clock menu more-arrow no yes

Filed under:

Barcelona vs. Advanced Statistics: The effects of tiki-taka

New, comments

We try to account for everything in the advanced statistics. But we can't account for truly peculiar tactics like Barcelona's tiki-taka and the consistently high-quality shots they create. I think that's cool.

David Ramos

Everyone knows how Barcelona attack. They play the ball out from the back with lots of quick, short passing, they slowly build attacks through the middle, depending on creative midfielders to pick out the killer pass and their false nine to make the runs that produce these opportunities. The tiki-taka style is so endlessly debated, its descriptions by now so clichéd, that I think we lose track of how peculiar it is. In this column, I will take a look at the effects of tiki-taka on goals scored compared to expectations.

Barcelona, peculiarly among all clubs in the major European leagues, consistently score dramatically more goals than their underlying stats would project. This is an entirely plausible effect of a system designed precisely to create high-quality chances. Barcelona have long been criticized for prioritizing the extra pass for the perfect chance. They lack, one might say, that old English directness. They don't just have a go when it opens up. The effects of this are seen in Barcelona's low rate of shots assisted by crosses, high rates of shots assisted by through-balls and shots attempted from the danger zone. But unsurprisingly, the stats we have cannot fully account for the effects of tiki-taka.

Barcelona, peculiarly among all clubs in the major European leagues, score dramatically more goals than their underlying stats project.

I have created advanced statistics pages for the Bundesliga, Serie A and La Liga. The stats break down the location and type of shots attempted and goals scored for each club in the league. I've also included an expected goals column, built from data on the average value of shots attempted, and the links on that page (also linked above) describe the xG method. The heart of this method is the breakdown of the pitch by locations, and the general shot zone map I use is posted below on the right. The xG method also accounts for the type of pass which assists the shot, whether the shot is taken as a free kick or from open play, and whether the shot is taken with the head or the foot. But the locations probably drive the majority of the model's effects.

Shot_matrix_mediumI don't think that any Expected Goals number is the end of the discussion, and I've included lots of the component data precisely so that readers can make their own judgments. I hope xG can be the beginning of a good discussion. In this article, I'll be using the limits of expected goals to demonstrate something about a football club. xG can be useful, I think, even when it doesn't capture something important—its failure can also tell us something interesting.

Perhaps the most striking thing about Barcelona's clinical striking is that it's a consistent aspect of their game all over the pitch. Whether shot created comes from directly on top of the goal mouth or 25 yards out, they are more likely to convert than anyone else in La Liga. There is no region of the pitch where Barcelona don't score more goals than would be expected. The following chart displays goals and expected goals from the basic six zones of the pitch that I use in expected goals. (Zone 6 is a combination of Zones 6-8 in the graphic.). Because so many more goals are scored from Zone 3 than anywhere else on the pitch, I have separated out the Zone 3 chart with a separate scale, so they are all readable together.


One of the things that tipped me off to checking on Barcelona's xG numbers was their ridiculous shot conversion from the danger zone (zones 1-3 in the above map) during the title run last season. In 2012-2013, Barcelona attempted 205 shots from the danger zone and scored 73. The average conversion of danger zone shots in La Liga that season was about 20 percent, and Barcelona nearly doubled that, converting 36 percent of their shots from the central area of the 18-yard box.

So I was expecting that the effects of tiki-taka would be seen most strongly from inside the box, where Barcelona work the ball around carefully, probing for that top-quality chance. However, it turns out that Barcelona's overperformance of expected goals is even more notable from outside the box than inside it. While Barcelona consistently score about 30% more goals than expected inside the box, they score almost double their expected goals from outside. My numbers show that Barcelona should have scored about 36 goals from outside the box based on their average shot quality, but they have scored 68.

Is it just about Messi?

The obvious question at this point, I think, is Messi. Is this some complex, overdetermined effect of tactics, personnel and philosophy, or is this just what happens when you put Lionel Messi on the pitch?

To test this hypothesis, I looked at Barcelona's numbers from the past two seasons when Messi was injured, compared to their numbers over the remainder of the season. If Messi is causing the xG numbers to go haywire by himself, then in the matches he misses, they should go back to normal.

With Messi out, Barcelona averaged about 1.75 expected goals scored and 2.45 actual goals scored per match. With Messi in the lineup, they averaged about 1.85 expected goals scored and 2.6 actual goals scored. That's pretty similar rate of overperformance. So the effect of Messi, in these numbers, is mostly that he is instrumental in creating high quality chances rather than causing them to be dispatched at a much higher rate.

Lionel Messi's shooting skill and his ability to drag defenses out of shape helps Barcelona to become peculiarly dangerous from outside the box.

There is one exception here. While Barcelona's G/xG rates from the danger zone are basically the same with or without Messi, the outside of the box shooting looks like it might be a Messi effect. Barcelona scored 89 goals from the danger zone compared to 67 expected with Messi, about 30 percent better than expected. Without Messi, they scored 26 goals from the danger zone compared to 20 percent, again about 30 percent better. But from outside the box, Barcelona scored just two goals compared to 2.5 expected without Messi. When they had Lionel Messi in the lineup, they scored 17 goals from outside the box compared to 8.5 expected.

The sample here is small, and there are any number of possible confounding factors. But this data suggests that Messi adds an important dimension to Barcelona's attack. Lionel Messi's shooting skill and his ability to drag defenses out of shape helps Barcelona to become peculiarly dangerous from outside the box as well as from inside of it.

The Statistical Effects of Tiki-Taka

The effect of tiki-taka, then, is indeed to create high quality shots from inside the box. Barcelona does this by working chances through the middle, eschewing crosses and attempting many through-balls. I have mentioned this before, but just to confirm, here's the data. This is total shots, shots from the danger zone, shots assisted by crosses and shots assisted by throughballs, both for Barcelona and for the rest of La Liga.

Tactical Stats Shots DZ S S Cr S Z345 S TB DZ% Cr% TB%
Barcelona 2503 1028 421 1396 223 41.1% 41.0% 16.0%
La Liga 42796 15358 7568 20971 2045 35.9% 49.3% 9.8%

So Barcelona attempt a higher percentage of their shots from the danger zone than other clubs in La Liga, but they produce significantly fewer of these shots with crosses. This is not a terribly surprising finding, but it's worth noting. I think that the effects of tiki-taka seen in this table are nicely congruent with the data on expected goals from above. Barcelona seek high-quality shots by passing through middle. They attempt 75 percent more shots assisted by through-balls than the average club in La Liga. It makes sense that this constant drive for higher-quality chances would not show up only in shot locations and shot types. They also just generally attempt better shots in a way that doesn't wash out perfectly in the shot matrix data.

Data Tables

Here's the basic data I've been using for goals and expected goals, season to season. The xG+ columns show how much Barcelona outperformed expected goals, scaled to 100. So 132 in DZ G+ in 2011-12 means that Barcelona scored 32 percent more goals than expected off their danger zone shots that season.

Season G1 G2 G3 G4 G5 G6 xG1 xG2 xG3 xG4 xG5 xG6 xG+ DZ G+ WG+ oB+
2009-10 15 5 45 9 1 17 12.7 4.2 38.5 9.3 1.5 7.8 124 117 92 217
2010-11 11 5 43 12 3 15 11.5 5.8 35.5 9.6 3.1 7.3 122 112 118 206
2011-12 11 12 50 13 4 12 8.0 6.1 41.2 9.1 3.8 7.0 136 132 131 172
2012-13 14 1 58 14 4 7 11.1 1.6 38.1 8.4 1.0 6.2 148 144 192 112
2013-14 9 4 29 7 1 12 7.9 2.6 26.3 6.3 0.9 4.8 127 114 111 252