• Dear Guest, Please note that adult content is not permitted on this forum. We have had our Google ads disabled at times due to some posts that were found from some time ago. Please do not post adult content and if you see any already on the forum, please report the post so that we can deal with it. Adult content is allowed in the glory hole - you will have to request permission to access it. Thanks, scara

Set Pieces

mudshark said:
Just to throw the cat amongst the pigeons a little bit in this debate, the data here (http://www.footcharts.co.uk/index.cfm?task=corr_corners) are quite interesting, and suggest that, in any given season, more often than not, there is a significant linear correlation between numbers of goals per game and numbers of corners. If you average the data for each club over all seasons, though (presumably what Anderson and Sally did), it more or less vanishes. There were one or two seasons in the range they used where the correlation was only very weak, as well.
Click to expand...

I think that'll be because teams that shoot more get more corners rather than there being a relationship directly between corners and goals.
 
milo said:
I think that'll be because teams that shoot more get more corners rather than there being a relationship directly between corners and goals.
Click to expand...

They are giving goals per game v corners per game, though, and showing a strong correlation in the majority of seasons since 2000-01. I certainly wasn't able to reproduce Anderson & Sally's graph using their data.
 
mudshark said:
They are giving goals per game v corners per game, though, and showing a strong correlation in the majority of seasons since 2000-01. I certainly wasn't able to reproduce Anderson & Sally's graph using their data.
Click to expand...

I know but I think what you are seeing is teams that shoot a lot getting more goals and more corners. There is certainly a direct correlation between shots and corners.
 
milo said:
I know but I think what you are seeing is teams that shoot a lot getting more goals and more corners. There is certainly a direct correlation between shots and corners.
Click to expand...

What's the difference between A&S's graph and mine, then, or why don't they look the same? And where are they getting this data from for teams with 10-20 corners a game?
 
milo said:
I know but I think what you are seeing is teams that shoot a lot getting more goals and more corners. There is certainly a direct correlation between shots and corners.
Click to expand...

Plus, stronger teams spend more time in the opposition half, making it more likely that they get a succession of corners alongside naturally scoring goals by dint of their superiority. Classic omitted variable bias.
 
@mudshark apologies I missed the edit on your post, that's what comes from posting in the pub.

Apart from the outliers, your chart and A&S'S look quite similar to me. The central cluster certainly has a similar shape. I'd be interested in seeing one overlayed on the other.

In terms of sources, A&S used StatDNA, I believe that Football charts get theirs from football-data.co.uk, who get it from the Press Association.
 
milo said:
@mudshark apologies I missed the edit on your post, that's what comes from posting in the pub.

Apart from the outliers, your chart and A&S'S look quite similar to me. The central cluster certainly has a similar shape. I'd be interested in seeing one overlayed on the other.

In terms of sources, A&S used StatDNA, I believe that Football charts get theirs from football-data.co.uk, who get it from the Press Association.
Click to expand...

I think even ignoring the issue of scale on the number of corners axis, if you drew a regression line through their points, it would have a gradient close to zero (which, after all, is their claim). I'm not sure you could draw a convincing zero-gradient line through my points without some pretty selective pruning. I would agree there's bound to be a mutual dependence of goals and corners on shots at goal, though, so it might not be that surprising that the teams featuring in the high numbers of goals and corners region are Chelsea, Liverpool, Arse and Man U. If you took those out the scatter plot would look rather flatter. The point (6.82, 1.23) is Reading.
 
mudshark said:
Just to throw the cat amongst the pigeons a little bit in this debate, the data here (http://www.footcharts.co.uk/index.cfm?task=corr_corners) are quite interesting, and suggest that, in any given season, more often than not, there is a significant linear correlation between numbers of goals per game and numbers of corners. If you average the data for each club over all seasons, though (presumably what Anderson and Sally did), it more or less vanishes. There were one or two seasons in the range they used where the correlation was only very weak, as well.

Edit: I say it more or less vanishes, but actually, I'm not sure it does. I plotted my own averages using the footcharts.co.uk data for 2001-02 to 2010-11, expecting my plot to look like Anderson and Sally's, and this is what came out:
goals-v-corners-0102-1011.jpg

There's 35 teams involved, and Spurs are represented by the point at (5.71, 1.41). Nobody averages anywhere near 20 corners a game, though ....
Click to expand...

The 20 corners a game (or 20 corners in a game) makes it entirely unlikely to me that your guess that A&S averaged the data for each club over all seasons. It would essentially mean there would have to be a club that averaged 20 corners a game over a decade...

I think the more likely analysis is that they actually used individual games as data points. Meaning that they had more extreme statistical outliers with a very low sample size like the 20 corners in a game one. I actually think that's a better statistical analysis than the one you did though I might be mistaken both about that and about what they actually did.
 
braineclipse said:
The 20 corners a game (or 20 corners in a game) makes it entirely unlikely to me that your guess that A&S averaged the data for each club over all seasons. It would essentially mean there would have to be a club that averaged 20 corners a game over a decade...

I think the more likely analysis is that they actually used individual games as data points. Meaning that they had more extreme statistical outliers with a very low sample size like the 20 corners in a game one. I actually think that's a better statistical analysis than the one you did though I might be mistaken both about that and about what they actually did.
Click to expand...

Thing is, it's not like they've got all but one or two points clustered within a more normal range for numbers of corners. They appear to have a smooth distribution from 0 to 20 corners per game (including points at both those extremes, the latter of which apparently also averaged zero goals). I am forced to admit that I just don't understand what I am looking at when I look at their graph. More information required, I think. Perhaps they used some clever treatment that they haven't explained because they thought it was outside the scope, but I would much rather they had explained it, because the fact that the graph doesn't make sense to me can only lead me to view their conclusion with more scepticism than I did when I first read the book.
 
For anyone interested, TalkSport discussion this hour is:
Corners, 3.6% of corners lead to goals so are they overrated?

Durham is in the "they are overated" corner (sic)
 
Would be interesting to tot up the number of points we've gained through goals from corners and from what type they've come from.

Two yesterday for sure. One from a short corner by Eriksen to Davies. His perfectly floated cross, Wanyama's superb header, bang.
 
Last edited:
mudshark said:
Thing is, it's not like they've got all but one or two points clustered within a more normal range for numbers of corners. They appear to have a smooth distribution from 0 to 20 corners per game (including points at both those extremes, the latter of which apparently also averaged zero goals). I am forced to admit that I just don't understand what I am looking at when I look at their graph. More information required, I think. Perhaps they used some clever treatment that they haven't explained because they thought it was outside the scope, but I would much rather they had explained it, because the fact that the graph doesn't make sense to me can only lead me to view their conclusion with more scepticism than I did when I first read the book.
Click to expand...
It's not one visual point per data point. Most likely the point on the graph for 10 corners represents the average for all their data. And the same with all the others. There's no reason to expect anything other than what you describe as a smooth distribution in a visual representation like this. It's not showing the distribution of the actual data.

As far as I can tell their data presentation is perfectly fine.
 
http://statsbomb.com/2017/02/changing-how-the-world-thinks-about-set-pieces/
The first project I officially worked on inside football didn’t involve statistics, but it did involve analysis. My task was to take all the video for teams we knew were unusually good at set pieces, analyse what they were doing, and build a guide of best practices. Needless to say, we had some success.

What follows is a translation of two presentations I typically give to teams about this topic. I believe my work here remains fairly unique in the football world, and it has certainly benefited from working with top coaches like Gianni Vio and Brian Priske.

I can also assure you that when it comes to set pieces, we are just barely scratching the surface of what is possible...
Click to expand...

https://twitter.com/x/status/871662131237748736
 
milo said:
This is an interesting read

https://github.com/hesussavas/corners_stats/blob/master/README.md
Click to expand...

Question: What is considered as a "goal from corner"?
Answer: In this project only "the second touch goals" is analysed.
That mean the simplest scheme: Cross from corner -> Shot. No 3rd touch. No intermediate passes. No direct goals from the corner spot.



There are hundred of goals in this guys stats that aren't included as goal from corners. He also hasn't taken out any of the short corners from the total.

If you are gonna do stats do them right.
Don't just use whatever brick dataset is available and come up with a wrong figure of goals from corners that goes all over the internet.
And try to hid the fact you hate corners, smugly calling them primitive while sitting in Moscow supporting Borussia Dortmund and Liverpool.

If he were to watch a season of Championship football, he'll see why English fans in English grounds gets excited when their teams get a corner.
 
You must log in or register to reply here.
Back