Shot Matrix III: The Incredible Through-Ball

In last Sunday's match, Newcastle were only dangerous for maybe a ten- or fifteen-minute stretch in the first half, but during that time they mostly sought to play one kind of attacking strategy. In transition, their striker or their wingers would make runs looking to get behind the back line for a through-ball. They nearly broke Spurs down a couple times, and when Loic Remy's goal came, it was the culmination of an effective strategy. While Newcastle managed only three shots on target from inside the box, and only one from the Danger Zone, they had an expected goals number of about one. These were high-quality, high-expectation shots that Newcastle were attempting.

It's not surprising that through-balls would create high-quality chances. Opta defines a through-ball as "a pass splitting the defence for a team-mate to run on to." So a shot off a through-ball should be shot under very little defensive pressure. The defenders have, by definition, been split and the attacking player has run beyond them. One of the big problems with any expected goals metric based on shot location is that, obviously, a free header is easier to score than a contested header. Same with any other kind of shot. But until the FA springs for the equivalent of the NBA's (incredible) SportVU system, we do not have good data on defensive pressure. The through-ball data gives us, I think, at least a possible proxy for the effect of defensive pressure on shot outcomes.

I have a total of 1738 shots off through-balls logged in my database. A negligible number come from either outside the box or from Zone 1. No professional keeper would stand rooted to his line as a pass split his defense which allowed an attacker to run onto the ball in the center of the six-yard box. I presume that the 12 shots off through-balls from Zone 1 in my database represent shots where the attacking player collected the ball further from goal, rounded the keeper, and then finding himself in front of goal, took the shot. (Of these 12 shots, 9 were goals and two were saved, one was missed.) Likewise, few shots off through-balls were attempted from Zone 2. The vast majority of shots from through-balls are played in Zones 3, 4 and 5.

Through-balls produce shots between twice and three times as likely to result in goals compared to normal shots taken from those areas of the pitch. Here's the data:

As you can see, wide areas which usually produce relatively low-expectation shots are still quite good places to shoot from if you're assisted by a through-ball. One example of such a high-quality wide shot off a through-ball would be Jermain Defoe's attempt against Chelsea in the waning minutes of that match, as Chelsea defended desperately with 10 men. Mousa Dembele threaded a ball to Defoe in a wide area of the box, and his resulting shot wasn't good enough to beat Petr Cech. It was probably the best chance Spurs produced in that closing flurry. It's not a shot that Defoe should score every time, but it was a much better chance than the typical shot from wide areas in the box. I remember that feeling of hoping/expecting Defoe to finish that I don't usually have on shots from wide. Because he'd received a through-ball, he had space to prepare his shot and relatively little defensive pressure. That makes for a better chance.

One thing I find really fascinating are those G/SoT numbers for regular shots and shots off through-balls from Zone 3. Shots off through-balls are far more likely to result in goals than the normal shot from Zone 3, but it's entirely because it's easier to put the ball on target when you don't have the usual level of defensive pressure. A normal shot from Zone 3, under usually higher defensive pressure, stands just as good a chance of hitting the back of the net if you can put the ball on target. What this suggests to me is that we can use SoT% for shots as a proxy for defensive pressure. If a player put a ball on target, it's likely he wasn't being tightly closed down or shooting into a crowd of defenders. That's not a strong proxy, as obviously many shots on target come under defensive pressure or through traffic, but on average, it's at least interesting.

Now, this doesn't apply nearly so much in wide areas. As i noted in my last piece, the lower expectation of shots from wide areas in the box comes not from the difficulty of putting a shot on target, but from the difficulty of beating the keeper. The keeper has a smaller area to cover on wide shots because with proper positioning he can cut off the angle. So with through-balls to wide areas, the player sees an increase both in the rate of shots on target and the percentage of shots on target that score goals.

Team Tendencies to Through-Balls

I find a significant team effect in the number of shots off through-balls played. This is hardly surprising, we would expect clubs to have distinct tactics, especially to have distinct tactics under distinct managers. And playing through-balls is a tactic. As you'll see, two clubs stand out like crazy, and they're both clubs known for their managerial stability and tactical peculiarity. These are shots from Zones 3-5, shots off through-balls from Zones 3-5, and the percentage rate for all clubs in my database with 750 or more total shots from these areas in the Premier League from 2009 to the present.

Most of these clubs are bunched together, but I think Arsenal and Stoke City stand out impressively. Stoke City were managed by Tony Pulis for nearly the entire time covered by the Shot Matrix database, and of course Arsene Wenger has been in charge at the Emirates for like forever. Stoke City are known just as much more the "classic English" long-ball game they play (or used to play) as Arsenal are for the "continental" fluid passing style they practice. And Arsenal take shots off through-balls more than three times as often as Stoke.

As I was building this database, as I mentioned in the first piece, I hypothesized that Arsenal would have peculiarly high rates of expected goals on their shots. That is, I expected to find that Arsenal take, on average, better shots than other Premier League teams. I'll write more on overall shot quality next time, but here's the result of my mini-study of shot quality and Arsenal.

The average shot taken in the English Premier League from 2009-2013 has an expected goals value on 8.7%. Arsenal's shots have an expected goals value of 10.1%. Over a sample of 2693 shots, that is a difference highly unlikely to occur by chance. To put a number of it, there is less than a 1% chance of a club averaging a shot quality that far above league average, on that many shots, by pure random variation. It's a real tactical effect.

Coming up next: (IV) Statistical profiles of Premier League clubs by shot type and (V) Unleashing the nerdery, statistical analysis of the persistence of shot quality by team-season.