The xG nonsense

[AC Mack]

It's complete bollocks. If there was a xT (expected tackles) or xB (expected blocks) stat for defending, we'd be top 5 to balance it out, hence us being 7th.

xGa (expected goals against) might be helpful - someone earlier said we were 4th.

My other main issue I forgot to mention is how it calculates multiple shots/chances in one move, which has always seemed off to me.

 

[SuperKevsLeftBoot]

I haven't got shares, it's just you're contradicting yourself throughout.

"xG is an opinion, not a stat". It's not an opinion, it's the statistical value of how many goals were scored from shots taken in that position. But you're arguing against it because IN YOUR OPINION you think Brobbey's chance was easier than that.

Your opinion. Subjective.

How you interpret the data is up to you - our 'xG conceded' has been high all season, you could think that to be a concern based on the overall number or could consider that we deliberately conceded a high volume of low quality shots by playing a deep defensive line and encouraging teams to shoot from long range knowing we had a really good shotstopper in goal. That's the subjective part. The 'high xG conceded' part isn't subjective, it's objective.

where have I contradicted myself?
There are far too many variables (which you’ve presented) for xg to be an informed stat to use especially when it’s used to slag us off.

My opinion is worth more than that fkn xg stat based on those chances and the vast majority would agree that the chance was far easier. If you need a stat to tell you that then meh

I bet you’re one of those slagging us off coz of xg

Shots on goal is accurate but not informative.

You could have loads of shots on goal in a match but all of them taken from outside the box with 3 or 4 defenders in your way.

xG gives a relative value of the quality of your shooting positions and situations.

As lots of people are saying it’s not the be all and end all but it’s a really useful comparative stat, especially for clubs themselves.

It’s not If it’s not accurate FFS

Don't know why people take notice of shit like this

Coz its used to slag us off

 

[GipseT]

xG is not nonsense but not many understand what it is...

What does xG predict for the rest of the season?

Another thread arguing the merits of xg by people who at best hold a gcse in maths. I think a lot of it is this. There was an article in Roker Report the other day about xG and the person writing it simply didn’t understand what xG is and what it can and can’t tell you

www.readytogo.net

 

[G Whizz]

xG is not nonsense but not many understand what it is...

What does xG predict for the rest of the season?

Another thread arguing the merits of xg by people who at best hold a gcse in maths. I think a lot of it is this. There was an article in Roker Report the other day about xG and the person writing it simply didn’t understand what xG is and what it can and can’t tell you

www.readytogo.net

Thanks for sharing. That's really interesting

Can you expand on this? I can't picture what you're referring to

Another problem for the model is that the top 4 usually sees games out in a way that the model cant handle. And yet another is that the bottom teams seams to collapse and are really bad att chasing games again setting up problems for the model.

 

[Hulkenburger]

I wondered this but compared our xG against to shots against and the xG per shot seemed right in the mix with most other clubs

It was true until about March. When we started playing more aggressively we ended up conceding better chances and, predictably, more goals.

 

[johnkaysleftleg]

I've been looking into Xg in the last few days using a spreadsheet (nerd alert!!) of actual results and Xg vales for each game. The Xg I'm using the the most advanced, i.e OPTA.
Unfortunately for the critics Xg isn't a nonsense, last season 1045 goals were scored in the EPL the total Xg was 1043, so remarkably accurate.
Fortunately expected points are a deeply flawed metric however.
Overall the correlation between Xg and actual results is around 52% not particularly great but better than any other non goal metric and not inconsequential.
If we have a deeper delve into the actual data an interesting trend can be seen.
If a team has an Xg advantage during a game of 0.75 or better they will win 67% of the time, now this is a strong correlation.
Below this the correlation falls to less than 36%, scarcely better than pulling Home, Away or draw out of a hat and certainly not something to build any kind of prediction on.
Given this accounts for 48% of all games played in the EPL this season we can safely call Expected points calculated this way a pile of arse.
For the record Sunderland had a significant Xg (0.75 or over) advantage 5 times this season and won the lot. In the games we were on the wrong end we managed only 6 points from 14 gmaes (D6L8).

 

[Kent Boy 2]

xG is one statistic that should not be used purely on its own, but it is relatively good to help describe a predicted outcome.

I’m not sure why people get so hung up on it being terrible by using one or two examples. Yes, Sunderland were an outlier, but all statistics have multiple outliers - otherwise there would be no betting industry. However, it’s a pretty good indicator when you consider that 5 of the top 6 in the xG table were in the top 6 of the actual table, and the bottom 5 in the actual table were all in the bottom 7 in the xG table.

It’s like saying that scoring more goals that you concede is a bad statistic because Sunderland finished 7th with a GD of -6, whilst Forest finished 16th with a GD of -3

 

[Definitely Maybe]

where have I contradicted myself?
There are far too many variables (which you’ve presented) for xg to be an informed stat to use especially when it’s used to slag us off.

My opinion is worth more than that fkn xg stat based on those chances and the vast majority would agree that the chance was far easier. If you need a stat to tell you that then meh

I bet you’re one of those slagging us off coz of xg

It’s not If it’s not accurate FFS

Coz its used to slag us off

How is '33,000 out of 100,000 shots were scored' inaccurate?

 

[FatDrummer]

…….xG has value as a comparator (it would show Harry Kane is a better finisher than me) but fairly useless alone

This.
And ONLY this.
Anyone that uses it as a single, sole arbiter of goalscoring prowess is completely missing the whole picture.

 

[SuperKevsLeftBoot]

How is '33,000 out of 100,000 shots were scored' inaccurate?

Because that’s clearly not happened has it
They’ve had a top professional striker take 100,000 shots from that position with the keeper in the same position hahaha Chesus wept

 

[A Visiting Villan]

XG had us down as 13th and Hull as relegated. It's a useless stat for the kind of people who live in their parents spare bedroom.

 

[Charmless Man]

It’s not If it’s not accurate FFS

It’s accurate to the extent it needs to be - i.e it gives you a fairly decent idea of a shot’s chances of going in given where it was taken from, the presence of defenders etc’ I don’t think I’m quite seeing how it’s not.

Yes different models exist so obviously when making comparisons you need to make sure you’re making comparisons with the same model.

xG is one statistic that should not be used purely on its own, but it is relatively good to help describe a predicted outcome.

I’m not sure why people get so hung up on it being terrible by using one or two examples. Yes, Sunderland were an outlier, but all statistics have multiple outliers - otherwise there would be no betting industry. However, it’s a pretty good indicator when you consider that 5 of the top 6 in the xG table were in the top 6 of the actual table, and the bottom 5 in the actual table were all in the bottom 7 in the xG table.

It’s like saying that scoring more goals that you concede is a bad statistic because Sunderland finished 7th with a GD of -6, whilst Forest finished 16th with a GD of -3

And outliers are useful data in and of themselves. Why is this thing confounding what we might expect is a useful question to ask and try to answer.

I really don’t get people’s problem with that

If Sunderland are an outlier then Sunderland should be interested in whether that’s sustainable or we need to improve in areas. Other teams should interested in how they can exploit whatever is making them an outlier.

 

[TJlivesaway]

XG is a flawed system as its only a rating of probability of where a goal would be scored from in relation to the position on the pitch. It doesn't take into account the ability of the player, the ability of the goal keeper, the perceived pressure of the moment, the technique involved in the strike, etc. Ultimately, its bollocks. The fact that we've scored more goals than we're 'supposed' to have scored is a bizarre concept, cos we still think we haven't scored enough anyway.

 

[SuperKevsLeftBoot]

It’s accurate to the extent it needs to be - i.e it gives you a fairly decent idea of a shot’s chances of going in given where it was taken from, the presence of defenders etc’ I don’t think I’m quite seeing how it’s not.

Yes different models exist so obviously when making comparisons you need to make sure you’re making comparisons with the same model.


And outliers are useful data in and of themselves. Why is this thing confounding what we might expect is a useful question to ask and try to answer.

I really don’t get people’s problem with that

If Sunderland are an outlier then Sunderland should be interested in whether that’s sustainable or we need to improve in areas. Other teams should interested in how they can exploit whatever is making them an outlier.

So you think brobbeys chance against Chelsea is a 1 in 3 chance when he’s 1v1 8/9 years out and has a huge gap on the right to score?

 

[Kent Boy 2]

It’s accurate to the extent it needs to be - i.e it gives you a fairly decent idea of a shot’s chances of going in given where it was taken from, the presence of defenders etc’ I don’t think I’m quite seeing how it’s not.

Yes different models exist so obviously when making comparisons you need to make sure you’re making comparisons with the same model.


And outliers are useful data in and of themselves. Why is this thing confounding what we might expect is a useful question to ask and try to answer.

I really don’t get people’s problem with that

If Sunderland are an outlier then Sunderland should be interested in whether that’s sustainable or we need to improve in areas. Other teams should interested in how they can exploit whatever is making them an outlier.

I couldn’t agree more. Every season a couple of clubs over achieve versus their xG. It was Sunderland and Villa this season. 2 from 20 is 10%. Statistically if we have the same relative xG next season then we will almost certainly be in the bottom 3 or 4. People completely ignoring a statistic because 2 clubs were anomalies, when 18 weren’t, is madness. Thankfully the club have a whole data analytics department that understands these statistic properly, and uses them to identify areas for improvement.

 

[Charmless Man]

XG is a flawed system as its only a rating of probability of where a goal would be scored from in relation to the position on the pitch. It doesn't take into account the ability of the player, the ability of the goal keeper, the perceived pressure of the moment, the technique involved in the strike, etc. Ultimately, its bollocks. The fact that we've scored more goals than we're 'supposed' to have scored is a bizarre concept, cos we still think we haven't scored enough anyway.

They do account for things like the perceived pressure of the moment (I.e relative defender positioning), how the shot was taken (header, left foot, right foot), the nature of the assist (e.g cross vs through ball) though.

Obviously there’s some approximation involved because it’s a mathematical model but that doesn’t mean it’s bollocks, it just means you need to be aware when using it.

 

[Kent Boy 2]

So you think brobbeys chance against Chelsea is a 1 in 3 chance when he’s 1v1 8/9 years out and has a huge gap on the right to score?

I’d say that the last 3 times a Sunderland player has been through clean on goal - Isidor at Villa, Diarra at Villa, and Brobbey v Chelsea, only one goal resulted. That seems to be 1 in 3 to me .

They do account for things like the perceived pressure of the moment (I.e relative defender positioning), how the shot was taken (header, left foot, right foot), the nature of the assist (e.g cross vs through ball) though.

Obviously there’s some approximation involved because it’s a mathematical model but that doesn’t mean it’s bollocks, it just means you need to be aware when using it.

Again, I couldn’t agree more

 

[SuperKevsLeftBoot]

I’d say that the last 3 times a Sunderland player has been through clean on goal - Isidor at Villa, Diarra at Villa, and Brobbey v Chelsea, only one goal resulted. That seems to be 1 in 3 to me .

So it’s based on players ability as well then? Coz if Wilson had all 3 he’d have have scored two if not all 3.

If it takes into account the player then I’d understand it more. Brobbeys finishing isn’t the best.. Wilson is far better

 

[Charmless Man]

So it’s based on players ability as well then? Coz if Wilson had all 3 he’d have have scored two if not all 3.

Would he? He’s missed some sitters from memory.

The good thing about using stats is we don’t have to rely on memory because I actually couldn’t hang my hat on any probability from my own recall.

So you think brobbeys chance against Chelsea is a 1 in 3 chance when he’s 1v1 8/9 years out and has a huge gap on the right to score?

See above.

 

[EternallySAFC]

So you think brobbeys chance against Chelsea is a 1 in 3 chance when he’s 1v1 8/9 years out and has a huge gap on the right to score?

Yes