 game rolling average basis.

Southampton

Remember: Lower PPDA value means a high level of pressing

During Nigel Adkins’ spell as Southampton boss the South Coast club had a 6 game rolling average PPDA that ranged from just under 10 to 14. The new PPDA metric easily captures the impact that Pochettino had as soon as he arrived at St Marys. The Saints’ PPDA value fell in an almost straight line fashion during their first six games, with the result that the rolling 6 game value was less than 8 at the end of this six game period. We can see that Southampton maintained this level of pressing right throughout the remainder of the 2012/13 season.

However, this metric suggests that Southampton hasn’t been nearly as intense in terms of pressing over the last half of the 2013/14 season as they had been during the first 12 months of Pocchetino’s reign. Ironically their 6 game rolling PPDA value at the end of the 2013/14 season was higher than when Nigel Adkins departed from Southampton in January 2013. I don’t recall this being widely picked up or reported upon.

In fairness, Southampton had “nothing to play for” in their final 3 or 4 games of the season. Perhaps their drop off in pressing intensity at the end of the season, which is apparent from the above plot, was solely due to that. Regardless of the reason for their drop off in intensity, I assume they will be very disappointed with the PPDA figure of 14.5 that the Saints recorded in their home game against Cardiff towards the end of the season.

This really uncharacteristic performance, unsurprisingly, resulted in a 1-0 away win for the relegated Cardiff side; in fact it was only Cardiff’s second away win of the campaign. Southampton’s lack of aggression in attempting to win back possession may help explain why they lost that game.

 

Barcelona after Guardiola

Pep Guardialo is another famed exponent of the high defensive press, first at Barcelona and also now at Bayern Munich. He clearly believes in pressing aggressively in offensive areas in an attempt to regain possession very quickly. Unfortunately, for Barcelona, the man that was chosen to succeed Pep, Tito Vilanova, at the end of the 2011/12 season seemed to have a different view of how his team should defend.

The following plot shows Barcelona’s rolling 6 game average Pressing Score (PPDA) over the last 4 seasons, starting with 2010/11):

Barca

During the two seasons that Pep was in control at the Camp Nou, Barca barely wavered from the range of 6 – 9. However, as soon as Vilanova took the reins we can see that the team’s PPDA value started to rise. In other words, Barcelona attempted fewer defensive actions per pass they allowed beyond the x = 40 line than they had previously done. They had abandoned the way that they played when they weren’t in possession of the ball.

Sadly for Vilanova, we are all aware of what happened next. After 17 games of the 2012 / 13 season he was admitted to hospital where he underwent cancer treatment. Officially, the Barcelona team was under Tito’s management for the entire season, however, it seems pretty clear to me that as soon as Vilanova was admitted to hospital Barcelona began to revert to the game plan that had held them in such good stead under Guardiola.

The rate at which Barcelona reduced their PPDA metric in early 2013 (Game number 91 onwards in the above plot) is nothing short of remarkable. The Catalan giants performed a huge about turn in how they approached the game when they weren’t in possession of the ball. Either Vilanova had a tactical rethink during his period of convalescence or it seems that Jordi Roura was actually in charge of the team’s tactics during this period. Given the stark differences in the PPDA values, I know which of those two options my money is on.

The above Barcelona plot is very useful as it shows the impact that a manager can have on the intensity that a team presses with, regardless of the players at their disposal. And I believe it also provides an answer as to why anyone would wish to measure this at all.

We know that attacking strategies can be measured / defined with reference to stats, is there a reason that we can’t strive to do the same for the defensive side of the game?

 

Pressing Intensity by League

We have seen in the two examples cited in this piece that whether teams press or not seems to be a decision of the manager or coaching staff. It appears that the right group of players is capable of playing whatever way is asked of them. It follows therefore that the position of the leagues in the next table is attributable to the tactics that are pervasive throughout the respective leagues:

Leagues

Out of the 20 leagues / seasons that we have the data for, the 2013/14 EPL season has got the highest PPDA value, ie it seen the least amount of pressing. It seems that Mauricio Pochettino is still very much the exception, and not the norm in the Premier League.

We can also see that the three leagues / seasons that witnessed the highest levels of intense pressing were all during the 2011/12 season.

Is this simply some variance associated with small sample sizes or is it a coincidence that this was the last season that Guardialo was in charge of Barcelona and tika taka was all the rage? We can’t be certain either way, but it is interesting that all of the 2013/14 season PPDA values are relatively high – this suggests that pressing as a defensive tactic is not as dominating as it has been in the recent past.

 

2013/14 Premier League

If we turn our attention to last season’s Premier League, we can see the spread of PPDA values across the twenty teams. It’s no surprise to see that Southampton were the team that recorded the strongest pressing value in England’s top flight last season, however the values posted by teams on the far right side of the chart below suggest almost a total lack of pressuring the opposition in deep positions.

EPL2013_14

Those teams on the right of the chart seemed to be content with basically conceding possession in the opposition half and presumably funnelling men back behind the ball to defend much closer to their own goal.

In fact, of the six teams that recorded the highest PPDA values (and thus the weakest press) in the Big 5 leagues last season, five of them played in the Premier League. Lorient in France’s Ligue 1 makes up the passive half dozen.

For all the influence of foreign coaches and exposure to continental tactics it is pretty clear that there are a large number of Premier League managers and coaches that are still very partial to the traditional method of English defending – getting men behind the ball.

 

More than One Way to Defend

This article concentrated on creating a method to define and measure the intensity that teams employed when attemping to win the ball back in high up-field areas. I am very aware that the use of the high press is not the only way to defend, and over the coming weeks I’ll have a look to see if it is possible to measure and define other defensive systems being used by coaches and managers.  In the meantime, fire away with any suggestions or other ideas you might have for how we can create some more defensive metrics.  You can either leave a comment here, on my twitter or drop me a mail.

 PoweredbyOpta

 

Appendix: Logic for settling on x = 40 as boundary for pressing

I ultimately decided on x =40 as the boundary for this pressing metric as I wasn’t comfortable with using just the actions that took place inside the opposition half (x=50). A team may be happy to concede possession in the opposition half but tenaciously apply pressure on the ball when they cross the half way line. The use of x = 40 ensures that such a team will receive credit for the pressure they apply as soon as the ball enters into their own half.

I didn’t want to go closer to the defending team’s goal than x = 40 because x =33 is right back into the defensive third of the pitch. We would expect all teams to make tackles and challenges in that area of the pitch as part, and so it seemed wrong to count those defensive actions in any potential metric that is intended to be used for identifying how defences press.

As a final check on which of the boundaries to use (x =33, 40 or 50) I actually collated and analysed the data for all three potential boundaries. Obviously some teams will differ in the intensity of their pressing depending on which of our potential cut offs we use, however, it was encouraging to see that there was actually quite a high degree of correlation between all the potential cut off values.

Correlation (r^2) values between boundaries on an individual match by match basis:

x= 33 with x= 40: r^2 = 0.84
x= 40 with x=50: R^2 = 0.81

The pressing metric value (PPDA) at x= 40 explains more than 90% of the PPDA at each of the other potential boundaries. So as well as making practical football sense to use x=40 as the boundary for this metric it is clearly a good approximation of how teams press regardless of the exact cut off value that is used.

  • http://bullenstats.blogspot.co.at/ Just ideas

    Good work. Interestingly, I am also going to collect data about tackles, interceptions and ball recoveries resulting from high pressing in the following season of RB Salzburg, and I also chose to draw the boundary line at 40m (exactly 42m because the official pitch length is 105m so that the line at 42m marks the boundary between the second and the third fifth of the pitch; 105*2/5=42. But it’s only cosmetical). Even my explanations are the same as yours.

    In terms of defensive metrics, I have some ideas, but neither of them is really convincing. Defence is really hard so that I am hoping to read something intelligent from another analyst.
    I’ve noted some suggestions here as a comment at the bottom: http://everyteamneedsaron.com/2014/07/24/theories-on-defensive-stats/ In short, I suggest to look at what happens “at the ball”. How often does a player/team press onto the opponent in possession? A main point is to build a counterpart to interceptions: There is one for tackles won (i.e. tackles lost, dribbled past), but there is no metric yet that tells us how often the ball is played past a defender. Only passes that could have been intercepted if the defender had had a better positioning.

    Another kind of “at the ball”-metric would be passes hurried: http://possessionwithpurpose.com/2014/05/27/hurried-passes-could-this-be-a-new-statistic-in-soccer/)

    Though, these metrics don’t include anyway off the ball-movement and positioning, and to be honest, I don’t have a clue how to measure it. Well, you could separate each situation and check if one team is in majority, but what exactly is a situation? And do the defenders maximise compactness? And is there a player who has done a very good job at building the majority? And if there is no majority, who is missing? Far too many hard-to-answer questions.

    As we can’t evaluate if a player/team defends well in open-play, my best of many not ideal approaches is to analyse what went wrong at shots/goals conceded. On team basis this is nothing new: Everyone knows stats such as “shots conceded by set pieces” etc.

    But I am not sure if shot analysis is also common on player basis. I know that Opta collects “errors leading to shot/goal”, in comparison to what I’ve done on Salzburg the season before (the GERMAN article can be found here: https://www.blogger.com/blogger.g?blogID=5946961677838620814#editor/target=post;postID=2961352319206598850;onPublishedMenu=posts;onClosedMenu=posts;postNum=4;src=link), this is very superficial. I regarded each goal conceded and decided what players contributed negatively to the goal. I distinguished between three levels of gravity: A player got 1 error point for an error that ” can happen”. The danger could have been avoided even with the error. Those situations occur quite often without leading to a shot. An example is a lost 50:50-duel in midfield.
    A 2-point error is one that “should not happen”. It also occurs quite often, but it often results in a good chance. E.g., a fullback (and every other player) gets 2 points if a winger dribbles past him before creating the goal. Or imagine a situation where a defender doesn’t track a forward so that the latter can receive a through ball.
    3 points are only awarded for really bad mistakes, such as clearing the ball straightly to an opponent who can shoot immediately.

    However, this method is not ideal because…
    -it’s subjective
    – a lot of effort and time is necessary for studying the shot’s creation carefully. One goal took me few minutes. This is worrying because I only regarded one team, not the whole competition, and because…-
    I only regarded goals, not shots. As always, goals alone can be misleading so that analysing shots (or at least big chances) would be indeed a step forward. And a step that would take even more time…

    As there are issues with time an probably also data, I of course don’t suppose you to put any of my suggestions into practice, but there are just ideas…

  • http://blogs.columbian.com/portland-timbers/ Chris Gluck

    Just ideas – thanx for the plug on my article about Hurried Passes :).

    The issue I confronted was when evaluating passes – the most frequently used method for moving the ball on the pitch I needed to ignore tackles, fouls or other events associated with dribbling the ball. I believe there needs to be a distinction between dribbling and passing… two very different things.

    The other issue is how do you measure what doesn’t happen? For example. many teams play zonal defense and the ‘lack of a dribble drive to an area’ or the ‘lack of a pass to an area’ has just as much value for the defending team (who does their job) as an unsuccessful pass or tackle/challenge that might win the ball.

    And how do you account for the tactical strategy that a team might employ where they CEDE penetration into their own Defending Final Third (up high and opposed to down low) in order to try and capitalise on gaining a turnover – getting a quick transition – to penetrate and create their own goal scoring opportunity while the defender is out of position…

    I don’t have the solution on those issues but I do have some ideas and my analyses focuses on those ideas.

    All told it’s a hard nut to crack and the more research conducted the better the story can be told in 5-10-15 years time… I applaud Colin and his work and only wish I could get access to the OPTA data like he does 🙂 Best, Chris

  • Colm

    Great work again Colin.

    Just wondering – does this take into account a phase where a team is forced into a clearance, unsuccessful pass or loss of possession without a tackle, interception, challenge or foul taking place?

    Also, I’d be curious how time could be used to supplement this – taking the supposed Barca rule of winning the ball back within 5 seconds as the upper benchmark?

    • Colin Trainor

      Colm,

      The formula simply operates as follows:

      Get the total passes played by the opposition more than 40 Opta co-ordinates away from our goal (a).

      Sum the total number of tackles, fouls, interceptions and challenges that we put in during the game in the same area (b)

      PPDA = a / b

      Not really sure that answers your first question, but I’m not looking at phases or passages of play per se Colm. Does that make sense?

      Yes, could easily (ish) add in some sort of time to regain possession. The reason I didn’t do it in this article is to do that you need to construct the passages of play. For this article I could simply work off a list of individual actions -all I needed was the x/y co-ordinates.

      Does that answer your question?

      BTW, feel free to drop me a mail to gutties@hotmail.com if you wish to chat anything over any time.

  • MattV

    Hi, I just lost my attempted post because by the time I’d read this article and formulated my comment, the captcha timeout expired, or somesuch. Then, going back erased the text… Very annoying, please fix 🙂

    In any case, the point of my previous (logically bulletproof and beautifully written!) comment attempt, was that I would also like to see figures for x=50 and x=60 in addition to x=40, in order to reveal the level to which patricular teams carry their press into the opposition half. The figures for x=40 do seem a sensible first stop, though.

  • Pingback: Nekromanták vs fekete lovak | Vörös Fonat()

  • Pingback: Foot - L1 - OM - Marseille: le pressing de Bielsa décortiqué آریاییان()

  • Pingback: Átalakul a Barca, és vele Messi is | NewsCafe Hungaria()

  • Pingback: Átalakul a Barca, és vele Messi is | Magyar sport()

  • Pingback: So how do you beat Manchester City? | Abel Lorincz()

  • Pingback: Quantifying ‘Gegenpressing’ | 11tegen11()

  • Pingback: Tour statistique de Ligue 1 : 3ème journée | Côté Stats()

  • Pingback: OM : le retour du pressing ? | Chroniques sporTVistes()

  • Pingback: Weights | irregularstats()

  • Pingback: Statistikanalyse: Wie intensiv und erfolgreich pressen Österreichs Bundesligisten? » abseits.at()

  • Pingback: Gary Neville’s Red Wedding – Deep xG()

  • Pieretto

    Interesting.

  • Pingback: Seven games in … so far so good – Andy Johnson's Magic Hat()