Monday, July 14, 2014

Lionel Messi and Soccer Equilibrium Outcomes

So another World Cup has come and gone. Enough water had passed under the bridge that I no longer resented Argentina for their dismal performance in 2002 when I wagered on them. I was vaguely hoping for an Argentine win, just because I would have liked to see Lionel Messi win a cup.

'Twas not to be, of course.

A very good starting point for understanding Messi is this excellent post by Nate Silver going through a whole lot of metrics of soccer success and showing that Messi is not only an outlier, he's such an outlier that his data point is visibly distinct from the rest even in simple plots. Like this one:

(image credit)

Seriously, go read the whole thing. If you're apt to be swayed by hard data, it's a pretty darn convincing case.

So what happened in the World Cup? Why didn't he seem nearly this dominant when you watched him play?

The popular narrative is that there's some inability to perform under pressure - in the big situations when it really counts, he doesn't come through with the goods. He's a choker, in other words.

This is hard to disprove exactly, but one thing that should give you pause is that with Messi on the team, Barcelona has won two FIFA Club World Cups and three UEFA championships. This at least suggests that the choking hypothesis seems more specific to World Cups.

So one explanation consistent with the choking hypothesis is that the World Cup is much higher stakes than the rest, hence the choking is only visible in that setting. It's possible, and hard to rule out.

But another possibility is that the difference comes from the way that opposing teams play against Messi in each setting.

Remember, a player's performance is an equilibrium outcome. It's determined by how skilfully the person plays that day (which everyone thinks about), but also by how many opposing resources are focused on the person (which very few people think about).

Let's take the limiting case, since it's easiest. Suppose I take a team comprised of Lionel Messi and ten guys from a really good high school team, and pit them against a mid-range club team. My guess is that Messi wouldn't perform that well there, and not just because he wouldn't have as many other good people to pass to. Rather, the opposing team is going to devote about 4 defenders just to covering Messi, since it's obvious that this is where the threat is. Throw enough semi-competent defense players on someone, and you can make their performance seem much less impressive.

Have a look at the pictures from the Daily Mail coverage of the game against the Netherlands. In one, Messi is surrounded by four Dutch defenders. In another, he's surrounded by three. The guy is good, but that's a pretty darn big ask of anyone.

In other words, Messi may be better than the rest of the Argentine players by a large enough margin that opposing teams will throw lots of resources into covering him, making it harder for him to shine. In soccer, like in martial arts reality (as opposed to martial arts movies), numbers matter. Jet Li may beat up 12 bad guys at a time, but it you try that in real life, you're on your way to the emergency room or the morgue, almost regardless of your martial arts skill.

The last piece of the puzzle for this hypothesis is the question of why this doesn't happen when Messi plays at Barcelona.

I'm a real newb at soccer (evidenced by me referring to it as 'soccer' - you can take the boy out of Australia, etc.), but my soccer-following friends can tell me if I'm right here or not.

My guess is that the rest of the Barcelona team is much closer to Messi's level of skill than the rest of the Argentine team. This means that if opposing teams try to triple mark Messi in a Barcelona game, the rest of the attackers will be sufficiently unguarded that they'll manage to score and the result will be the same or even worse than if Messi were totally covered. As a result, Messi goes less covered and scores more.

There's a reason that the sabremetricians (who tend to be among the most sophisticated of sports analysers) talk about wins above replacement. You need to think about the counterfactual of if the person wasn't there, not the direct effect of what they did or didn't do in equilibrium.

Of course, the skeptics will point out the cases where great stars did manage to indivdiually play a big role in lifting their national teams to great success. What about Maradona, they say?

This is a fair question. Sometimes you really can get it past five defenders to win a world cup. Maybe that's what a true champion would have done yesterday.

Or maybe the English just weren't marking as well as the Dutch were.

Or maybe, even more pertinent, the rest of the Argentine team in 86 was sufficiently better in relative terms that England couldn't afford to mark Maradona as hard. The effect of this, if true, would be for Maradona's performance to look more spectacular relative to the rest of his team - having a good team means less defenders on you means more heroics. And when that happens, you look individually more brilliant, leading to you getting all the credit and making it look like you won the game single-handedly. If you really were that much better than everybody else, you would be less likely to deliver a performance that showed this fact to a novice observer.

Not many people think in equilibrium terms. This is why we analyse data.

The data case, however, is clear. Viva Messi!

No comments:

Post a Comment