The venue for the first OptaPro German Conference. Westfalenstadion. 12 hours ago

This Thursday sees OptaPro host its first football analytics conference in Germany. Presentations. 21 Mar

.@Worville provides an in-depth analysis of Napoli, assessing various aspects of their playing style. Informed. 20 Mar

BLOG: @Worville applies a possessions model to this data-driven analysis of @sscnapoli. Se…… 17 Mar

BLOG: @Worville uses a possessions model for an in-depth analysis of @sscnapoli. Partenope…… 17 Mar

The OptaPro Blog

RSS Feed

Welcome to the OptaPro blog, featuring news and analysis from OptaPro's cutting-edge research team.

BLOG: La Liga bias

The general consensus across the greater football community is that the Spanish are king when it comes to possession oriented football. When we consider Barcelona and the Spanish national side, this is undisputed. But, do these truly exceptional squads unfairly skew our perception of the Spanish football as a whole? Most non-Spanish fans have a woefully incomplete view of La Liga. Since international fans have considerably more access to Barcelona and Read Madrid in comparison to the other La Liga clubs, have we developed a bias?

It seems that we have.

Using Opta's passing data from the 2011-2012 EPL and La Liga seasons, we have developed a robust passing model that estimates the likelihood of each individual pass being completed successfully. If the Spanish are indeed kings of possession football, we should see Spanish squads attempting high efficiency passes at a much higher rate than their competitors.

This is a comparison of La Liga and the English Premier League by the type of passes that they generally attempt. 


The x-axis measures the likelihood of a pass being completed. For example, a pass in the 0.90 section has, by my estimation, between a 90 and 95 percent chance of being completed. The y-axis is the total number of passes that belong in that particular efficiency bucket.

The results here are contrary to my - and likely most people's - general preconception. The English Premier League actually had more high-efficiency passes attempted in 2011-2012 than La Liga.

But, like most football statistics that roll-up across many different teams, this doesn't tell the whole story.


La Liga



This is a team-by-team breakdown of every 2011-2012 EPL and La Liga club by their "strong pass" to "weak pass" ratio. To clarify, "strong" passes are classified as passes that are estimated to have more than a 50 percent chance of completing. Conversely, "weak" passes are passes that have less than a 50 percent chance of completing. It's important to articulate that classifying a "weak" pass does not by any means suggest that that attempting that pass is detrimental to a team - in fact, practically every cross falls into this designation. Instead, these two contrived descriptors are simply used to give us a better understanding of team style.

A good rule of thumb in football analysis is that any graph that plots Barcelona and Stoke City at opposite ends of the spectrum is probably a good one. And, in this case, it makes a ton of intuitive sense as well.

Curiously, outside of Barcelona and Real Madrid, the rest of the high-efficiency teams are English clubs. In fact, Swansea City's ratio almost perfectly matches that of Real Madrid.

A majority of Spanish clubs reside just beyond this top English tier, suggesting that the league is generally more possession oriented - but not a significantly large amount more than the average English EPL side. On the tail end for the Spanish is the club CA Osasuna who's playing style most closely resembles Aston Villa.

While we can't find any Spanish equivalents to Stoke, we shouldn't be so fast to condemn this particular style of play. While Stoke does attempt a higher percentage of "weak" passes than any other squad in these two top leagues, they also complete these attempts at an alarming rate.



In fact, we can build similar graphs to show the trends of particular players to glean a better understanding of their particular roles in their team's system. Ryan Shawcross is one of Stoke City's defenders. Shawcross, like Stoke as a whole, attempts "weak" passes at an exceptionally high rate. While this isn't surprising, this is not the case for every Stoke player.



Steven N'Zonzi, who started 35 EPL games in Stoke's midfield this season, has an entirely different distribution. N'Zonzi attempts a much higher amount of "strong" passes than the average player in either the EPL or La Liga. Is this what you would have expected from a Stoke player? Given their reputation of employing a very direct style of play? Clearly, it's important not just to look at passing trends on a team-wide scale since there can be such tremendous deviation across players on the same squad.

By wading a bit deeper into the stats behind La Liga's pass-centric mentality and Stoke's long ball infatuation, we gain the appreciation that some of our common talking points are a bit more nuanced that we previously expected.

Posted by Devin Pleuler at 10:37


Big Fat Goalie said...
Would you care to explain exactly how do you categorize passes into difficulty groups? For example, "a pass is likely to be completed with 90% chance", is the 90% estimated by a mixture of direction, length and position of the pass, or just an ex post statistic? Thanks a ton! (Neat graphs BTW)
May 23, 2013 07:02
Devin (author) said...
Exactly correct. The passing model is built using a handful of different metrics including all of the different metrics that you have mentioned (direction, length and position) as well as a handful of others (is it a cross?, etc.). The model not perfect, and we expect to improve it, but it's relatively robust and suitable for this kind of analysis.
May 24, 2013 04:18
tommo said...
Doesn't it seem counter intuitive that all types of passes were completed on average over 20% of the time (EPL average)? Because surely a pass you have decided should be completed <5% of the time should not average out to over 20% over 380 games? Surely a better judgement would be to group passes together and then see the season-long average completion and set that as the difficulty. But then I guess there are still problems, I mean e.g. Shawcross completes a higher % of your "difficult" passes than the average EPL player, but obviously it is far easier to complete a long ball to Jones or Crouch than nearly anyone else in the league.
May 24, 2013 07:15
idimik said...
Spanish teams are best interceptors (Proof - but have worst pass completion rate (Wenger said it here - - backed by stats) in top 5 leagues. But I'm curious whether it's because of skilled interceptors or worse passers on average. Can you make graphic showing pass completion rate for every pass difficulty in Spain vs PL?
May 25, 2013 08:33
Devin (author) said...
We must be careful making a claim like the Spanish being the best interceptors of the ball. It might be true, but the source you have cited actually looks at just the number of the interceptions. Does the league that shoots the most have the best shooters? Of course not -- and it's the same with interceptions. A graph showing pass completion rate vs. pass difficulty would be interesting. I would expect that La Liga teams complete "easy" passes at a higher rate than expected, while completing "hard" passes (e.g. crosses) at a lower rate.
May 28, 2013 06:00
Devin (author) said...
Yes, it's certainly counter intuitive -- and a product of an imperfect model. But, if you look at the first graph, you will notice that these <5% passes are an *incredibly* small sample size. So, I'm not terribly worried. There probably remains some kind of bias that could be worked out here. That bias may indeed be caused by a team's particular playing style. The Shawcross -> Crouch situation certainly skews our results.
May 28, 2013 06:07
Manny said...
Interesting piece. I love the degree to which Barcelona are outliers! Is one of the variables for determining difficulty of a pass, whether the pass was lofted or on the ground? Maybe this could be used to alleviate the imperfection of the model relating to Shawcross' lobbed balls to Crouch. A list of inputs used in the difficulty calculations would be great to see.
June 4, 2013 06:54

Post a comment