Alex Stewart looks at set pieces, an area of the game where there is still room to gain significant advantage through planning, strategy, and data use
Every fortnight for 20 weeks, we run a course for the PFA exploring the use of data in football. We have a range of guest speakers, and recently one, Oliver Gage, was talking about opposition set-piece reports. He showed this visual, courtesy of 21st Club, to make the point that set-piece success and general, on-pitch dominance are not strongly (if at all) correlated.
Instinctively, this feels right. We all know that some teams may have a James Ward-Prowse who elevates that facet of the game, or that a fairly stolid side might have a few aerial monsters capable of causing issues at corners; neither of this necessarily translate to being generally a dominant side.
But, this graph was from 2015/16. We know that plenty of clubs have been spending money on set-piece coaches. Indeed, set-piece goals do account for a significant proportion (around 25-30% of goals) so why not make the effort, especially if the above viz holds true? But, as we know, money often buys success in football – generally, wealthier clubs are more dominant in open play and can afford the best coaches, so surely the ground has shifted? Let’s have a look.
To conduct this quick analysis, I looked at open-play and set-play xG and xGA from Europe’s big five leagues last season. I used Opta Analyst’s excellent Data Hubs to access the material.
The below visual shows (using median-derived quartiles) which teams are good and bad at open-play and set-piece xG generation. I have highlighted four groups: league champions, teams finishing in Champions League places, relegated sides, and Bochum, who were in the relegation play-off place, but won it (Metz were not so lucky, hence being in red).
And what we can see is that in attacking terms, what was true in 2015/16, albeit measured slightly differently, is indeed still true. There is a weak positive correlation between open-play xG and set-piece xG (Pearson correlation of 0.347). This is probably because the teams who generally dominate open play are able to generate more field position from which to win dangerous set-piece opportunities. One thing that is fairly obvious too is that poor teams are just (mostly) poor. Only Luton Town and FC Koln are in the strong at set-pieces/weak at open-play xG creation quartile, although Bochum is also there, while only Almeria is in the strong in open-play but weak at set-pieces quartile.
That said, there are a number of teams such as Genoa, Heidenheim and Torino who do punch above their weight in SPxG, while teams like Sevilla, Everton, and West Ham also do well by that measure. And while Brentford and Bournemouth are both in the ‘good’ quartile, there is no doubt that their SPxG helped them punch above their financial weight. Generally, too, wealthier teams are good in open-play as is to be expected, but several are poor, or even very poor, in terms of SPxG. It’s hard to know whether the resources put into set-pieces by clubs such as PSG or Stuttgart are commensurate with their overall spending, or indeed match the rough spend of clubs near them in the table, but it is clear that there is room to develop in that department.
Crucially, though, it clearly still holds that there is a weak correlation between open-play xG and set-piece xG generation and so this area is still ripe for exploitation by teams who cannot afford to stack their teams with the kind of talent available to their wealthier competitors.
Let’s look now at defence. Because here we see something altogether different, and very interesting. Running open-play xGA against set-piece xGA, we can see pretty clearly below that there is a moderate to strong correlation (Pearson 0.705, right on the line) between conceding chances in open play and at set-pieces. Again, some of this is explained by poor defensive teams giving up set-piece opportunities by conceding corners, being forced into fouls, and probably defending deeper (note to self, do average height of defensive actions vs xGA from set-pieces some time), but also this is probably because some of the same core elements of defending apply to both skillsets/scenarios – heading, blocking, clearing, organisation, communication – in a way that is more fundamentally alike than the skillsets involved in being good at attacking.
Before you ask, the numbers suggest essentially no correlation at all between being good at generating xG from set-pieces and giving it up (in simple terms the Pearson correlation between xG and xGA from set-pieces is -0.0699). If you rank all these teams by xG for from set-pieces, Liverpool are fifth and Arsenal sixth with 15.2 and 14.9 respectively, but Liverpool conceded 12.5, while Arsenal only conceded 6.6 (meaning that Liverpool were 21st worst, to Arsenal’s sixth best). This means two things – that just having a bunch of players who are good in the air is insufficient to monster both aspects of the set-piece situation, and that having a really good set-piece taker probably adds significantly to one’s xG for (and obviously adds nothing to the defence).
Given, though, that there is a moderate to strong correlation across open- and set-play defence, clubs who are weak only at set-pieces need to consider whether they might be able to work on specific, simple structural things that would help, as they obviously possess the tools to be good otherwise. Two other things to note from the defensive numbers: being good at set-piece defence alone does not give the same kind of table bump that being good at set-piece attack alone can do; and Manchester United and Spurs really were poor defensively last season when compared to the kinds of teams that their spending bracket puts them alongside. Indeed, it’s interesting how the majority of the worst teams here are Premier League teams, and not just ones in the relegation zone.
Lastly, let’s get closer to the original visual, and plot xG difference in open-play vs xG created from set-plays (firstly, this is closer to the original visual and secondly, plotting set-piece xG difference is basically irrelevant given the clear lack of correlation between SPxG and SPxGA if we are trying to show that set-pieces can yield marginal gains for weaker overall sides, which is what we are seeking to do, I think).
So, what do we see? In obvious terms, the good teams (champions or CL sides) are again generally good and the weak teams are generally weak. But, interestingly, there is actually a weak negative correlation between being an all-round good side (high number for open-play xG difference) and being good at creating from set-pieces (Pearson of -0.356). This tells us that weaker open-play sides can absolutely gain an advantage from set-piece attacking that does not mean they need to have throw money at open-play quality but, crucially, that a few probably already are quite deliberately leveraging that (no-one, obviously, is deliberately bad at open-play xGD but some sides, like Everton, Brentford, Mainz, etc, know that they cannot compete with the financial heavy hitters in their leagues and have clearly leveraged set-pieces effectively).
Crucially, as Luton Town and FC Koln and Bochum will tell you, being good at set-piece attack is probably not enough on its own to save a genuinely (relatively) poor all-round team from going down. But it can bring significant benefits to teams who might otherwise struggle in open-play chance creation to make up some of the deficit.
We have seen that there is a weak negative correlation between open-play xG difference and set-piece xG created. This means that teams who are relatively weaker in open-play xG can still generate good xG from set-pieces and, indeed, that some teams have been able to gain an advantage in this area.
And although there is a moderate correlation between open play defence and set-piece defence, some teams who are very strong defensively in open-play, and who therefore tend to be the best teams (and have the most by way of resources), are not necessarily any good at defending set-pieces. When you add that weakness to the possibility of being strong at set-pieces (mostly) in isolation to your open-play attacking ability, it clearly demonstrates a route for smaller (financially), weaker sides to sting the big sides.
Most importantly of all, though, is the financial element to this marginal gain. While player quality obviously affects set-piece ability, and set-piece coaching cannot account for all the xG gained or given up from set-pieces, the really top sides in both set-piece facets like Arsenal, Bayern, Juventus, Manchester City, Everton, or Brentford could find that the +7 or more xGD in set-pieces they gain was cheaper (by hiring a great set-piece coach) than finding a +7 or more xGD in open-play by buying someone. Yes, OK, this is kind of a false equivalence but a) Everton or Brentford cannot compete with those other clubs otherwise and b) set-pieces are clearly still done poorly by teams with loads of money to spend on players and c) you need to buy players anyway to just play, but you may or may not bother to spend lots of time on their set-pieces.
So it does, to a degree, hold up: set-pieces are still an area where small investments can have disproportionate pay-offs and gain an advantage over sides who can spend more on players but neglect set-pieces.
Header image copyright IMAGO/Kieran Cleeves/Sportimage
