Table based on xG

[And Its Smith]

Some will know I’m an xG sceptic.  I don’t mind the principle behind it but find the execution someone too flawed to prove reliable.  However, I know that some people take a lot of notice of it so thought I’d share this that some might find interesting.  It would suggest, if you believe in xG that we are currently overperforming and may well ‘regress to the mean’ (an annoying phrase that gets used a lot nowadays) at some point.  The caveat is that 8 games isn’t a lot to go on.  It certainly doesn’t feel that we are getting more points from games than we deserve 

[mozo]

23 minutes ago, And Its Smith said:

Some will know I’m an xG sceptic.  I don’t mind the principle behind it but find the execution someone too flawed to prove reliable.  However, I know that some people take a lot of notice of it so thought I’d share this that some might find interesting.  It would suggest, if you believe in xG that we are currently overperforming and may well ‘regress to the mean’ (an annoying phrase that gets used a lot nowadays) at some point.  The caveat is that 8 games isn’t a lot to go on.  It certainly doesn’t feel that we are getting more points from games than we deserve 

Could this also say that our strikers are all razor sharp this season, scoring chances that only top strikers score? Maybe they'll stay sharp 

[Mr Popodopolous]

Certainly not a perfect model but some of our more basic underlying numbers feel a bit midtableish to me. Enjoyable football without doubt and the attacking football is great to see but is Conway going to sustain his brilliant form eg? Wells in the best form of his time here...then again a fully fit Semenyo adds a new dimension as we know and Kalas as and when he is fully fit will only help.

On the other hand, BAD refereeing decisions have influenced these numbers.

Even if you disregard the penalty claims as maybe maybe not, the one against at Hull, the non red card for Bennett in the heat and when we were ahead and on top and the Sykes red but Freeman reprieve, all of these negatively impacted us. Not just results wise but statistically too, badly- we were so dominant at 11 v 11 vs Luton for example.

Edited September 13, 2022 by Mr Popodopolous

[Davefevs]

as per my usual response….trying to take the difference between XG for and XG against in a 90 minute period and turn it into a result is futile.  Trying to then put that into a standard league table is even worse.  I love E361’s work…apart from this!!!  An average game has circa 10 shots each team, so it’s such a small sample.

XG does not include own goals, so for example Ekpiteta’s og means we don’t get any XG for that.  Had it fallen to Wells he’d have probably got 0.7/0.8 for that…and we shoot up the imaginary table!  Millwall scored from 2 og’s in a 2-2 draw with Swansea a couple of weeks ago too.

I do think it’s fine to look at XG though, but more as a general trend of “are we creating scoring chances / are we conceding chances”.  If looking at individual’s performance, again you can start to see trends.  Championship POTM Estupinan has scored 7 goals versus 6.49 XG, so that’s pretty good.  Delving deeper you can see that he’d scored those 7 goals from his first 12 shots at this level, that’s incredible (in his last 2 games - 0 goals from his last 5 shots).  What it highlights is that he was getting very good chances…and scoring them.  If you watch the highlights on ITV you’ll see two of them were virtually on the goal-line, one of which “just hit him”.  The question becomes will he continue to keep getting gilt-edged chances?

FWIW here’s City’s XG per player.

Along the bottom, total XG this season.

Up the side, average XG per shot.

Size of blob equals goals scored.

Nahki, Chris and Andi get better chances (on average) than Tommy, but Tommy is converting well above his average.

The pic below shows every shot and it’s XG value.  Grey blob - no goal, red diamond - goal, green line - average XG per shot.

What we see is that we generally convert our “big chances” - XG 0.3 or better.

Of our 14 big chances, we’ve scored 9 of them (64%)

Of our 42 lower percentage chances, we’ve scored 6 (14%)

Plus of course Ekipeta’s og!!

 

 

[pongo88]

I prefer real goals 

[chinapig]

8 minutes ago, pongo88 said:

I prefer real goals 

Which are a result of how many chances you create, the quality of those chances and how efficient you are at converting them into actual goals.

Though performance, in football or any business, should not be judged on a single metric.

[cidercity1987]

44 minutes ago, Mr Popodopolous said:

Certainly not a perfect model but some of our more basic underlying numbers feel a bit midtableish to me. Enjoyable football without doubt and the attacking football is great to see but is Conway going to sustain his brilliant form eg? Wells in the best form of his time here...then again a fully fit Semenyo adds a new dimension as we know and Kalas as and when he is fully fit will only help.

On the other hand, BAD refereeing decisions have influenced these numbers.

Even if you disregard the penalty claims as maybe maybe not, the one against at Hull, the non red card for Bennett in the heat and when we were ahead and on top and the Sykes red but Freeman reprieve, all of these negatively impacted us. Not just results wise but statistically too, badly- we were so dominant at 11 v 11 vs Luton for example.

We are three points behind on XG compared to our actual points

Knock out the 0.8 or something XG from Hull's penalty and BANG our XG = actual points

I dont think we are overperforming at all and XG doesn't take into account our strikers are better than many others

[Davefevs]

Thought I’d have a look at Preston - as they are stinking out goals for and against):

Wigan 0-0 Big chances 2 (0.54 / 0.52 both to Riis) first one a 1-on-1 saved, second one ball hits him 2 yards out but he deflects it over the bar.

Hull 0-0 Big chances 1 (0.35 Alan Browne) far post scramble, hits outside of the post

Luton 1-0 Big chances 0 (absolute worldie from Brad Potts, like Weimann’s volley at Sheff Utd, but miles better)

Rotherham 0-0 Big chances 1 (0.43 Liam Lindsay) free header from 5 yards which he gets under, sensing keeper is coming.

Watford 0-0 Big chances 0

Cardiff 0-0 Big chances 0 (total XG for the match 0.15 )

Coventry 1-0 Big chances 1 (0.53 Ledson from which he scored)

Birmingham 0-1 Big chances 0

- summary

big chances 5 scored 1 (20%)

lower chances 86 scored 1 (1%)

34 minutes ago, Davefevs said:

Of our 14 big chances, we’ve scored 9 of them (64%)

Of our 42 lower percentage chances, we’ve scored 6 (14%)

Sorry, calculated ours wrong.

lower chances 72 scored 6 (8%)

PNE have had more a few more shots than us….but 41 (45%) are from outside the box.  City’s 86 shots - 23 from outside the box (27%).  PNE are 21st for average XG per shot, we are 4th.  We create better chances.

 

 

 

[chinapig]

11 minutes ago, Davefevs said:

PNE have had more a few more shots than us….but 41 (45%) are from outside the box.  City’s 86 shots - 23 from outside the box (27%).  PNE are 21st for average XG per shot, we are 4th.  We create better chances.

Which means that despite the mockery box entries do in fact matter!

[Davefevs]

2 minutes ago, chinapig said:

Which means that despite the mockery box entries do in fact matter!

Only if you finish the job off

 

Also, a half-chance is anything but!!!

Edited September 13, 2022 by Davefevs

[ExiledAjax]

1 minute ago, chinapig said:

Which means that despite the mockery box entries do in fact matter!

If they end in a shot yes. If you're just taking the ball into the box, wandering around with it, then losing possession, then not so much.

Shots to Box entry ratio would be what you need.

1 minute ago, Davefevs said:

Only if you finish the job off

Exactly, no point going in, out, in, out, in out, and never getting your shot off.

[chinapig]

16 minutes ago, ExiledAjax said:

If they end in a shot yes. If you're just taking the ball into the box, wandering around with it, then losing possession, then not so much.

Shots to Box entry ratio would be what you need.

Quite, I suppose that was the assumption behind my post.

I remember David Moyes when he was at Man Utd (after a home draw I think) stressing how many crosses they had got in as if that was an end in itself.

Edit: though I'll state the obvious that if you don't get in the box in the first place you won't create high quality chances.

Edited September 13, 2022 by chinapig

[Mr Popodopolous]

1 hour ago, cidercity1987 said:

We are three points behind on XG compared to our actual points

Knock out the 0.8 or something XG from Hull's penalty and BANG our XG = actual points

I dont think we are overperforming at all and XG doesn't take into account our strikers are better than many others

We certainly could have lost at Wigan however, conceded a lot of possession and territory 2nd half  but again the non red card for Bennett.

From memory, it's 0.74 for a penalty so yes draw at Hull?

I'm sure it had us worse off than 2-0 winners in the Huddersfield game but better off than we got in the Sunderland game.

It's not even solely on xG, just don't think we're dominant enough to be considered a top 6 side...yet. We're good yes.

Edited September 13, 2022 by Mr Popodopolous

[CyderInACan]

1 hour ago, pongo88 said:

I prefer real goals 

My preference is athletic bilbao goals 

[Davefevs]

Deviating from the OP, sorry @And Its Smith, but we hear a lot about “front foot football” and “eyes always forward” and stuff like that.  We often make comments like “he always passes backwards” or “his passing stats are down because he’s trying riskier passes”.  I’m not attempting to confirm one way of the other, as there are lots of dynamics at play, e.g. team style.

So I thought I’d have a look over the weekend at which direction do players pass it in.  I’ve initially started with CMs as they are the most likely to have variance over 360 degrees.

@Lrrrbefore you jump on me, yes I’ve recalculated all of Wyscout’s passing stats to account for the fact that they don’t include lateral passes under 12m in their lateral passes per 90 figure (no idea why not )!

So, here goes:

(All championship CMs who’ve played 180 mins or more)

Overall Distribution:

 

If we just take the two red lines, the average percentage of forward passes (of the total passes) is 30.64%.  The average backwards is 19.33%, therefore the average of lateral passes is 50.03%.  If you take a compass 25% is facing North (forwards), 25% south (backwards) and 50% either east or west (lateral / sideways).

Bristol City:

We can see that HNM does indeed make the highest percentage of backwards passes, and Joe Williams the highest percentage of forward passes.  Some of that will of course be down to the positions they play, where they make their passes from, etc…and another example of why data alone doesn’t tell the whole story.

If you zoom in (sorry phone users) you can see the numbers of each pass type - back / lateral / forward / (total).

Josh Brownhill:

Makes more backwards and less forward passes than any of our players, but he is playing a bit higher this season and therefore will be receiving balls back to goal, on the turn, etc.

Swansea:

The pass heavy team.

Basically, lots of sideways passing!!

Not trying to draw too many / any real conclusions other than there is a general trend that the more forward passes you hit, the less backwards you hit. How obvious, although I was surprised that lateral passes were 50%.  Hence why I had to think in terms of a compass.

Who are the two blobs right on the average intersection:

Jamie Allen / Coventry

Alan Browne / PNE

 

 

Edited September 13, 2022 by Davefevs

[OldlandReddies]

I expect us to be mid table after the next two games. Hope City prove me wrong. If they do then we are going places. Two of the toughest away days we will have. Cmon Nige!!! 

[Mr Popodopolous]

Dunno when the xG table will be updated but the individual match rating fwiw up this morning was Norwich 1.40 v Bristol City 1.20.

Which to all intents and purposes is a draw. 1.5 or above do you round up?

Edited September 15, 2022 by Mr Popodopolous

[Hxj]

On 13/09/2022 at 15:04, chinapig said:

Though performance, in football or any business, should not be judged on a single metric.

Unless of course you get promoted or relegated in which case that will be the only metric which counts.

[Davefevs]

1 hour ago, Mr Popodopolous said:

Dunno when the xG table will be updated but the individual match rating fwiw up this morning was Norwich 1.40 v Bristol City 1.20.

Which to all intents and purposes is a draw. 1.5 or above do you round up?

Think E361 bases result on XG differential being less than 0.333.

[ExiledAjax]

10 minutes ago, Davefevs said:

Think E361 bases result on XG differential being less than 0.333.

Yeh he'd call it a draw.

FWIW xG values varied a fair bit last night. Infogol gave it as 1.99 v 1.05 to Norwich, so they saw Norwich as having clearly the better chances.

FotMob gave it as 1.7 v 1.5. so the same differential as E361, but +0.3 to each team.

The other source I use, which I am perennially reviewing my use of, gave it as 1.5 v 1.7. so they actually saw us as being on top.

Overall the average I have from 4 sources is 1.65 for Norwich to 1.35 for us. So a differential of 0.3. 

Close enough as to be unclear that either side generated a better aggregate chance of scoring than the other.

 

[Davefevs]

16 minutes ago, ExiledAjax said:

Yeh he'd call it a draw.

FWIW xG values varied a fair bit last night. Infogol gave it as 1.99 v 1.05 to Norwich, so they saw Norwich as having clearly the better chances.

FotMob gave it as 1.7 v 1.5. so the same differential as E361, but +0.3 to each team.

The other source I use, which I am perennially reviewing my use of, gave it as 1.5 v 1.7. so they actually saw us as being on top.

Overall the average I have from 4 sources is 1.65 for Norwich to 1.35 for us. So a differential of 0.3. 

Close enough as to be unclear that either side generated a better aggregate chance of scoring than the other.

 

Wyscout gave in 2.17 (Norwich) v 1.15 (Us)

Norwich big chances (0.3 XG):

Pukki 8 mins (2nd block)

Pukki 11 mins goal

Pukki 23 mins goal

Pukki 75 mins (saved by Bents face)

City’s:

Dasilva 28 mins (blocked after Wells block)

In general play we’ve restricted them to scraps really, goals from errors and a set-piece (then breakaways).

Still need to improve in certain game situations, but positives from overall play.

[Mr Popodopolous]

1.99 v 1.05 to Norwich for such a well fought game seems like a huge outlier to me- 2.17 vs 1.15 also.

[Davefevs]

14 minutes ago, Mr Popodopolous said:

1.99 v 1.05 to Norwich for such a well fought game seems like a huge outlier to me- 2.17 vs 1.15 also.

Yes and no.  Pukki’s two goals account for 0.8…0.8 avoidable.  That starts to bring parity to the XG total.

My view is that in general play we were pretty good / the better side, but I don’t think we fashioned the same quality of chances as in more recent games.  We were a bit wasteful really.  Sykes had a super game but his final ball rarely found it’s man.  One cross was brilliant…but nobody was there.

Hence why XG isn’t a good use for determining a “result”.

[And Its Smith]

As xG seems to be talked about as how good a team is at chance creation it is interesting that it gives a team credit for a shot that they themselves did nothing to create. Haaland goal last night also interesting in that xG rated that as a better chance than Bellingham’s goal despite popular opinion being that there are hardly any strikers who would finish that ‘chance’ 

[Davefevs]

1 hour ago, And Its Smith said:

As xG seems to be talked about as how good a team is at chance creation it is interesting that it gives a team credit for a shot that they themselves did nothing to create. Haaland goal last night also interesting in that xG rated that as a better chance than Bellingham’s goal despite popular opinion being that there are hardly any strikers who would finish that ‘chance’ 

You need to read up about xGT (threat)…that measures each pass and it’s likelihood that the move will end in a shot!  It’s the new xG!!!

PSxG is interesting…post-shot xG….measures the XG of the shot, e.g. gives the shot a better score for a thunderbastard into the corner, than a tame shot straight at the keeper.  Players like Harry Kane tends to have a PSxG much greater than his XG, ie he finds the corners.

[Mr Popodopolous]

One for Dave and EA before I forget. Anyone else in fact but feel these two may know a lot.

@Davefevs @ExiledAjax

ExP.

Is there such a metric yet invented as Expected Penalties?

Granted it's a bit clunky but penalties per big chance created as an average or in a rudimentary way, penalties to shot or shot on target ratio over time?

Am keen to look at the latter, the more basic one and make some graphs, charts to reflect the numbers.

Edited September 15, 2022 by Mr Popodopolous

[Davefevs]

16 minutes ago, Mr Popodopolous said:

Is there such a metric yet invented as Expected Penalties?

Yes, there is….….for Bristol City it’s 0.00 and for everyone else it’s 1.00

[ExiledAjax]

7 hours ago, Mr Popodopolous said:

One for Dave and EA before I forget. Anyone else in fact but feel these two may know a lot.

@Davefevs @ExiledAjax

ExP.

Is there such a metric yet invented as Expected Penalties?

Granted it's a bit clunky but penalties per big chance created as an average or in a rudimentary way, penalties to shot or shot on target ratio over time?

Am keen to look at the latter, the more basic one and make some graphs, charts to reflect the numbers.

Are you trying to look at expected penalties or are you trying to look at ratios of actual penalties versus shots, shots on target etc?

If you want an xP metric in the mould of xG or xT then you need to first work out what causes penalties to occur, and then measure that. xG works because shots lead directly to goals* and crucially are always attempting to become goals. It's clear what you need to measure, and you can use the record of results from millions of shots to predict the likelihood of any type of future shot becoming a goal.

To do the same for penalties is far more complicated. There's a subjective human element - the referees - plus a far broader range of actions that result in a penalty. You'd have to account for all the various types of foul that can occur, and also for each referee's likelihood of awarding it. You'd have to account for VAR. You have to account for all of these things in all the different areas of the box. By all means give it a go but it's a big ask even for someone like Opta or Prozone to do all of that.

The other way though, ratios of shots etc to penalties awarded. That's easier as those numbers are out there for probably the last 20 seasons or so. @Olé made a good start on that in the big thread we had going on our own penalty quirks recently. Probably best to find that again and buil from there.

*a tiny number of goals do occur without a shot, but it's a negligible number.

[Mr Popodopolous]

Thanks both.

I think I'll try the more basic stuff all told...shots for vs penalties awarded over the last couple of seasons and this. Maybe Shots on Target too and I'll try and make some graphs or charts this time too.

Shots and Shots on Target for can be indicators of attacking threat or intent even if more basic.

[Mr Popodopolous]

Had a quick look at Experimental 3-6-1.

Actually had us actually if anything, shading it at Burnley, 1.4-1.2.

Burnley 1.20 v Bristol City 1.40

Although rounding wise and xG margin for error, that feels like draw.