xG in the top 8
#21
(03-12-2024, 09:58 AM)MrBater Wrote:
(03-12-2024, 09:42 AM)ColliersWoodBaggie Wrote: Goal Difference is the main thing to look at. Not this stupid XG bollocks.

Unless you lose 4 games 1-0 then thump someone 7-0... Then GD is misleading... Then again in the 1-0 losses you could get 3 or 4 xg and not score, so I agree it's one of the daftest stats.

It's like possession, against QPR it showed as 50/50 but we were barely out of our half so that's misleading too.

As the old saying goes you can prove anything with stats.

(03-12-2024, 09:58 AM)AnelkasBeard Wrote: People really need to get better at understanding what xG is showing us, and what Lurker is analysing from that. As ever, it's a stat that helps to corroborate certain theories when coupled with other stats, rather than the be all and end all.

It backs up the fact we're a defensive team first, keeping our shape, not over committing players forward and probing teams with possession and trying to draw them out of their block/defensive shape. We don't ever really create "easy" chances through overloads etc. This is also shown by a low amount of shots on target, and shots in general. It all paints a picture that we can see from watching our play as well.

Sorry, I didn't need stats to know any of that, just watching some games tells you all that.

Right because scientific method just requires you to say a hypothesis without any need to back it up with data or stats. Donut.
Reply
#22
Only top 8, CC out!!!@!!!!!!!
Reply
#23
In terms of xG we are actual 12th.

The 4 best teams in the league? Leicester, Leeds, Southampton and Ipswich. Who has the highest xG? Yep those 4.

Out of the rest we have the biggest discrepancy between our xG and our GF. But we also have the lowest GA by some distance which is buffering our lack of consistency in attack.

CC talks about it a lot in interviews. We had to defend and "suffer" too much and we didn't attack enough. This will be largely down to our injury's, hopefully.
"IMO of course..."
Reply
#24
xG is misused as it’s one indicator of many. I think here all it shows is that we don’t have a clinical finisher in the team which we can see with our own eyes. Equally it shows we have to create more to score, which makes winning harder, especially away from home, which again we can see.

I do think where this analysis is let down like many stats is looking across the season, sample size is normally good, however xG is a time sensitive statistic due to form.

I’d rather look at xG in 10 game periods than across a season. The reason being isn’t simply form, it tracks how a coach can change the system and tactics to mitigate the issue.

EDIT: forgot to mention by looking at xG in isolation (regardless of the correlation to the current top 4, correlation isn’t causation) you can have an effective attacking team with a poor defensive unit. You need to look at a number of indicators.
Reply
#25
You never heard Steve Bruce talking about XG.
Reply
#26
(03-12-2024, 11:23 AM)BurleyBaggie Wrote: xG is misused as it’s one indicator of many. I think here all it shows is that we don’t have a clinical finisher in the team which we can see with our own eyes. Equally it shows we have to create more to score, which makes winning harder, especially away from home, which again we can see.

I do think where this analysis is let down like many stats is looking across the season, sample size is normally good, however xG is a time sensitive statistic due to form.

I’d rather look at xG in 10 game periods than across a season. The reason being isn’t simply form, it tracks how a coach can change the system and tactics to mitigate the issue.

EDIT: forgot to mention by looking at xG in isolation (regardless of the correlation to the current top 4, correlation isn’t causation) you can have an effective attacking team with a poor defensive unit. You need to look at a number of indicators.

Its not the point I was highlighting in the OP, mate. I was highlighting that the majority of our goals come from poor quality of chances as we do not produce enough good quality shooting opportunities. 

We are 16th in the league for number of shots, but we are 5th in terms of average distance of shots. So we are actually trying to play into areas to create good opportunities, but its been a struggle with our attacking injuries. Our patterns of play have improved with the reintroduction of Grady and the addition of Johnston who are both capable of manipulating the ball in tight, difficult areas. Fellows is showing signs of becoming more balanced, tactically aware player as well which bodes really well.
"IMO of course..."
Reply
#27
(03-12-2024, 11:10 AM)AnelkasBeard Wrote:
(03-12-2024, 09:58 AM)MrBater Wrote:
(03-12-2024, 09:42 AM)ColliersWoodBaggie Wrote: Goal Difference is the main thing to look at. Not this stupid XG bollocks.

Unless you lose 4 games 1-0 then thump someone 7-0... Then GD is misleading... Then again in the 1-0 losses you could get 3 or 4 xg and not score, so I agree it's one of the daftest stats.

It's like possession, against QPR it showed as 50/50 but we were barely out of our half so that's misleading too.

As the old saying goes you can prove anything with stats.

(03-12-2024, 09:58 AM)AnelkasBeard Wrote: People really need to get better at understanding what xG is showing us, and what Lurker is analysing from that. As ever, it's a stat that helps to corroborate certain theories when coupled with other stats, rather than the be all and end all.

It backs up the fact we're a defensive team first, keeping our shape, not over committing players forward and probing teams with possession and trying to draw them out of their block/defensive shape. We don't ever really create "easy" chances through overloads etc. This is also shown by a low amount of shots on target, and shots in general. It all paints a picture that we can see from watching our play as well.

Sorry, I didn't need stats to know any of that, just watching some games tells you all that.

Right because scientific method just requires you to say a hypothesis without any need to back it up with data or stats. Donut.

No something's you don't actually need to keep over analysing and can pick up on by watching. In this case the scientific method is just visual evaluation.

How the feck did the great teams in world history get by without needing to invent stats for everything?
Reply
#28
(03-12-2024, 11:45 AM)Lurker#3 Wrote:
(03-12-2024, 11:23 AM)BurleyBaggie Wrote: xG is misused as it’s one indicator of many. I think here all it shows is that we don’t have a clinical finisher in the team which we can see with our own eyes. Equally it shows we have to create more to score, which makes winning harder, especially away from home, which again we can see.

I do think where this analysis is let down like many stats is looking across the season, sample size is normally good, however xG is a time sensitive statistic due to form.

I’d rather look at xG in 10 game periods than across a season. The reason being isn’t simply form, it tracks how a coach can change the system and tactics to mitigate the issue.

EDIT: forgot to mention by looking at xG in isolation (regardless of the correlation to the current top 4, correlation isn’t causation) you can have an effective attacking team with a poor defensive unit. You need to look at a number of indicators.

Its not the point I was highlighting in the OP, mate. I was highlighting that the majority of our goals come from poor quality of chances as we do not produce enough good quality shooting opportunities. 

We are 16th in the league for number of shots, but we are 5th in terms of average distance of shots. So we are actually trying to play into areas to create good opportunities, but its been a struggle with our attacking injuries. Our patterns of play have improved with the reintroduction of Grady and the addition of Johnston who are both capable of manipulating the ball in tight, difficult areas. Fellows is showing signs of becoming more balanced, tactically aware player as well which bodes really well.

I know what you’re trying to say in the OP but you’re mixing correlation with causation and using a specific indicator to evidence some quite nuanced points. You can’t use individual stats in this way. On the UEFA B licence you’re taught how to use statistical analysis and there’s a fantastic exercise that proves Man City (in my course) are the worst team in the league defensively based on tackles won in the final third during key pass phases. The indicatir suggests one thing, however, it doesn’t show that actually the indicators on interceptions and final third errors by opposing players result in fewer shots and lower xGC - because one indicator doesn’t show the press or the player distances in defence etc etc
Reply
#29
(03-12-2024, 12:26 PM)BurleyBaggie Wrote:
(03-12-2024, 11:45 AM)Lurker#3 Wrote:
(03-12-2024, 11:23 AM)BurleyBaggie Wrote: xG is misused as it’s one indicator of many. I think here all it shows is that we don’t have a clinical finisher in the team which we can see with our own eyes. Equally it shows we have to create more to score, which makes winning harder, especially away from home, which again we can see.

I do think where this analysis is let down like many stats is looking across the season, sample size is normally good, however xG is a time sensitive statistic due to form.

I’d rather look at xG in 10 game periods than across a season. The reason being isn’t simply form, it tracks how a coach can change the system and tactics to mitigate the issue.

EDIT: forgot to mention by looking at xG in isolation (regardless of the correlation to the current top 4, correlation isn’t causation) you can have an effective attacking team with a poor defensive unit. You need to look at a number of indicators.

Its not the point I was highlighting in the OP, mate. I was highlighting that the majority of our goals come from poor quality of chances as we do not produce enough good quality shooting opportunities. 

We are 16th in the league for number of shots, but we are 5th in terms of average distance of shots. So we are actually trying to play into areas to create good opportunities, but its been a struggle with our attacking injuries. Our patterns of play have improved with the reintroduction of Grady and the addition of Johnston who are both capable of manipulating the ball in tight, difficult areas. Fellows is showing signs of becoming more balanced, tactically aware player as well which bodes really well.

I know what you’re trying to say in the OP but you’re mixing correlation with causation and using a specific indicator to evidence some quite nuanced points. You can’t use individual stats in this way. On the UEFA B licence you’re taught how to use statistical analysis and there’s a fantastic exercise that proves Man City (in my course) are the worst team in the league defensively based on tackles won in the final third during key pass phases. The indicatir suggests one thing, however, it doesn’t show that actually the indicators on interceptions and final third errors by opposing players result in fewer shots and lower xGC - because one indicator doesn’t show the press or the player distances in defence etc etc

Of course, when there so many variables its hard to nail down definite causation. 

However my point in the OP is far from a correlation causation fallacy. Total xG, xG/individual shot, total number of shots and distance the shots are taken from are correlated to our performance at both ends of the spectrum. All of those factors highlight the quality of chance we produce, and it is all pointing to us not regularly enough creating high quality shooting opportunities. 

If a muppet like me can pull that from a limited data set I am sure the pros at the club can do much more and then its the coaches job to overcome any weaknesses. For the first time in quite some years I have full faith in our manager to do that on a regular basis too.
"IMO of course..."
Reply
#30
Clear example on Sunday of why xG is flawed. The Yokuslu goal: xG suggests he 'should' only score that goal once if he had 100 attempts at it, but if you ask the player he'd be extremely disappointed to miss his timing 99 times out of 100 when the ball falls so perfectly to him to strike it clean and true into the net. I'd argue he'd expect score that goal at least 10 times more than the xG suggests if he had 100 goes at it, purely due to how it fell to him and how he was able to time the shot perfectly due to that. As opposed to the arbitrary distance/angle calculations done by the xG model/algorithm.
Reply


Forum Jump:


Users browsing this thread: