Last week we looked at some basic team shot statistics (total shot ratio, shots on target ratio, and conversion rate) with an eye to how they measure performance, and alert us to anomalies unlikely to last. Now we’re going to do the same thing with more shot-related numbers, but with a twist: these numbers show something different from what you’d expect.
[The numbers in this piece come from the really wonderful site Objective Football, run by Ben Pugsley, who gathers a ridiculous amount of team data and makes it available for public consumption. Note that these numbers do not include the games just played midweek.]
Because the only shots that go in are shots already on target, measurements dealing with shots on target should give us important performance data. The two primary measures are shooting percentage (Sh%, sometimes called scoring percentage), which is the percentage of your shots on target that go in, and save percentage (Sv%), which is the percentage of opponents’ shots on target that your keeper saves.
Shooting percentage measures how good your shooters are. If you put it on target and the keeper doesn’t save it, that means you placed the ball well or shot particularly hard, or both. Let’s first take a look at Sh% for the season so far (league average is 33.3%):
West Brom 43.5
West Ham 37.7
Man City 35.5
Crystal Palace 35.3
Man United 25.2
Last week we saw that Southampton’s conversion rate was abnormally low, and here we see it’s partly because their shooting percentage is dreadful. In fact, their 21.4 would be the fourth-lowest recorded in the eight-year period for which these stats are available. We also see one of the things holding Manchester United back, and to a lesser degree Everton too.
At the other end, strange bedfellows West Brom and Arsenal, Pulis and Wenger, are well ahead of the pack. In fact, their numbers would be the two highest ever recorded in the eight-year period.
There are some oddities here. Crystal Palace are high on the list, which means not only (remember last week) is their SoTR noticeably higher than their place in the table, their Sh% suggests that they’re scoring lots more goals with those shots on target. And indeed, Palace have scored 30 goals, tied for ninth in the league. Their problem is at the other end.
West Ham are also a surprise. We saw that their SoTR roughly fits their place in the table, but their Sh% is near the top. They’ve scored 29 goals, only one behind Palace, and like Palace, defence is their problem.
Let’s go now to save percentage. That’s clearly a measure of your keeper’s shot-stopping ability. If your opponent’s shot is on target, and your keeper saves, he’s doing his primary job. Here are the listings for Sv% (league
West Brom 73.6
Man United 69.1
West Ham 67.6
Crystal Palace 62.7
Man City 57.6
None of the top numbers are abnormally high, but the bottom three are quite low, and would be respectively the second, fourth, and sixth lowest recorded in the eight-year period.
Most of these figures are unsurprising. They tally roughly with how we rate the shot-stoppers. Hugo Lloris leads the pack, Tom Heaton and Thibaut Courtois follow, Ben Foster is having a great year, Petr Cech and Victor Valdés are quality keepers. At the other end, Claudio Bravo hits the headlines for all the wrong reasons.
At the same time, it’s a bit unexpected to see Fraser Forster and Lukas Fabianski helping prop up the table; they’re quality keepers too. Plus, the teams’ defensive records are so very different: Southampton have one of the best defences in the league, Swansea one of the worst. But Swansea have allowed 46 more shots on target, which explains the difference. (We’ll look closer at the Forster/Fabianski comparison in our xG pieces; for the moment, just remember they’re at or near the bottom.)
All of this seems pretty straightforward. But it isn’t. Because these numbers, although 100% accurate, don’t quite tell us what they appear to. They seem to say that Arsenal and West Brom are shooting brilliantly, and that Hugo Lloris is right now the best keeper in the league. But as hard as it may be to believe, Sh% and Sv% are more luck than skill. I’ll repeat that: the percentage of your shots on target that go in, and the percentage of opponents’ shots on target that your keeper saves, depend more on luck than skill. The percentage is roughly 60-40 for Sh%, and 65-35 for Sv%. This has been proven repeatedly.
But how can this be? Surely some players and teams are better shooters than others, and we know that some keepers are better. There are three explanations. The first is that roughly 35-40% is indeed skill, so it’s not entirely absent from the ratings. The second is the easily verifiable fact that putting the football exactly where you want, at the proper pace, is hard. And the third is the obvious but rarely noted fact that the goal is big, much bigger than the area a keeper can cover.
Think about it: even the best strikers miss hatfuls of chances. We’ve already seen how wildly conversion rates can swing. And if the best shooters struggle, what must ordinary shooters do? Only a small percentage of footballers have a clear idea where and how fast the football is going when they strike it. Add to that the size of the goal, and any given shot, depending on power and placement, can be completely unstoppable or savable by a six-year-old. And so, at both team and individual level, there’s a huge slice of luck involved.
The situation of the keepers is simply the converse. Of course some keepers are better than others. But to save a shot, the shot has to be savable in the first place. And since your opponents’ shots will be all over the place, a lot of it is luck.
Among other things, what those studies in the link show is that, over time, high and low Sh% and Sv% won’t last long. A look at the above tables provides corroboration. Most of the teams are clustered within a very narrow range, only a few points from the league average. That’s very different from TSR and SoTR, where the numbers show a wide spread.
So there are two lessons here. First, statistics can be deceiving if you don’t know the full story. Common sense isn’t always right. Second, like very high or low conversion rates, high and low Sh% and Sv% can be red flags. If your team, or a particular striker, is doing well or poorly, a look at these numbers tells you if you’re riding your luck or can’t get the breaks.
So if we’re talking attack, West Brom and Arsenal may be headed for a fall, and Chelsea are over-performing a bit as well. For shot-stopping, Southampton and Swansea, both of whom have decent keepers, have good hope for improvement in the weeks to come. Man City changed their keeper and their luck did immediately change.
Shooting percentage and save percentage are sometimes combined into a single number, called PDO. (Don’t worry how the name came about.) It’s a controversial stat among analytics people, but it does provide a catch-all figure that to a degree measures how lucky your side happens to be at the moment. It’s also very easily understood, because the league average is always exactly 100. Here are the current rankings:
West Brom 117.1
West Ham 105.2
Crystal Palace 98.0
Man United 94.3
Man City 93.1
As with their shooting percentage, West Brom’s and Arsenal’s PDOs are abnormally high; they’d be the second- and third-highest recorded in the eight-year period. Chelsea’s would be the sixth-highest. So more likely than not their numbers will decline, and possibly their results with it. As with their save percentage, Southampton’s and Swansea’s are abnormally low, and more likely than not they’ll improve.
Remember, though, PDO isn’t all luck: it’s still roughly 35-40% skill. And some teams, in the way Tony Pulis beats the odds in TSR and SoTR, have managed to record consistently strong PDOs. Sir Alex Ferguson’s sides were usually at or near the top in this ranking. In fact, United’s 2009/10 side had the highest recorded full-season PDO, at 117.7. But they didn’t win the title. Chelsea, under Carlo Ancelotti, edged it with a decent but unexceptional 105.3 PDO. That’s because by every available measure, that Chelsea side was the most statistically dominant team in the period under consideration. They didn’t need much luck.
It should be clear that despite the luck factor, Sh% and Sv% will to a degree depend on the positions you shoot from. In fact, that’s true of any goal- or shot-related measure. Better spots, better shots. It’s been a recurring theme here, and next week we’ll face it squarely. As you probably know by now, the stat is called Expected Goals, or xG, and for the moment it’s the most sophisticated performance stat we can widely access. But it’s a lot less complicated than you might think. Next week, then.