XG

[Tamesaint]

I know that there is a lot of scepticism about XG on this forum but tonight must take the biscuit.

Match Stats 

Southampton 2 goals / 4 shots on target.

City 0 goals / 0 shits on target.

XG

Saints 0.69

City 1.0

Do City get 1 XG automatically because they come onto the pitch ???

 

 

 

 

 

[SotonianWill]

millions upon millions spent and they can’t even get a shit on target

3 minutes ago, Tamesaint said:

City 0 goals / 0 shits on target.

 

[SotonianWill]

maybe the bazunu save when the ball went across the box if i’m remembering right?

[S-Clarke]

I will forever believe that Xg is just nonsense, it's a stat put out there to make people feel better when they lose. ''well, we had a good xG so we're doing the right things''.

Irrelevant in my opinion. Stats can be worthwhile, but it's things like this just cloud the game. Shots/shots on target and the score in the top right hand corner are all that matter.

Edited 11 January, 2023 by S-Clarke

[John Boy Saint]

Maybe it just means Xtremely Gullible 

[31cc]

9 minutes ago, SotonianWill said:

millions upon millions spent and they can’t even get a shit on target

 

Perhaps they should sign Jason Puncheon..

[notnowcato]

I kind of like XG as a stat but like many stats they can miss context.
 

If I understand XG correctly, the position from where Djennapo shot would have a very low probability of scoring from; the missing context here was Ortega’s horrific positioning. 

[Saint Garrett]

I think like any stat you have to take it in context with everything else. Both our goals would have been low XG, especially Djenepo’s. I’m not sure how City got 1, but they did have quite a few decent opportunities inside the box. 

Pinch of salt. 

[John D]

This tweet shows how the cumulative xg is broken down.  Whether it is relevant or right I'll leave for you all to debate!

 

https://twitter.com/OptaAnalyst/status/1613439333544894464?t=SCCa3dHbwmOV_iYZl13PdA&s=19 

[Sheaf Saint]

27 minutes ago, John D said:

This tweet shows how the cumulative xg is broken down.  Whether it is relevant or right I'll leave for you all to debate!

 

https://twitter.com/OptaAnalyst/status/1613439333544894464?t=SCCa3dHbwmOV_iYZl13PdA&s=19 

That graph shows a big jump of 0.5 for City around 77 mins. What was that then? Would it have been the cross that Haaland couldn't quite connect with but would have been an almost certain goal if he had made any kind of contact?

[Saint86]

.

Edited 12 January, 2023 by Saint86

[Saint86]

15 hours ago, Sheaf Saint said:

That graph shows a big jump of 0.5 for City around 77 mins. What was that then? Would it have been the cross that Haaland couldn't quite connect with but would have been an almost certain goal if he had made any kind of contact?

Alvarez chance on circa 78/79 min, assisted by ilkay gundogan. Commentator was stating haaland scores it easily due to greater size / leg length.

Alvarez got on the cross but sliced it onto salisu. Pretty much dead centre 6 yards out, bazunu got down low to smother the deflection.

Notably the chart increased by 0.5xG at this exact time, which i think is reasonable as 50% of the time you'd expect the striker to score a dead centre tap in from 6yards... not including che who would absolutely do his best to pass to the keeper.

The marginal increases (particulary for Djenpo's chance) reflect how clinical we were. Not many shots go in from that kind of distance lets be honest.

A case of xG again showcasing its value in complimenting fan perceptions of a game?

Edited 12 January, 2023 by Saint86

[Turkish]

[verlaine1979]

Surprised Mara's goal had such low xg. Once he made space, he had a free strike from about ten yards at a ball already traveling with pace.

It seems to me that the other major weakness of xg is that it doesn't account for things like Adams being offside or on his heels in the six yard box, when any competent striker would be alive and alert to the defensive line.

Very difficult to watch that game and conclude that statistically we deserved less than one goal.

[Nolan]

So xG is essentially a stat showing how many Sitters you missed....

I agree with Turkish. B*llocks and simply nudge theory for gamblers.

[Ex Lion Tamer]

17 minutes ago, Nolan said:

So xG is essentially a stat showing how many Sitters you missed....

Yes that's exactly what it is. Why is that problem?

[Ex Lion Tamer]

7 hours ago, verlaine1979 said:

Surprised Mara's goal had such low xg. Once he made space, he had a free strike from about ten yards at a ball already traveling with pace.

It seems to me that the other major weakness of xg is that it doesn't account for things like Adams being offside or on his heels in the six yard box, when any competent striker would be alive and alert to the defensive line.

Very difficult to watch that game and conclude that statistically we deserved less than one goal.

I think you're underappreciating how much he still had to do and what a great finish it was

[egg]

It's bollox. Someone or something makes up a number. They then say that someone has over or under achieved their xG, but fail to point out the obvious, ie that their guess was wrong and that no stats take account of a worldy save, the striker skying it, or some other event that happens on a football pitch but not a spreadsheet. It's bollox.  

[aintforever]

8 hours ago, verlaine1979 said:

Very difficult to watch that game and conclude that statistically we deserved less than one goal.

I'm not sure that is what you are supposed to conclude from xg, it's just to give an idea of how many goal scoring opportunities a team had. It's a crude tool but is no different to any stat in that respect. Possession stats don't always tell you who deserved to win, neither do shots on goal or shots on target. All the stats need to be used in context to mean anything.

[gecko]

xG is like every other statistic that gets published about a game - it can be useful, but probably shouldn't be taken as fact.

The standard stats that we get (shots/on-target etc) rarely ever tell the whole story and all of them, xG included, need to be considered in the context of actually watching the game. For instance, the Brighton match:

https://understat.com/match/18371
Saints: 14 shots, 4 on target
Brighton: 7 shots, 4 on target
That would suggest that it was a reasonably balanced match to some degree, but anyone who actually watched it would know we got our arses handed to us and at no point did we ever look like we'd win the game.

Coincidentally, the xG on that match also fails to show things brilliantly as well.
Saints: 1.71 xG
Brighton: 1.01 xG

The Perraud O.G. is given an xG of 0.0, which seems to twist the numbers a little bit, and the JWP penalty and follow up are treated differently as they're two efforts, with xG's of 0.76 and 0.37 respectively (1.13 combined) - so discounting the penalty, the xG scores would be:
Saints: 0.58
Brighton: 1.01 xG
I'd also point out that Solly March's goal was spectacular, and rightly given an xG of just 0.04 - i.e. only 1 in 25 shots like that would hit the back of the net, which seems about right.

So taken in isolation on a specific match, they obviously don't always tell the complete story, just like all the other stats. But over the course of a season across all teams/shots/goals/saves etc, it probably spits out a reasonable estimate/average of how good a striker has performed. We know Che misses a lot of chances that he should score and his xG reflects that; we know Kane + Haaland score a lot of their chances, and again their xG reflects that.
tl;dr - it's not an exact science, it _can_ be useful, but it's not total bollocks

[Saint86]

56 minutes ago, Ex Lion Tamer said:

Yes that's exactly what it is. Why is that problem?

Yeah i don't the issue myself, but to each their own. The people that don't rate it must just not use any stats for games? Not even the simpler ones like possession percentage, shots on target etc. 

[Weston Super Saint]

5 minutes ago, Saint86 said:

Yeah i don't the issue myself, but to each their own. The people that don't rate it must just not use any stats for games? Not even the simpler ones like possession percentage, shots on target etc. 

Or score?

[Saint86]

43 minutes ago, gecko said:

xG is like every other statistic that gets published about a game - it can be useful, but probably shouldn't be taken as fact.

The standard stats that we get (shots/on-target etc) rarely ever tell the whole story and all of them, xG included, need to be considered in the context of actually watching the game. For instance, the Brighton match:

https://understat.com/match/18371
Saints: 14 shots, 4 on target
Brighton: 7 shots, 4 on target
That would suggest that it was a reasonably balanced match to some degree, but anyone who actually watched it would know we got our arses handed to us and at no point did we ever look like we'd win the game.

Coincidentally, the xG on that match also fails to show things brilliantly as well.
Saints: 1.71 xG
Brighton: 1.01 xG

The Perraud O.G. is given an xG of 0.0, which seems to twist the numbers a little bit, and the JWP penalty and follow up are treated differently as they're two efforts, with xG's of 0.76 and 0.37 respectively (1.13 combined) - so discounting the penalty, the xG scores would be:
Saints: 0.58
Brighton: 1.01 xG
I'd also point out that Solly March's goal was spectacular, and rightly given an xG of just 0.04 - i.e. only 1 in 25 shots like that would hit the back of the net, which seems about right.

So taken in isolation on a specific match, they obviously don't always tell the complete story, just like all the other stats. But over the course of a season across all teams/shots/goals/saves etc, it probably spits out a reasonable estimate/average of how good a striker has performed. We know Che misses a lot of chances that he should score and his xG reflects that; we know Kane + Haaland score a lot of their chances, and again their xG reflects that.
tl;dr - it's not an exact science, it _can_ be useful, but it's not total bollocks

Agree entirely.

[Sarisbury Saint]

9 minutes ago, Saint86 said:

Yeah i don't the issue myself, but to each their own. The people that don't rate it must just not use any stats for games? Not even the simpler ones like possession percentage, shots on target etc. 

if I was describing a game to someone that wasn’t there, I would tell them how many sitters we misssed and half chances created while dominating for most of the game ect. I certainly wouldn’t tell them our XG score is this and our XA score is that. 
We all know that Chez Adam’s should have scored a lot more than he has, but I don’t need a spreadsheet to tell me that.

[tajjuk]

9 hours ago, verlaine1979 said:

Surprised Mara's goal had such low xg. Once he made space, he had a free strike from about ten yards at a ball already traveling with pace.

It seems to me that the other major weakness of xg is that it doesn't account for things like Adams being offside or on his heels in the six yard box, when any competent striker would be alive and alert to the defensive line.

Very difficult to watch that game and conclude that statistically we deserved less than one goal.

It's a first time shot at an angle that beats a keeper at the near post, it's a tough chance and a lot of those chances get missed all the time so the XG seems pretty spot on to me.

It was a very good finish, I don't think you could ever call that a clear cut chance or even really a half chance. Both goals being low XG makes sense considering they are hard finishes, even accounting for the poor positioning the execution to lift it like that whilst on the run was very good so again a pretty low percentage chance. 

Alvarez's chance in front of goal makes sense for being the best chance of the game, he's 6 yards out with the whole goal to aim at and doesn't get a clean connection. Highly likely had Haaland had the same chance then he would have least got that on target I reckon.

I do find the overall vitriol quite hilarious, the old men seem to get very touchy about it for no good reason, it's a pretty self explanatory statistic and makes sense most of the time. It's not perfect but it basically assesses how good the chances you got in the game were so you can see if you created decent chances, conceded decent chances, which is pretty helpful when analysing the game.  As said above as with any statistics you have to view them in context and their will be caveats, like mentioned above a penalty miss and subsequent rebound can create a huge amount of XG, but the latter chance only comes from the penalty/save miss and if that is all you did in the game then it will create a false position, but it's a better guide than shots/shots on target. 

 

Edited 13 January, 2023 by tajjuk

[Lee On Solent Saint]

18 minutes ago, Saint86 said:

Yeah i don't the issue myself, but to each their own. The people that don't rate it must just not use any stats for games? Not even the simpler ones like possession percentage, shots on target etc. 

Jesus Christ. The only stat that makes any difference to me is the one that says Saints got more balls in the back of the net than the other team

[Nolan]

22 minutes ago, Saint86 said:

Yeah i don't the issue myself, but to each their own. The people that don't rate it must just not use any stats for games? Not even the simpler ones like possession percentage, shots on target etc. 

Stats are fantastic.  xG just looks subjective though, and therefore, as a stat, its useless.

 

 

[aintforever]

7 minutes ago, Sarisbury Saint said:

We all know that Chez Adam’s should have scored a lot more than he has, but I don’t need a spreadsheet to tell me that.

Of course, but IF a spreadsheet does tell you that it has done it's job. There are only so many games a scout/manager can watch, you can feed stats from thousands of games onto a spreadsheet.

[Ex Lion Tamer]

1 hour ago, egg said:

It's bollox. Someone or something makes up a number. They then say that someone has over or under achieved their xG, but fail to point out the obvious, ie that their guess was wrong and that no stats take account of a worldy save, the striker skying it, or some other event that happens on a football pitch but not a spreadsheet. It's bollox.  

The whole point of it is you see how well a team has performed without taking into account worldie saves and bad finishing

[gecko]

14 minutes ago, Lee On Solent Saint said:

Jesus Christ. The only stat that makes any difference to me is the one that says Saints got more balls in the back of the net than the other team

I like the simplicity here.

However, if there is a statistic that points out to the club that our strikers regularly under perform, and it highlights who might be able to do a better job, then that could show who to bring in to improve the number of goals in the back of the net.
xG correctly tells the club that Che misses loads, and if we'd kept Ings, we would (probably) have more goals than we do now - not sure many would argue that even if there wasn't a stat to back it up.

[Weston Super Saint]

14 minutes ago, Ex Lion Tamer said:

The whole point of it is you see how well a team has performed without taking into account worldie saves and bad finishing

Seems completely redundant for us then...

[Ex Lion Tamer]

5 minutes ago, Weston Super Saint said:

Seems completely redundant for us then...

I for one hadn't clocked that City actually had the best chances in that game so the stat was useful. Doesn't take away from the fact that we finished better and deserved to win

[Lee On Solent Saint]

9 minutes ago, gecko said:

I like the simplicity here.

However, if there is a statistic that points out to the club that our strikers regularly under perform, and it highlights who might be able to do a better job, then that could show who to bring in to improve the number of goals in the back of the net.
xG correctly tells the club that Che misses loads, and if we'd kept Ings, we would (probably) have more goals than we do now - not sure many would argue that even if there wasn't a stat to back it up.

I don't need a stat to understand that our forward players regularly under perform. Answer is every week out on the pitch.

[Ex Lion Tamer]

37 minutes ago, Sarisbury Saint said:

if I was describing a game to someone that wasn’t there, I would tell them how many sitters we misssed and half chances created while dominating for most of the game ect. I certainly wouldn’t tell them our XG score is this and our XA score is that. 
We all know that Chez Adam’s should have scored a lot more than he has, but I don’t need a spreadsheet to tell me that.

if I was describing a game to someone that wasn’t there, I wouldn't take my clothes off and jump in the shower, but it doesn't mean I shouldn't wash 🙂

[Whitey Grandad]

10 hours ago, Saint86 said:

Alvarez chance on circa 78/79 min, assisted by ilkay gundogan. Commentator was stating haaland scores it easily due to greater size / leg length.

Alvarez got on the cross but sliced it onto salisu. Pretty much dead centre 6 yards out, bazunu got down low to smother the deflection.

Notably the chart increased by 0.5xG at this exact time, which i think is reasonable as 50% of the time you'd expect the striker to score a dead centre tap in from 6yards... not including che who would absolutely do his best to pass to the keeper.

The marginal increases (particulary for Djenpo's chance) reflect how clinical we were. Not many shots go in from that kind of distance lets be honest.

A case of xG again showcasing its value in complimenting fan perceptions of a game?

Nope. It’s rubbish. It doesn’t take into account that it was Alvarez and not Haaland.

And I think you meant to write ‘complementing’ and not ‘complimenting’. This bloody autocorrect gets everywhere;)

[Whitey Grandad]

1 hour ago, aintforever said:

I'm not sure that is what you are supposed to conclude from xg, it's just to give an idea of how many goal scoring opportunities a team had. It's a crude tool but is no different to any stat in that respect. Possession stats don't always tell you who deserved to win, neither do shots on goal or shots on target. All the stats need to be used in context to mean anything.

And all the stats put together mean nothing. 

[gecko]

10 minutes ago, Lee On Solent Saint said:

I don't need a stat to understand that our forward players regularly under perform. Answer is every week out on the pitch.

yes, but the point is who do we get to improve on this?
Players who outperform their xG are likely to be the ones we need to get in - and if we can use this amongst other factors to focus our scouting on people we can actually afford, then great

[aintforever]

4 minutes ago, Whitey Grandad said:

And all the stats put together mean nothing. 

0 points from 3 games - that's a stat.

[Whitey Grandad]

Why is it that xG never comes close to the actually result of a game, even with the benefit of hindsight?

Anyone who was at that game on Wednesday will have come away thinking that Saints could well have scored three or four whilst City might have got one. If they were very lucky.

[Whitey Grandad]

1 minute ago, aintforever said:

0 points from 3 games - that's a stat.

Only if you want to think of it as such. The rest of us know that we are on a shit run and don’t need any figures to prove it.

[aintforever]

5 minutes ago, Whitey Grandad said:

Only if you want to think of it as such. The rest of us know that we are on a shit run and don’t need any figures to prove it.

But if you hadn't watched the games the 0 points from 3 games would give you a clue though wouldn't it?

[Saint86]

1 hour ago, Whitey Grandad said:

Nope. It’s rubbish. It doesn’t take into account that it was Alvarez and not Haaland.

And I think you meant to write ‘complementing’ and not ‘complimenting’. This bloody autocorrect gets everywhere;)

I mean, this is the point. It's an average expectation of a striker scoring a chance. What it's saying is that on average a player can be expected to score that chance 50% of the time... With haaland, the stats shows he converts at a rate of 1.3 goals per xG (ditto ings) So they'd be expected to score that chance 65% of the time on average 🙃. 

The fact that it counts so heavily on xG is a far better demonstration of the threat city carried than simply saying they had 1 shot off target 🤷‍♂️. 

You dont have to use any stats, but it you ever look at (or quote) stats such as total shots, shots on target, possession, corners, clear cut chances etc. Then you should consider xG and others. 

That Alvarez chance was counted as city's clear cut chance by sky etc, saints didn't have one listed because none of our chances were clear cut - instead we had very clinical finishing - something that will now be recorded against mara and djenpo overall stats.

People are welcome to ignore these more detailed stats if they want to. But it's dumb to consider the basic ones and ignore the more detailed and nuanced ones - They clearly give a more detailed view of player and team performances over the course of a game or over a season etc.

 

Edited 13 January, 2023 by Saint86

[Saint86]

2 hours ago, Ex Lion Tamer said:

if I was describing a game to someone that wasn’t there, I wouldn't take my clothes off and jump in the shower, but it doesn't mean I shouldn't wash 🙂

Telling someone how many sitters we missed is a simplistic way of telling them our xG and goals scored 🤔.

Edited 13 January, 2023 by Saint86

[notnowcato]

Essentially, there are 2 types of people, those that can extrapolate information from data

 

It’s easier to call it bollox. The data directly derived from video is brilliant and is less subjective than some think. 

[Weston Super Saint]

43 minutes ago, notnowcato said:

Essentially, there are 2 types of people, those that can extrapolate information from data

 

It’s easier to call it bollox. The data directly derived from video is brilliant and is less subjective than some think. 

You've only listed one type...

I'm assuming you've deliberately left out the second in a hilarious way of saying all others are stupid, yet you neglect to factor in a third group that can extrapolate information from data and still think it's shit.

Lies, damned lies and statistics ..... 

[Patrick Bateman]

I work closely with data, I understand data, but xG is a complete load of crap.  According to that stat, Man City (with rounding) SHOULD have won 2-1.  They didn't, that wasn't the score.  So a retrospective at what was expected is a load of rubbish. 

You are supposed to take learnings from data, what can you learn from that? "You SHOULD have scored" and "you SHOULD have hit the target" - brilliant. I think the qualitative data from staff and fans grumbling in the stands is enough to form that view.

Data for data's sake. There are examples of it everywhere.

Look at a shampoo advert; "loved by 89% of people" ... "note: 89% of 63 people" ... WTF is that? That's not a rep sample. Who are these people, what's their demographics, are they a broad enough spectrum and what would give more significance? 1000, 2000, probably 20000.  Anyway, I'm ranting now, your fault. Bloody xG my arse. 

[Whitey Grandad]

2 hours ago, Saint86 said:

I mean, this is the point. It's an average expectation of a striker scoring a chance. What it's saying is that on average a player can be expected to score that chance 50% of the time... With haaland, the stats shows he converts at a rate of 1.3 goals per xG (ditto ings) So they'd be expected to score that chance 65% of the time on average 🙃. 

The fact that it counts so heavily on xG is a far better demonstration of the threat city carried than simply saying they had 1 shot off target 🤷‍♂️. 

You dont have to use any stats, but it you ever look at (or quote) stats such as total shots, shots on target, possession, corners, clear cut chances etc. Then you should consider xG and others. 

That Alvarez chance was counted as city's clear cut chance by sky etc, saints didn't have one listed because none of our chances were clear cut - instead we had very clinical finishing - something that will now be recorded against mara and djenpo overall stats.

People are welcome to ignore these more detailed stats if they want to. But it's dumb to consider the basic ones and ignore the more detailed and nuanced ones - They clearly give a more detailed view of player and team performances over the course of a game or over a season etc.

 

But xG is never anywhere close to the actual goals scored, which it should be if it ever meant anything.

[Whitey Grandad]

3 hours ago, aintforever said:

But if you hadn't watched the games the 0 points from 3 games would give you a clue though wouldn't it?

A clue as to what? 

[Whitey Grandad]

1 hour ago, notnowcato said:

Essentially, there are 2 types of people, those that can extrapolate information from data

 

It’s easier to call it bollox. The data directly derived from video is brilliant and is less subjective than some think. 

It's bollox because the data taht is extracted bears no relationship to what happened on the pitch.

[Swils]

Expected and actual are different.  You can't say XG is bollox because that's not what the score was.  XG is representative of the average, not a single instance.