United and xG (now that Ole is gone will things change?)

sullydnl

Ross Kemp's caf ID
Joined
Sep 13, 2012
Messages
34,063
I’d be interested to know how often the team with the lower xG in a game still wins. I get the feeling it could be more often than one might instinctively think.
The figure I saw before is that just under 81% of PL games were won by the team with the highest xG. Though what particular model that was and how many games they looked at, I'm not sure.

Which (heavily context dependent though even any look at xG in a single game should be) would still make it a better indicator in one-off games than the amount of shots, for example. So if someone was arguing that team x should have won because they had more shots, pointing to them also having the lower xG could be a reasonable point depending on the context of the game.

In general though the above are correct, it's not a stat designed to be used in one-off games. So heavy caveats apply if you try to do so.
 
Last edited:

largelyworried

Full Member
Joined
Feb 10, 2021
Messages
2,101
The figure I saw before is that just under 81% of PL games were won by the team with the highest xG. Though what particular model that was and how many games they looked at, I'm not sure.

Which (heavily context dependent though even any look at xG in a single game should be) would still make it a better indicator in one-off games than the amount of shots, for example. So if someone was arguing that team x should have won because they had more shots, pointing to them also having the lower xG could be a reasonable point depending on the context of the game.

In general though the above are correct, it's not a stat designed to be used in one-off games. So heavy caveats apply if you try to do so.
What got me thinking was wondering what stat was the best indicator of a result. For example if I knew nothing about a game other than what the possession stats were and had to guess who won, I’d guess the team with more possession. But particularly in midtable games I bet that doesn’t hold true for very long. Is there a better indictor than xG on a single match basis to predict who won?
 

anant

Correctly predicted Italy to win Euro 2020
Joined
Feb 28, 2015
Messages
8,258
What got me thinking was wondering what stat was the best indicator of a result. For example if I knew nothing about a game other than what the possession stats were and had to guess who won, I’d guess the team with more possession. But particularly in midtable games I bet that doesn’t hold true for very long. Is there a better indictor than xG on a single match basis to predict who won?
Have a feeling it would a combination of big chances and xT
 

Noodle

New Member
Newbie
Joined
Oct 24, 2019
Messages
323
Supports
Chelsea
Have a feeling it would a combination of big chances and xT
I bet there is an equation to be done whereby you combine possession, xG and big chances and get the right winner >95% of the time
 
Joined
Jan 1, 2021
Messages
872
As for the general claim that plenty of teams have based entire seasons and won titles on the strategy of having lower xG than the other team, I'd love to see a few examples. Because that sounds... completely wrong.
I don't think anyone's gotten back with an example of this yet. Just a lot of jokes about fans going into meltdown for winning individual games but losing the XG.

Still waiting!
 

bosnian_red

Worst scout to ever exist
Joined
Aug 13, 2011
Messages
57,577
Location
Canada
There's nothing lucky about winning games while having lower xG than the opponent. Plenty of teams literally plan entire seasons based on that strategy and have come out with league wins, let alone CLs or WCs. Chelsea won back to back leagues by parking the bus and playing 2 DMs with a flat back 4 and they'd win those leagues 100 times over if it were repeated. 'Smaller probability'.

Imagine if this lot was around after the CL semis in 07/08 when we defended for pretty much the entire second leg won 1-0 on aggregate from a 30 yarder rocket from Scholes. They'd have probably cried themselves to sleep the night we got to the final.
What? No team goes out with a gameplan to let the opponent get quality chances and to limit yourself to only poor chances. That's essentially essentially your suggesting. Every single team makes a game plan about how to create as many good chances as possible, and how to limit the opponents to only half chances at best. There is literally no stat that measure how effective a team has been at executing their game plan than xG. Teams can plan to park the bus and limit to shots from range. A team can have 30 shots from range and 70% possession and still have below 1 xG, while the other team can have 4 shots and 30% possession and end up with over 1 xG. That would be a successful game plan. XG is a sum of the quality of chances created during a game. That's it. Over time it's been proven to be a very valuable stat to read into, and all the little flukes that happen over a couple of games tend to balance out.

You seem to equate defending or parking the bus to a lower xG which is flat out incorrect.

An interesting note - people always bring up the 1999 Champions League final that United played like shit. Well the xG for that game has us at 2.26 to Bayerns 1.54. People will bring up how they hit the post twice, as they would for Chelsea in 2008, except both of those were good shots, not good chances. In my opinion, a team is unlucky if they lose from a couple of shots from range and hitting the post from a shot from range is usually just something you can disregard. However a team is lucky if they give away big chances and survive on the night. Which is what PSG did yesterday. They created very little and no big chances, City created plenty. United in those CL finals created a good chunk of big chances and conceded few. Hence we were deserving winners in both.
 

bosnian_red

Worst scout to ever exist
Joined
Aug 13, 2011
Messages
57,577
Location
Canada
In PL last season, if we classify the games where xG difference of =<0.3 as a draw, the xG results and actual results were same in 212 games (55%) and there was an undeserving winner in 41 games (11%)

If we classify the games where xG difference of =<0.5 as a draw, the xG results and actual results were same in 199 games (52%), and there was an undeserving winner in 29 games (8%).

So, on a game by game basis, xG isn't that great a model like a lot of people have mentioned earlier
Even on a game by game basis it's not bad, those are good numbers. 10% of the time there is an undeserving winner for an individual game? For a game like football where there is an average of under 3 goals per game, that's not bad at all. So if a team is set up well to outplay their opponents, they'll on average only get an upset 10% of the time. And I'm sure if you look at xG differences of 1 and higher for an individual game, there'd be a much higher success rate too.

Nothing surprising about it of course, doesn't take a genius to say that if you create better chances than the opposition, you will win more often than not. That's literally the basics of football, it's just quantified yet some people are so resistant to something so simple.
 
Joined
Jan 1, 2021
Messages
872
What? No team goes out with a gameplan to let the opponent get quality chances and to limit yourself to only poor chances. That's essentially essentially your suggesting. Every single team makes a game plan about how to create as many good chances as possible, and how to limit the opponents to only half chances at best. There is literally no stat that measure how effective a team has been at executing their game plan than xG. Teams can plan to park the bus and limit to shots from range. A team can have 30 shots from range and 70% possession and still have below 1 xG, while the other team can have 4 shots and 30% possession and end up with over 1 xG. That would be a successful game plan. XG is a sum of the quality of chances created during a game. That's it. Over time it's been proven to be a very valuable stat to read into, and all the little flukes that happen over a couple of games tend to balance out.

You seem to equate defending or parking the bus to a lower xG which is flat out incorrect.
Precisely.

Said the same thing in my response to this post.

The usual counter-argument to your example above is "yeah, well, if the team with 30 shots racks up an XG of 2 just by taking potshot after potshot, they haven't actually created anything but the stat still says they 'should' have scored twice". Yes, and that's an example of the stat telling you exactly what it says: you can let the other team have 30 shots a game and keep a clean sheet now and then. If you let them have 30 shots a game in the long run, you will concede more often than not.

I have no idea why you always get people rushing to dismiss XG with the usual "lolstats" stuff every time it comes up in a discussion. It's obviously useful - nobody's saying it's literal gospel, the most extreme viewpoints are invariably from the likes of that Zen86 chap who's posted nothing of any substance in the entire thread.
 
Last edited:

Siorac

Full Member
Joined
Sep 1, 2010
Messages
23,789
Even on a game by game basis it's not bad, those are good numbers. 10% of the time there is an undeserving winner for an individual game? For a game like football where there is an average of under 3 goals per game, that's not bad at all. So if a team is set up well to outplay their opponents, they'll on average only get an upset 10% of the time. And I'm sure if you look at xG differences of 1 and higher for an individual game, there'd be a much higher success rate too.

Nothing surprising about it of course, doesn't take a genius to say that if you create better chances than the opposition, you will win more often than not. That's literally the basics of football, it's just quantified yet some people are so resistant to something so simple.
I was going to write something similar but you said it better.

The resistance to using stats to better understand football is something I'll never really understand.
 

bosnian_red

Worst scout to ever exist
Joined
Aug 13, 2011
Messages
57,577
Location
Canada
I was going to write something similar but you said it better.

The resistance to using stats to better understand football is something I'll never really understand.
Yeah. Football in its nature is a sport with a lot of variance or luck that plays a part, but xG is obviously a stat that tells a better story than shots, shots on target or possession, or any other stat on any given game. Like all stats you have to use them to help explain things, not point at xG and nothing else. Some are so anti using any analytics as if it will diminish their experience of watching football.
 

bosnian_red

Worst scout to ever exist
Joined
Aug 13, 2011
Messages
57,577
Location
Canada
Differing views on a football forum, shocking.
But what is it that you don't like. Seriously. It's an analytical stat. How can somebody have an emotional opinion about a statistic? Do you feel like that when somebody says "we had more shots than the opposition"? Like... what is there to be annoyed about (other than some people who just use it completely incorrectly)?
 
Joined
Jan 1, 2021
Messages
872
"Stats should f*ck off and nobody gives a shit about them except smug journos" is a rather extreme "view", I'd say.

Like, what's the basis for it? Usually just get some variation of "you don't need stats, just watch the game". To which ... nobody said it had to be one or the other?
 

anant

Correctly predicted Italy to win Euro 2020
Joined
Feb 28, 2015
Messages
8,258
Even on a game by game basis it's not bad, those are good numbers. 10% of the time there is an undeserving winner for an individual game? For a game like football where there is an average of under 3 goals per game, that's not bad at all. So if a team is set up well to outplay their opponents, they'll on average only get an upset 10% of the time. And I'm sure if you look at xG differences of 1 and higher for an individual game, there'd be a much higher success rate too.

Nothing surprising about it of course, doesn't take a genius to say that if you create better chances than the opposition, you will win more often than not. That's literally the basics of football, it's just quantified yet some people are so resistant to something so simple.
But we're talking about success rate - and that's 55%.

I love xG but on a game by game basis, drawing conclusions about a team is not that wise as there are a lot more variables at play. Think about it this way - as per xG, we deserved to win vs Villa but if you look at quality of chances we created (not including ifs like Greenwod passing to Bruno, etc.), I doubt anyone would say that we had higher quality chances. Of the best 3 chances in the game, they had 2, we had 1 (the penalty).

EDIT: Add to that, on a game by game basis, the match state plays a crucial role as well. If you're the underdog and somehow leading 1-0 against a top side, there's a good chance that you'll probably end up with a high quality chance in the closing stages of the game thereby boosting the final number.
xG as a predictive tool is a great metric, but not so great if you want to describe a team's performance on a given day (unless of course the xG at the end of game is something like 2 vs 0.4 or something)
 
Last edited:
Joined
Nov 11, 2015
Messages
2,596
Location
Whalley Range
But what is it that you don't like. Seriously. It's an analytical stat. How can somebody have an emotional opinion about a statistic? Do you feel like that when somebody says "we had more shots than the opposition"? Like... what is there to be annoyed about (other than some people who just use it completely incorrectly)?
Football cannot be boiled down to the xg data model. Journalists with no more knowledge or insight that anyone on this forum act like they have the definitive and only way to analyse football.

They use it to boost their self importance and try to put some weight on their pure speculation.

The big proponenets of XG on certain podcasts and certain papers use XG when it suits their narrative and ignore it when it doesn't.

I don't feel xg has brought about any better or more accurate coverage of football than there was previously.

Ultimately if xg was followed Alex Ferguson would have been sacked. Klopp would have been sacked. Because data like xg cannot quanify all the factors that go into football.
 

Daydreamer

Full Member
Joined
Feb 8, 2014
Messages
1,276
Supports
Arsenal
I find the resistance to xG kinda weird. It's a pretty harmless stat. My only issue with it is that it only measure chances that involve a shot. So instances like:
  • A forward being put clean through who but takes a poor touch allowing the defence to regroup
  • A cutback to the penalty spot that is only just toe-poked away by a desperate defender
  • A rebound in the six yard box that narrowly evades the strikers control
aren't covered by xG, even though they would be considered 'chances' and would most likely be part of the match highlights. But all stats have their drawbacks. I think xG is one of the best due to its simplicity and applicability.

It also seems incredibly accurate over time. Brighton massively underperformed their xG last season. Therefore their current good form is not a surprise - they create a lot of good chances and eventually that has translated into points.

I'm curious, if you're not a fan of xG - why is that? Genuine question, not being aggro. I just personally find it a weird thing to be against.
 
Joined
Jan 1, 2021
Messages
872
Ultimately if xg was followed Alex Ferguson would have been sacked. Klopp would have been sacked
We have absolutely no justification whatsoever for that first sentence, because xG wasn't calculated when Fergie's job was in danger. And we have possibly even less for the second, because Klopp's Liverpool teams did very well on the metrics in his first couple of seasons when they won nothing.

Genuine question: do you know what XG is?

I find the resistance to xG kinda weird. It's a pretty harmless stat. My only issue with it is that it only measure chances that involve a shot. So instances like:
  • A forward being put clean through who but takes a poor touch allowing the defence to regroup
  • A cutback to the penalty spot that is only just toe-poked away by a desperate defender
  • A rebound in the six yard box that narrowly evades the strikers control
These are good points, though I'd only focus on the first as an actual chance that isn't captured by the data. The second is really just good defending, and the third is only a chance if the attacker does control the ball.
 
Last edited:

bosnian_red

Worst scout to ever exist
Joined
Aug 13, 2011
Messages
57,577
Location
Canada
Football cannot be boiled down to the xg data model. Journalists with no more knowledge or insight that anyone on this forum act like they have the definitive and only way to analyse football.

They use it to boost their self importance and try to put some weight on their pure speculation.

The big proponenets of XG on certain podcasts and certain papers use XG when it suits their narrative and ignore it when it doesn't.

I don't feel xg has brought about any better or more accurate coverage of football than there was previously.

Ultimately if xg was followed Alex Ferguson would have been sacked. Klopp would have been sacked. Because data like xg cannot quanify all the factors that go into football.
The problem is you're assuming people look at it and it's the most important thing in the world. It's a very important stat - for sure. It essentially just quantifies the quality if chances per game, which over time gives a trend. That's a very normal thing that I assure you, every team keeps track of regardless.

Some people are dickheads yes but it doesn't make the stat itself any less valuable. It's a more relevant stat than standard shots or possession. But at the end of the day, it's a stat.

Your last paragraph is entirely meaningless and without reason. Are you saying that Sir Alex's teams were shit but by luck we won the title every other year? We were the best team in the league and the xG would reflect that, if it was around. It reflects that Man City were a top team the year Pep took over, but that they had a decent chunk of bad luck and they finished 4th. Fast forward a year and they walked to the league title. Over a long period of time it provides a stat to quantify if a team was truly lucky or unlucky. But it's just a stat so it should be used as one, just a piece of analysis to help give the real story.
 
Joined
Nov 11, 2015
Messages
2,596
Location
Whalley Range
The problem is you're assuming people look at it and it's the most important thing in the world. It's a very important stat - for sure. It essentially just quantifies the quality if chances per game, which over time gives a trend. That's a very normal thing that I assure you, every team keeps track of regardless.

Some people are dickheads yes but it doesn't make the stat itself any less valuable. It's a more relevant stat than standard shots or possession. But at the end of the day, it's a stat.

Your last paragraph is entirely meaningless and without reason. Are you saying that Sir Alex's teams were shit but by luck we won the title every other year? We were the best team in the league and the xG would reflect that, if it was around. It reflects that Man City were a top team the year Pep took over, but that they had a decent chunk of bad luck and they finished 4th. Fast forward a year and they walked to the league title. Over a long period of time it provides a stat to quantify if a team was truly lucky or unlucky. But it's just a stat so it should be used as one, just a piece of analysis to help give the real story.
Alex Ferguson's Manchester United teams were shit in the first 4 years. The xg would have been bad - xg tells a small picture of what's actually going on.

It's an interesting tool, but the way it's hyped up as the definitive way to analyse football is annoying. I think it's limited and xg hasn't improved football analysis imo.

Certain journos who love it use it to push their narratives. It's hilarious listening to Totally Football etc. And the wankers on there use xg to hammer Ole, but Arteta doesn't get reduced to a limited data set because he apparently has a plan...
 

Swedish_Plumber

Full Member
Joined
Jan 23, 2021
Messages
4,847
Location
Edinburgh
I don’t think we should have to deal with the 8 games of amazing form followed by the 5/6 games of shite anymore. It’s been 3 years and he’s been backed to the hilt so far. Not gonna win anything with that as your status quo.
He’s done brilliantly to rebuild a broken squad, but he’s not imprinted any sort of identity for longer than those 8 game before form drops off a cliff.
 

Abraxas

Full Member
Joined
Mar 6, 2021
Messages
6,023
Hard to understand what the fuss about it is. What manager and side would not want to be creating better and more chances and conceding less to their opponents? That is the essence of football and it is distilled into a number.

Completely pointless using for one fixture and probably even a handful but if you have a decent sample size it is almost certainly going to inform an analytical team in terms of a very high level overview of how the team is performing. You then have to drill into why you have the numbers you do and that's the job of a great manager, and maybe more precise statistics.
 
Joined
Nov 11, 2015
Messages
2,596
Location
Whalley Range
So it wouldn't have been "following xG" that got Fergie sacked, then. It would have been the results.

Sounding more and more like you don't actually know what xG is.
No ferguson had decent players that had done better under the previous manager, he also spend a lot to bolster the team. The performances were shit and didn't reflect what it was capable of.

I was replying to someone else's comment anyway. Your previous reply to my initial views on xg wasn't worth replying to.
 

Daydreamer

Full Member
Joined
Feb 8, 2014
Messages
1,276
Supports
Arsenal
These are good points, though I'd only focus on the first as an actual chance that isn't captured by the data. The second is really just good defending, and the third is only a chance if the attacker does control the ball.
This is illustrating my point. A 'chance' is defined as period of play culminating in a shot. Here's an example:
  • Pogba dissects the defence with a brilliant through ball to Lingard.
  • Lingard now has a simple five yard square ball to an unmarked Ronaldo for a tap-in.
  • However Lingard plays it slightly behind Ronaldo and the chance is lost.
Except... according to xG there was no chance - because there was no shot. But anyone who's ever watched a game of football knows that United have wasted a golden chance to score.

Another obvious example is someone clean through who is cynically taken out by the last man. The referee's report would say that he was sent off for 'denying a goal-scoring opportunity'. Yet xG would say... nothing

Stats require consistent definitions to be meaningful so we have decide what we're calling a chance. I'm not even saying I can come up with a better definition. But it's limited by it's definition of a chance due to the nature of football as a low-scoring game.

Still, it's a decent proxy for chance creation.
 

Skills

Snitch
Joined
Jan 17, 2012
Messages
42,010
This is illustrating my point. A 'chance' is defined as period of play culminating in a shot. Here's an example:
  • Pogba dissects the defence with a brilliant through ball to Lingard.
  • Lingard now has a simple five yard square ball to an unmarked Ronaldo for a tap-in.
  • However Lingard plays it slightly behind Ronaldo and the chance is lost.
Except... according to xG there was no chance - because there was no shot. But anyone who's ever watched a game of football knows that United have wasted a golden chance to score.

Another obvious example is someone clean through who is cynically taken out by the last man. The referee's report would say that he was sent off for 'denying a goal-scoring opportunity'. Yet xG would say... nothing

Stats require consistent definitions to be meaningful so we have decide what we're calling a chance. I'm not even saying I can come up with a better definition. But it's limited by it's definition of a chance due to the nature of football as a low-scoring game.

Still, it's a decent proxy for chance creation.
Statistically it's actually not a golden chance though. Because multiple things need to happen for the goal to actually materialise - one of the attackers needs to play and execute the correct pass at the correct time, and then the player in the goal scoring position needs to make sure he's in the correct position to receive it and then score it.

The second isn't a chance because the foul stopped the chance from materialising. Unless it's a penalty or a shot from free kick range.
 
Joined
Jan 1, 2021
Messages
872
Your previous reply to my initial views on xg wasn't worth replying to.
Please feel free to read it again and come back with a genuine response to the Klopp bit!

I'll understand if you don't, though. And while I wait, here's some more on the last post that you did (somewhat reluctantly) deem worth replying to:

No ferguson had decent players that had done better under the previous manager, he also spend a lot to bolster the team. The performances were shit and didn't reflect what it was capable of.
So Fergie spent money, had players that had performed better under the previous manager, and didn't have performances or results to show for it.

What does any of this have to do with xG? Literally nothing. You're making no sense, and the more you post the more I'm convinced you have no idea what XG is except that you've read Michael Cox or some other slightly self-absorbed journalist (that also doesn't like Solskjaer much) write about it, and you've decided you're not having it.
 
Last edited:

bosnian_red

Worst scout to ever exist
Joined
Aug 13, 2011
Messages
57,577
Location
Canada
:drool:

A stat that says what everyone's eyes say - appalling performance that we got away with due to excellent work from De Gea. A result that gives us a lifeline but a performance that gives us bigger worries.
 

bosnian_red

Worst scout to ever exist
Joined
Aug 13, 2011
Messages
57,577
Location
Canada
No ferguson had decent players that had done better under the previous manager, he also spend a lot to bolster the team. The performances were shit and didn't reflect what it was capable of.

I was replying to someone else's comment anyway. Your previous reply to my initial views on xg wasn't worth replying to.
Yeah, but it was a different time and Sir Alex is a massive exception. He also had a very good record before joining United. Based on the results which is what a manager gets judged off of at the end of the day, he could've been sacked and there wouldn't have been huge complaints at the time. He stayed because of other reasons and just faith and it worked out. Fail to see the relevance of that.

xG is nothing a stat that details if a team is underperformed results wise or not. Sir Alex and his start have nothing to do with it. A relevant example is with Brighton. They were in a relegation struggle for a while despite xG saying they should be a top half team, and it was bad luck keeping them down there. Eventually their form turned around and they started pushing up the table. Their board kept faith in Potter because they were playing well but weren't getting results. XG was a start that showed they were playing well and just not getting the results.