Main Discussion - Norwich City

Topic has 9 replies.
Print Search Sort Posts:

Interesting...

But is it meaningful?



Stattos, over to you...

   Report  

Re: Interesting...

Their table seems to do little more than inherently undermine the use of ‘expected goals/points’ as a predictor of reality.

If we’re really talking statistically, each team would have to play each other many more times than they do each season, in order to approach a valid sample for making predictions. By that point the season would be over anyway...

Boringly, it all suggests you may as well stick to the actual league table.
   Report  

Re: Interesting...

Xg isn't supposed to be a predictor though- it is often misused in that way.

It is there to analyse issues but also these things usually regress to the mean over time.

I remember the former head of analytics at Brentford mentioned us being in a false position when we were top under Neil- we were giving up chances at a rate that wasn't sustainable and it proved to be true. Similarly with Ipswich's strong start last season, they were taking chances at a rate that was unlikely to last and fell away when that regressed over time.

This seems to suggest we're pretty much where we should be for goals but that basically we're letting teams score too much from half chances. For me that points maybe to a keeper not saving shots he should be.
   Report  

Re: Interesting...

As King says, as an analytical tool for a club to use internally, it’s probably useful.

If you’ve had 20 xgoals but only scored 6, it suggests you need to buy a better striker. If you’ve got a low xgoals tally which matches your actual goals tally, you might have a top quality poacher but nobody in midfield to create the chances.

However it’s limited in its usefulness when comparing clubs against each other. Football still ultimately comes down to the individuals; if we are below our expected points total whilst other clubs are above theirs it suggests they have better/more clinical individuals than us. They make something out of little, we don’t.
   Report  

Re: Interesting...

Agreed Aggy and King. Gx has some usefulness in context. I found the table interesting and a bit of fun.

The article was tagged “what to expect after the international break”, so they’re selling its predictive capacity. Sillies.

Still, some of those predicted to do better will actually do so... hope to the God of Stats it’s us. Of course if we do, it will be because we play better and finish our chances, and get more of the rub of the yellow & green.
   Report  

Re: Interesting...

Reading the article I don't think they really are- they are just pointing who is over and underperforming their xg and suggesting what *should* happen as things regress to the mean.
   Report  

Re: Interesting...

I wouldn't hold your hopes up Gen.

Sorry I can't do a link today but if you look back to "what to expect after the international break" from last October the actual table, admittedly after 11 games, was more accurate than what they were "expecting".

I couldn't find one for September last year. The league table from last September was not so accurate with more anomolies than the October one.
   Report  

Re: Interesting...

 king canary wrote:
Reading the article I don't think they really are- they are just pointing who is over and underperforming their xg and suggesting what *should* happen as things regress to the mean.


Thanks NN - very interesting. I largely agree with your analysis, KC. Indeed, the table supports what a lot of people have said in threads - e.g. over our defensive set piece "issues."

Hopefully that will regress to the mean, but as I am sure you know, this will only be the case if it is a random factor - if we have a definite weakness (e.g poor defensive technique/ bad goalkeeper) our Goals against will remain worse than our expected goals against.


   Report  

Re: Interesting...

Infogol could have done that analysis on our xG at virtually any point last season and they would still have been waiting for us to revert to the mean when the season ended! Our conversion rate (GF/xG) stayed consistently around 0.8 throughout. 
   Report  

Re: Interesting...

Aha.. that doesn’t surprise me, Nutty. Good comparison.


   Report