On the analysis of defensive corner kick strategies in football

When Canada lost to Panama two-nil in a September World Cup qualifier, many fans and journalists bemoaned how the first goal went in. It came off a Panama corner kick as Rolando Blackburn headed the cross in just inside the far post. People could not understand why Canada did not station a man to mark the far post – in that situation it would surely have saved the goal.

There is, of course, a massive problem with this line of reasoning. It is the king of the small sample size error in logic. That was literally one event to help support the idea that you should have a man on the far post. There is a reason why such a point gets brought up, however, and that’s because it is a very salient event. It is obvious what could have prevented a goal on that corner – a man on the far post. Not only that, but putting a man on the far post is common place, so people are quick to criticize when going against tradition so obviously leads to the undesired event.

The question then becomes, why not have a man on the far post? Perhaps the answer is, why stop there? Why not have a man on both posts, and one in the middle of the net who can go across and guard the net behind the keeper? Say the opposition sends five guys forward to attack a corner, why not man mark everyone the opposition sends forward then use the other five to guard the line?

Clearly the final scenario is ridiculous and no one does it. Why they don’t do it gives the possible answer as to why not have a man on the far post. When you think about it, guarding the posts is pretty arbitrary. The idea is surely that those are the areas where a keeper is least likely to get to so you try and protect them, but do you really need a guy on each post each time? What if the other team has a guy unmarked and you have both posts marked? What’s more dangerous, leaving him open or leaving the post free?

I play keeper, albeit at a pretty recreational level, but I have a set up I always use for corners. I put a man on each post and a man in the space ahead of the near post, about five metres closer to the corner and three metres out from the line. If they have an open man I will tell the far post marker to take him, and if there’s still another one I’ll peel off the near post man. I never move the zonal marker beyond the near post. Is it successful? I have no idea, I don’t have the data. The reason I do it is from facing hundreds of corner kicks and realizing what gives me the most trouble, how most of the deliveries come in, and how chances are created and goals are scored from corners. But I’m just a rec league player, I’m not expected to have the data. You would think that corner kicks, one of the few discrete events in football, would be rife with data and analysis on defensive corner kick strategies.

Suprisingly, there isn’t that much out there when it comes to meaningful analysis of raw data, possibly because the data isn’t really that great. In 2005 there was a study published in Science and Football V titled ‘Notational analysis of corner kicks in English Premier League soccer.’ The researchers collected and defined their own data of 217 corner kicks over the course of a Premier League season. It is a fairly short paper and probably the most key finding was that the type of ball that had the most success was an in-swinger across the front of the six yard box.

This helps explain why many teams line up four or so defenders marking zonally across the six yard box (in fact it’s what Canada did on the goal conceded to Panama). Does it mean that it’s smart? Well, the intuition seems sound: if most goals scored in area x then defend area x. That’s not enough though, as researchers understand logic yields to practice. We need real world data to show that such a system is better, not just the inductions. Of course, there are a litany of research method issues to take into consideration, the biggest one I think being a type of selection bias that doesn’t allow us for a proper control group. We could do a between groups comparison between teams that mark zonally and teams that mark man to man. Controlling for number of corners faced would be pretty straight forward, and as long as we operationalize what zonal marking on a corner is (we could even have multiple groups) then what set up goes in what group should be easy. We just need the grunt work done. But what teams practice zonal marking on corners? Is it possible it is the teams that have players with less aerial ability than other teams are more inclined to such a system? Do managers view it as something that is advantageous with a short side, and had they a bigger team they’d do strict man to man marking?

Imagine this scenario. In a twenty team league ten teams mark zonally and the other ten mark man to man. At the end of the season we find that these two groups (zonal vs. marking) concede corners at the exact same rate. We can conclude that the type of marking system has zero effect, right? Well, what if the ten teams that mark zonally do so because they are shorter and think it gives them an advantage? If the ten teams that mark zonally are also the ten shortest teams, would it be fair to say that zonal marking is a better strategy than man marking? You could imagine a bunch of scenarios where the system and aerial ability interaction could present a headache for statistical analysis. That’s without even mentioning how leaving players up the pitch effects the game. Do you leave one guy up to corral a clearance? Does that save you more goals in the long run? What about two guys? How much does it depend on how the other team sets up? There are too many variables to consider, just like all statistics, so we’ll never get the perfect model, but we can at least start controlling for some.

The point here is that analysis of defensive corner strategies as practiced by fans and (often) journalists is naïve and lazy. We rely on salient events that conform to our traditional views of defensive strategies – not at all unlike why humans make errors in reasoning in every walk of life. Sadly, even controlling for obvious factors like number of corner kicks faced seems to not be part of much of the popular discussion, let alone other possible variables. The tape is there, it is up to clubs and researchers to analyse, define, and interpret the data in a meaningful way for us to have any sort of definitive answer on what system is the best for each team. But would that time, money, and effort even be worth the advantage gained? Would a team save enough goals to justify putting those resources into that aspect of research? That’s the other big question for the analysis of corner kicks.

UPDATE SEPTEMBER 12th, 2013: In this article I didn’t mention the idea that attacking teams would devise certain types of corner kick strategies depending on the system the defending team is using. Certainly if I was managing I would give my team different instructions depending on the defending teams set up, and if teams did do this it would make for a muddier picture of the stats involved in analyzing success rates of defensive strategies and corner kick behaviour. This could potentially be another fruitful avenue for research on the subject.