xG is absolutely pointless

[Kenai]

PSG where awful infront of goal and could have played for another 90 minutes. xG doesn’t tell you that at all.

Surely that's exactly what it's telling you?

On a statistically 'average' day, with the chances they had, you'd expect them to have ended up with about four goals.

The fact they actually only scored one rather than four immediately tells you they were absolute garbage in front of goal and weren't converting their chances the way that they should be.

Edit - it also provides insight when averaged out. Taking Southampton as an example, given they're who I follow, it paints a picture that reflects what i've seen from them this season.

Their average xG For is 1.83 per game but the actual goal count is 1.63, indicating the forwards are slightly underperforming and should be a bit more clinical - this definitely reflects the frustrating experience of watching them this season despite good results, not putting games to bed or relying on scrappy last minute goals. Their average xG against is 1.15 per game but the actual is 1.42 conceded, indicating we're also leaking a few goals that we ought not to be, which again ties in with watching them concede quite a few really embarrassing goals that would have had comparatively low xG.

 

[gingergundog]

No, it isn’t. There’s an obvious point to betting odds. A trade off between enticing punters to gamble and losing money, that’s the entire “game”.

Your last point is what I was really getting at. A lot of fans don’t understand much about it so say things after a loss like, “Well, the xG was x for us and y for you, so we should have won”.

The moneyball stuff and statistical analysis is great, I’m unsure what value is derived from putting it out to fans.

Agreed, not sure how many fans are even interested in it.

Fine if you like stats I guess(not like there are a shortage of them) and each to their own etc but personally I find it dull and id rather just enjoy the football(or not in the case of watching UTD).

 

[wesimmo]

BTW, the link I shared shows that xG typically has about a 0.68 correlation to goals scored. So not really an indicator if you go by their analysis.

The problems they highlight include weighting of xG against player ability.

For example the same chance is given the same xG for Haaland as Maguire.

Then you get into weighting the chance on player ability. But those change on the type of chance. I.e. Andy Cole's poacher instinct and touch vs Le Tissier's ability to magic a strike from nowhere (for those of us of an age).

And then you start down the 'Man City will inherently have a higher xG because their players will be rated higher' route.

So, yes, it's AN indicator of a match but by no means the only one.

 

[Hotwired]

Surely that's exactly what it's telling you?

On a statistically 'average' day, with the chances they had, you'd expect them to have ended up with about four goals.

The fact they actually only scored one rather than four immediately tells you they were absolute garbage in front of goal and weren't converting their chances the way that they should be.

Edit - it also provides insight when averaged out. Taking Southampton as an example, given they're who I follow, it paints a picture that reflects what i've seen from them this season.

Their average xG For is 1.83 per game but the actual goal count is 1.63, indicating the forwards are slightly underperforming and should be a bit more clinical - this definitely reflects the frustrating experience of watching them this season despite good results, not putting games to bed or relying on scrappy last minute goals. Their average xG against is 1.15 per game but the actual is 1.42 conceded, indicating we're also leaking a few goals that we ought not to be, which again ties in with watching them concede quite a few really embarrassing goals that would have had comparatively low xG.

You sure that's insight and not just reaching for a stat and abusing it when you feel like it.

What right has anyone to future performance based on previous performance. A single number tells you nothing about any meaningful reasons and is this weird bogeyman of the past when used as a measuring stick for the present and the future.

 

[MartinPrince]

I quite like xG, though as someone that studied Maths and Statistics at Uni that's probably to be expected!

I normally only get time to watch highlights (which often don't show all chances created) so I find xG a more refined and concise metric as to which team created the better chances.

Before xG I would look at shots/shots on target, corners and possession but now xG saves me doing that so much.

In modern times stats have become an integral part of many mainstream sports; you only have to watch the film Moneyball to see how useful they can be.

 

[Kenai]

You sure that's insight and not just reaching for a stat and abusing it when you feel like it.

Yep

What right has anyone to future performance based on previous performance. A single number tells you nothing about any meaningful reasons and is this weird bogeyman of the past when used as a measuring stick for the present and the future.

No idea what this is getting at, xG is a reflective statistic to help assess a performance, not a predictive measure, so don't see the relevance of comments about 'right to future performance'.

 

[BaZ87]

I really don't understand what people find hard about xg. When we say that's a great chance, all xg does is quantify that and tells you just how good a chance it was.

 

[fez]

I really don't understand what people find hard about xg. When we say that's a great chance, all xg does is quantify that and tells you just how good a chance it was.

People can tend to be a little black and white these days. Something is either perfect or awful rather than just being useful and nuanced.

I'm sure that eventually we will have so much information on players, teams etc that we will be able to predict when they will have good games, when they will get injured and what chance a single player has of taking a given chance was pretty good accuracy purely based on all the data.

Probably won't even need to watch the matches

 

[HangTime]

Yeah, I think it's nonsense personally. It's utterly meaningless. You can tell from other stats how a game is going if you aren't watching it live, and if you are, you'll be focussing on the ACTUAL score rather than some made up 'expected' score.

I think it helps summarise 'how the game is going' in a simple stat without having to look at possession/corners/shots/shots on target bla bla. I don't mind it, you just have to take it with a pinch of salt.

 

[toshj]

Yeah, I think it's nonsense personally. It's utterly meaningless. You can tell from other stats how a game is going if you aren't watching it live, and if you are, you'll be focussing on the ACTUAL score rather than some made up 'expected' score.

It's crazy that Bloom and Benham have probably made over £1bn between them on something that's utterly meaningless! They must just be terribly lucky...

 

[Reuben klang]

It's crazy that Bloom and Benham have probably made over £1bn between them on something that's utterly meaningless! They must just be terribly lucky...

Yeh, i’m pretty sure they aren’t simply going sort by xG.

They’ll be using an entire boat load of stats and they know how to read and use then correctly.

 

[toshj]

Yeh, i’m pretty sure they aren’t simply going sort by xG.

They’ll be using an entire boat load of stats and they know how to read and use then correctly.

Of course, they'll have access to loads of stats that us laymen couldn't comprehend. But they're still gaining an edge by using stats to their advantage, of which predictive xG seems a fairly crucial one.

 

[PikeyPriest]

xG can show trends overall. Like all stats the more data points you have the better the quality you can give.

If a players scores 6 goals from 14 xG and 40 games, it would make you question their ability in front of goal if they are getting frequent good quality chances. If a team generate few chances due to ability or tactics, recognising the striker has a high conversion rate (even if overall goals are lower) would have value.

In a single game it can be a poor marker as a team missing a sitter in the 90th minute could give a final score and XG of 2-0 (1.6 - 1.5) and give the impression that the game was close. XG doesn’t take in to account the overall game, it only cares about the scoring chance and some players outperform the expected.

 

[doopydug]

I have always found it really confusing and di9dn't understand how it was measured but I suppose it makes sense in a kind of way