• Dear Guest, Please note that adult content is not permitted on this forum. We have had our Google ads disabled at times due to some posts that were found from some time ago. Please do not post adult content and if you see any already on the forum, please report the post so that we can deal with it. Adult content is allowed in the glory hole - you will have to request permission to access it. Thanks, scara

The Stats Thread

Bishop said:
This is what I'm actually most impressed by. We suddenly started keeping clean sheets, something we just haven't been able to do regularly since 2018. Bodes ready well if we can keep the opposition out, gives us a great base to play from.

Sent from my XQ-BC72 using Fapatalk
Click to expand...

We were defensively solid at times under Mourinho, even for a shorter spell under NES. But always at the cost of being limp and ineffective going forward.

Since the start of march we've only conceded 8 goals in 13 games. Scoring 34 in the same period. Really impressive numbers, close to City and Liverpool even though they have been slightly better.

XG numbers aren't wildly different. Xg at 27.5 and xG against of 9.6 in the same period. Given that both Son and Kane fairly regularly outperform xG and Lloris fairly regularly saves more than average xG against would indicate it should be fairly close to sustainable. If we can keep performing at a similar level over time, or better of course.
 
Just wanted to provide some evidence that the xG model is pretty accurate, based on this season's Premier League stats (using Understat data).

Firstly: Total xG across all teams was 1070. In total 1071 goals were actually scored. So overall the model was off by 1 goal, or 0.1%.

Breaking that down by team, there was a 96% correlation between xG and G:

upload_2022-5-31_14-53-26.png

So whilst xG obviously isn't a perfect measure of chance quality, I think this indicates that it's a pretty decent indicator, especially over multiple games.
 

Attachments

  • upload_2022-5-31_14-50-45.png
    upload_2022-5-31_14-50-45.png
    9.1 KB · Views: 2
elltrev said:
Just wanted to provide some evidence that the xG model is pretty accurate, based on this season's Premier League stats (using Understat data).

Firstly: Total xG across all teams was 1070. In total 1071 goals were actually scored. So overall the model was off by 1 goal, or 0.1%.

Breaking that down by team, there was a 96% correlation between xG and G:

View attachment 14151

So whilst xG obviously isn't a perfect measure of chance quality, I think this indicates that it's a pretty decent indicator, especially over multiple games.
Click to expand...

Not meaning to be a contrarian, but what it probably points to is over a large enough group of games, it's fairly indicative.

I'd suspect that game by game, it's a lot more variable.
 
Raziel said:
What I meant was despite the season view seemingly validating the model (<1% variance), it (in my opinion) still did not validate it as a per game judgement model which is primarily how it used.
Click to expand...
But as said, that's how stats work! It's a measurement tool! Create x amount of x valued chances, and you usually end up with x amount of goals! Sometimes a team/player will score more, sometimes less. But the presicion of the model remains true!

What's your alternative that is better?
 
Kandi1977 said:
But as said, that's how stats work! It's a measurement tool! Create x amount of x valued chances, and you usually end up with x amount of goals! Sometimes a team/player will score more, sometimes less. But the presicion of the model remains true!

What's your alternative that is better?
Click to expand...

I know how stat's work (I understand concepts of sample size), however lets be clear (as we seem to be talking around each other)

- As a season evaluation tool it's probably quite useful, as the closeness of the results/predicted shows (this is the place it will precise)
- What I'm saying is I'm quite sure if you took that and applied to every individual game, the variance from game to game will probably make it nothing more than a novelty
- However people/fans/media seem to use it as a true measurement of how a game went, when like every stat (e.g. possession), it's often useless without some kind of context, e.g. a team can have brick possession because a/ they couldn't get/hold on to the ball or b/ because they sat in a mid/low block. kept the opposition in front of them and nullified the threat.

Not my gig to make an alternative, I just made a comment that what was being portrayed as an impossibly precise <1% accuracy, is likely only good in the context of a season evaluation not a per game.
 
Raziel said:
I know how stat's work (I understand concepts of sample size), however lets be clear (as we seem to be talking around each other)

- As a season evaluation tool it's probably quite useful, as the closeness of the results/predicted shows (this is the place it will precise)
- What I'm saying is I'm quite sure if you took that and applied to every individual game, the variance from game to game will probably make it nothing more than a novelty
- However people/fans/media seem to use it as a true measurement of how a game went, when like every stat (e.g. possession), it's often useless without some kind of context, e.g. a team can have brick possession because a/ they couldn't get/hold on to the ball or b/ because they sat in a mid/low block. kept the opposition in front of them and nullified the threat.

Not my gig to make an alternative, I just made a comment that what was being portrayed as an impossibly precise <1% accuracy, is likely only good in the context of a season evaluation not a per game.
Click to expand...
xG gives you a very good idea of which team had the best chances, and whether a result reflects the game, at least more so than any other stats.
 
Little Luka said:
View attachment 14446

Just a table for the hell of it.
Click to expand...
200.gif
 
Kandi1977 said:
Surely only Arsenal can get max possible points???
Click to expand...

Yes, I think my badly worded table has deceived you. 114 as max is true for all teams before the season starts. It's really just there to show what the max for a whole season is, but it really isn't needed. :)
 
Little Luka said:
Yes, I think my badly worded table has deceived you. 114 as max is true for all teams before the season starts. It's really just there to show what the max for a whole season is, but it really isn't needed. :)
Click to expand...
I would have thought that column was there to show what is the maximum number of points a team could get from the current position, ie we can now get 112.
 
metalgear said:
https://twitter.com/x/status/1562028783615574016

Describes how we play. Perhaps it's about denying opponents xg positions (closer to goal) and otherwise in attack?

Sent from my SM-T865 using Fapatalk
Click to expand...

We do sit back a fair bit, but defend well when we do. Low xG, long shots, average crosses will happen.

On the flip side we're rather good at counter attacking, situations where getting closer to goal, higher xG chances is easier.

Still a bit surprising. May be a sample size thing, will be interesting to see if it continues in a similar way.
 
metalgear said:
https://twitter.com/x/status/1562028783615574016

Describes how we play. Perhaps it's about denying opponents xg positions (closer to goal) and otherwise in attack?

Sent from my SM-T865 using Fapatalk
Click to expand...

Every teams goal defensively is to deny opponents good goal scoring opportunities, i think that stat just shows we're defending well.

The attacking one is interesting but with only 3 games played it doesn't really tell us anything about whether it is by design or not
 
You must log in or register to reply here.
Back