Thanks for this. So it's actual conclusion is it's useless for predicting results. Takes no account of who's got the chance, how much pressure they are under or even which way their facing and body shape when receiving the ball. The bit it would seem useful for is comparing strikers ability to convert XG. The fact Rotherham have a high Xg but low goals could simply be giving the ball to the wrong player (one under pressure) or having useless strikers. It may well be they will improve, but for people to say it is likely because of Xg doesn't seem logical, and that is actually stated in the video.
So these 'stats' are actually based on subjective opinions then?
Hmmm.
Next we'll be having people tell us that a goal has 1.4 chance of going to VAR where it will have a 0.82 chance of being disallowed, and trying to work out the outcome of the game on that.
Not at all. It’ll be a collection of data points on hundreds of thousands of shots and the position each was taken from, with a result outcome attached.
Let’s take a 1m square around the penalty spot for example… a massive database has a collection of 100,000 shots from that specific patch of grass. Now we see that 8,000 shots were goals.
8,000 / 100,000 = 0.08
That’s the starting point for an xG model. Most models will have it broken down further into right foot, or left foot, or header - regular play, or counter attack, or set piece. The most advanced models that the public don’t have access to will include pressure from opponents, goalkeeper positioning, time since turnover of possession etc.
But whatever the parameters, the calculations should be consistent or it’s a crap model.
Noticed one of the pundits refused to mention xg on Match of the Day last week (Might have been Shearer) when the graphic came up.
Not a fan of it and it looks a nonsense when you look at the start we have made this season
Not really, it shows we’ve won 3 games that have been quite low margin games, in which we’ve barely conceded any good chances. The xG tells the story quite well
Thanks for this. So it's actual conclusion is it's useless for predicting results. Takes no account of who's got the chance, how much pressure they are under or even which way their facing and body shape when receiving the ball. The bit it would seem useful for is comparing strikers ability to convert XG. The fact Rotherham have a high Xg but low goals could simply be giving the ball to the wrong player (one under pressure) or having useless strikers. It may well be they will improve, but for people to say it is likely because of Xg doesn't seem logical, and that is actually stated in the video.
It can’t take into account who had the chance because that becomes an opinion and means you wouldn’t be able to compare players, i.e. top class strikers who out perform their xG vs an average striker.
The video is 6 years old and there are different versions of the model out there, some will take into account different factors or interpret them in a different way. I would imagine there have been improvements and tweaks in that time, and as has been said it’s only one statistic to use alongside various others, rather than the full picture.
Noticed one of the pundits refused to mention xg on Match of the Day last week (Might have been Shearer) when the graphic came up.
Not a fan of it and it looks a nonsense when you look at the start we have made this season
Not really, it shows we’ve won 3 games that have been quite low margin games, in which we’ve barely conceded any good chances. The xG tells the story quite well
Perhaps but not as well as the actual match results do.
It’s just like any other stat. Feel free to use it or ignore it.
With xG, a common complaint is that “it doesn’t take into account who took the shot” - but it was never meant to.
It’s like seeing England having 20 shots in a match and asking “but how many shots were taken by Harry Kane?” - I don’t think I’ve seen that many times, if ever.
I still think the stats generally show what everyone can see. We can see we create very few chances, don't get in the opposition box enough, have few shots and are likely to score few goals. We also have the minority of possession, win possession in the final third by pressing, defend resolutely and concede very few chances. The stats tell us what virtually everyone can see, which is why the majority expect games to be 0-0, 1-0, 0-1 or 1-1, with a chance of one team scoring 2.
If you are 19th in the league for shots on target then you have to have the very best players to convert the limited chances you are creating. Or you dont sell the leagues top goalscorer.
Surprised our ‘touches in oppo box’ is ranked so high given our crossing from outside of the box tends to bypass the box completely and head straight for the opposite touchline.
I am sure xg has it's place and many, my son included, take an interest and enjoy discussing it. For me, chuntering on about Xg is amongst a number of things that renders football an utterly joyless experience. A boring ever changing variable to tittilate nerds (sorry, son )
Best quality chances created, worst quality chances conceded. #JonesBall
Only team to average more than 0.14 xG per shot taken and less than 0.06 xG per shot conceded.
Quite encouraging that the teams that show up the best in this currently sit 2nd 3rd 4th and 5th. Who knew xG over a stretch of matches might actually be a good indicator of team strength?
Best quality chances created, worst quality chances conceded. #JonesBall
Only team to average more than 0.14 xG per shot taken and less than 0.06 xG per shot conceded.
Quite encouraging that the teams that show up the best in this currently sit 2nd 3rd 4th and 5th. Who knew xG over a stretch of matches might actually be a good indicator of team strength?
who defines if a shot is quality, good or just an average shot ?
Best quality chances created, worst quality chances conceded. #JonesBall
Only team to average more than 0.14 xG per shot taken and less than 0.06 xG per shot conceded.
Quite encouraging that the teams that show up the best in this currently sit 2nd 3rd 4th and 5th. Who knew xG over a stretch of matches might actually be a good indicator of team strength?
who defines if a shot is quality, good or just an average shot ?
A computer and some fancy models made by clever people
Best quality chances created, worst quality chances conceded. #JonesBall
Only team to average more than 0.14 xG per shot taken and less than 0.06 xG per shot conceded.
Quite encouraging that the teams that show up the best in this currently sit 2nd 3rd 4th and 5th. Who knew xG over a stretch of matches might actually be a good indicator of team strength?
Except that wasn't the measure, or to be more accurate combined measures, being used earlier in the season.
Not just XG but per shots attempted but also quality of chances conceded, which as many pointed out was a clear strength in our early games.
And the best indicator of team strength still remains the league table
Best quality chances created, worst quality chances conceded. #JonesBall
Only team to average more than 0.14 xG per shot taken and less than 0.06 xG per shot conceded.
Quite encouraging that the teams that show up the best in this currently sit 2nd 3rd 4th and 5th. Who knew xG over a stretch of matches might actually be a good indicator of team strength?
Except that wasn't the measure, or to be more accurate combined measures, being used earlier in the season.
Not just XG but per shots attempted but also quality of chances conceded, which as many pointed out was a clear strength in our early games.
And the best indicator of team strength still remains the league table
Except it's not, at earlier stages of the season.
Points are won by goals, and since there are few goals in each game, statistically it is not as significant as a softer measure like xG, which has more data points.
It's clear you don't get it which is fine, but there's no need to be so dismissive of something you personally can't see the value in.
Best quality chances created, worst quality chances conceded. #JonesBall
Only team to average more than 0.14 xG per shot taken and less than 0.06 xG per shot conceded.
Quite encouraging that the teams that show up the best in this currently sit 2nd 3rd 4th and 5th. Who knew xG over a stretch of matches might actually be a good indicator of team strength?
Except that wasn't the measure, or to be more accurate combined measures, being used earlier in the season.
Not just XG but per shots attempted but also quality of chances conceded, which as many pointed out was a clear strength in our early games.
And the best indicator of team strength still remains the league table
Except it's not, at earlier stages of the season.
Points are won by goals, and since there are few goals in each game, statistically it is not as significant as a softer measure like xG, which has more data points.
It's clear you don't get it which is fine, but there's no need to be so dismissive of something you personally can't see the value in.
No need to be so patronising either but you were
And I "get it", I just think it is given fair too much significance, especially when it is used in isolation, as above table and many post on this thread show.
And points are not won by goals as you incorrectly state.
Points are won by winning or drawing games.
That generates points which make up the league table and it's the league table that decides whether you get promoted, relegated or stay in the same division, not XG.
Best quality chances created, worst quality chances conceded. #JonesBall
Only team to average more than 0.14 xG per shot taken and less than 0.06 xG per shot conceded.
Quite encouraging that the teams that show up the best in this currently sit 2nd 3rd 4th and 5th. Who knew xG over a stretch of matches might actually be a good indicator of team strength?
Except that wasn't the measure, or to be more accurate combined measures, being used earlier in the season.
Not just XG but per shots attempted but also quality of chances conceded, which as many pointed out was a clear strength in our early games.
And the best indicator of team strength still remains the league table
Except it's not, at earlier stages of the season.
Points are won by goals, and since there are few goals in each game, statistically it is not as significant as a softer measure like xG, which has more data points.
It's clear you don't get it which is fine, but there's no need to be so dismissive of something you personally can't see the value in.
No need to be so patronising either but you were
And I "get it", I just think it is given fair too much significance, especially when it is used in isolation, as above table and many post on this thread show.
And points are not won by goals as you incorrectly state.
Points are won by winning or drawing games.
That generates points which make up the league table and it's the league table that decides whether you get promoted, relegated or stay in the same division, not XG.
Thanks for explaining the basics of a league table, but sadly you've missed the point that statistics are more significant with more data points.
If you couldn't grasp that goals are the drivers of what generates points, then I feel explaining any further would be a fruitless exercise for both of us.
Comments
Let’s take a 1m square around the penalty spot for example… a massive database has a collection of 100,000 shots from that specific patch of grass. Now we see that 8,000 shots were goals.
8,000 / 100,000 = 0.08
That’s the starting point for an xG model. Most models will have it broken down further into right foot, or left foot, or header - regular play, or counter attack, or set piece. The most advanced models that the public don’t have access to will include pressure from opponents, goalkeeper positioning, time since turnover of possession etc.
But whatever the parameters, the calculations should be consistent or it’s a crap model.
Not a fan of it and it looks a nonsense when you look at the start we have made this season
The video is 6 years old and there are different versions of the model out there, some will take into account different factors or interpret them in a different way. I would imagine there have been improvements and tweaks in that time, and as has been said it’s only one statistic to use alongside various others, rather than the full picture.
With xG, a common complaint is that “it doesn’t take into account who took the shot” - but it was never meant to.
It’s like seeing England having 20 shots in a match and asking “but how many shots were taken by Harry Kane?” - I don’t think I’ve seen that many times, if ever.
We can see we create very few chances, don't get in the opposition box enough, have few shots and are likely to score few goals.
We also have the minority of possession, win possession in the final third by pressing, defend resolutely and concede very few chances.
The stats tell us what virtually everyone can see, which is why the majority expect games to be 0-0, 1-0, 0-1 or 1-1, with a chance of one team scoring 2.
Said it from the start, if we win, it's going to be great, but the losses are going to feel very draining.
We just simply don't play exciting football. It's just effective. Hopefully effective enough to get us promoted.
For me, chuntering on about Xg is amongst a number of things that renders football an utterly joyless experience. A boring ever changing variable to tittilate nerds (sorry, son
Only team to average more than 0.14 xG per shot taken and less than 0.06 xG per shot conceded.
Quite encouraging that the teams that show up the best in this currently sit 2nd 3rd 4th and 5th. Who knew xG over a stretch of matches might actually be a good indicator of team strength?
There's a helpful video that explains this at the top of the page, originally posted by Scoham.
Not just XG but per shots attempted but also quality of chances conceded, which as many pointed out was a clear strength in our early games.
And the best indicator of team strength still remains the league table
Points are won by goals, and since there are few goals in each game, statistically it is not as significant as a softer measure like xG, which has more data points.
It's clear you don't get it which is fine, but there's no need to be so dismissive of something you personally can't see the value in.
And I "get it", I just think it is given fair too much significance, especially when it is used in isolation, as above table and many post on this thread show.
And points are not won by goals as you incorrectly state.
Points are won by winning or drawing games.
That generates points which make up the league table and it's the league table that decides whether you get promoted, relegated or stay in the same division, not XG.
If you couldn't grasp that goals are the drivers of what generates points, then I feel explaining any further would be a fruitless exercise for both of us.