Xg stats

[It's Character Forming 1,160]

1 hour ago, Thirsty Lizard said:

You beat me to it ICF. The poster in question was Westcoast Canary. We had a very long thread about it - even in February he was still maintaining till he was blue in the face that we were outperforming our XG for goals scored and that we would probably drop away and finish 4th. It was impossible to get through to him that Teemu Pukki in particular was a better than average Championship striker. (Remember the XG stats are based on what an average   striker at that level would be expected to convert. 

Ah, I think I missed that, I thought he'd given up before then !

 

My favourite quote from him was around Christmas when he said something like "No sign of Leeds imploding yet."  Which was true at the time

[It's Character Forming 1,160]

12 minutes ago, Upo said:

So much good stuff here from Character Forming. I highlighted major points I wanted to comment on in aggregate, not necessarily specifically.

Being a good striker isn't mainly about being better at taking advantage of any given situation. Being a good striker is about creating those situations that can be taken advantage of. A great striker should not only outperform his xGs, but he should create the kind of xGs for himself that pretty much anybody could score from.

XGs are "flawed" in that they don't hold any more information than what is put in, and it always is up to discussion what is relevant to put in. The quality of the chance.. Yes, 1000 players may have scored 30% from that position, but what if there were 3 defenders and a goalkeeper in between...You probably should count that as relevant! But what if two defenders were flat on their **** and GK blood pouring from a cut above his eye? Maybe relevant...if you can put those to numbers? Well, what if the striker was left by his girlfriend the day before and couldn't sleep whole night? We know that's relevant, but let's see anyone put a number on it...

If you keep putting information in the stat, you're going to get a whole lot of noise in addition to the signal. I heard Messi pukes before games. So it's normal for him. If Van Dijk puked before game, you'd worry he has food poisoning. Out of 1000 games, Van Dijk has food poisoning maybe once or twice. Messi pukes 100 times. You need 1000 years to get a sample size big enough. But anyone with brain will realize Liverpool is going to struggle a bit with a puking Van Dijk and if Messi pukes, he'll probably do a hattrick.

Over time all things equal, random fluctuations balance. But long term might mean - like you said - longer than a single football season. And by that time you'd have a new coach and the abnormal stats persist and now you're thinking whether those stats were real all along, or whether the new guy has improved the team to match the abnormal stats, or whether the new guy is also just "lucky."

It comforts me to know that what an absolute fertile mess they're trying to put the figures on. There is space for the magic. The speculation. The frustrations and joy, that all make football the beautiful game. There is no "solving" it. You will only find more interesting mysteries.

Very true - the thing that struck me about Baseball from reading Moneyball is that Baseball is a brilliant sport for statisticians.  One of the things they did in Moneyball was use a grid to analyse where the ball was being hit on the baseball field and the outcome for the batting team.  They then fed in the results of this for zillions of games, which allowed them to give a very accurate view on the expected outcome if you hit the ball to a specific location, i.e. stripping out the effect of errors or an outstanding bit of fielding etc on the specific play. 

 

This is what xG is trying to achieve, but as you've pointed out, football has so many more variables to be taken into account, to get an xG which is as good as for baseball is much harder.

 

Having said this, I can see why DF finds xG a useful tool because of the vagaries of finishing in football, so for the manager the amount of xG you're creating is important.

[Aggy 752]

On 05/02/2020 at 14:14, lake district canary said:

So looking at all these posts it looks as if my view is a fair one. Xgs, unless you understand the context and all the variables, are a bit of a waste of time and the majority of people that quote them and those that hear them being quoted have little or no understanding of them, making them fairly pointless to the majority of people.

Can’t this be said about literally any fact?

[Hillhead 114]

On 05/02/2020 at 14:14, lake district canary said:

So looking at all these posts it looks as if my view is a fair one. Xgs, unless you understand the context and all the variables, are a bit of a waste of time and the majority of people that quote them and those that hear them being quoted have little or no understanding of them, making them fairly pointless to the majority of people.   I suppose what is slightly quaint, is that people quote them as if they do understand them when they plainly don't - and others nod sagely as if they also understand them, leading to a sort of communal delusion that they think they know something of value, but really don't.

 

I see where you are coming from but it does seem you're trying to dismiss them based on what you believe is a majority of people misunderstanding them.

As stated above a higher xG than you'd expect for a teams position could indicate a team being poor at putting chances away as opposed to just bad luck.

There was one about goalkeepers a few weeks back, from what I could gather (and I may well be mistaken) it seemed to only count shots on target as opposed to any science into how good the shot was. 

[It's Character Forming 1,160]

15 hours ago, Hillhead said:

I see where you are coming from but it does seem you're trying to dismiss them based on what you believe is a majority of people misunderstanding them.

As stated above a higher xG than you'd expect for a teams position could indicate a team being poor at putting chances away as opposed to just bad luck.

There was one about goalkeepers a few weeks back, from what I could gather (and I may well be mistaken) it seemed to only count shots on target as opposed to any science into how good the shot was. 

I've highlighted in bold they key disagreement I had with the poster (westcoastcanary) who quoted these stats a lot.

 

If the xG is different from the actual goals scored or conceded by a team, the xG stat in itself tells you nothing at all about why there is a difference.  It may be good/bad luck, good/bad performance, or some other factor not taken account of by the stat.  You then have to analyse why there's a difference to come to a view on whether it's good/bad luck or whatever.

 

WCC always assumed our xG against being less than the goals we conceded was due to "bad defending" (which was likely to continue) whereas when our xG for was less than the actual goals we scored, this was basically good luck which couldn't be expected to carry on.  Sort of "heads I win, tails you lose" thinking.  My problem was that he never put forward any reasons why.