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

sullydnl

Ross Kemp's caf ID
Joined
Sep 13, 2012
Messages
34,063
For what it's worth, going into this gameweek fbref had us fifth in the league for xG, xGA and xGD.

Obviously 6 games is way too short a sample size to read much into. Chelsea were 11th, 9th and 8th in those same stats for example, so fixture difficulty is obviously skewing things heavily (as presumably is being down to 10 men in the game against Liverpool).

That said, it should probably be concerning that even after just 6 games and even having had more difficult fixtures than us, City and Liverpool are already on top of xGD with +10.9 and +10.0 respectively, while we're trailing well behind at +3.2. With our fixtures getting tougher ahead, the early trend isn't positive. And I doubt today's game helped much.
 

AjaxCunian

vexingwijsneus
Joined
Mar 10, 2021
Messages
4,230
Supports
Ajax & United
For what it's worth, going into this gameweek fbref had us fifth in the league for xG, xGA and xGD.

Obviously 6 games is way too short a sample size to read much into. Chelsea were 11th, 9th and 8th in those same stats for example, so fixture difficulty is obviously skewing things heavily (as presumably is being down to 10 men in the game against Liverpool).

That said, it should probably be concerning that even after just 6 games and even having had more difficult fixtures than us, City and Liverpool are already on top of xGD with +10.9 and +10.0 respectively, while we're trailing well behind at +3.2. With our fixtures getting tougher ahead, the early trend isn't positive. And I doubt today's game helped much.
It is so clear Liverpool and City are much better sides than us, but some still think us finishing in front of Liverpool was by our brilliance and not them having an absurd injury crisis.
 

Robaldo

New Member
Newbie
Joined
Jun 25, 2015
Messages
344
xG is not a perfect art, but it does belittle Ole's constant references to players not taking chances or the team not being clinical enough.

Great player and a good manager, but by no means a great latter, which is precisely what we need to break our duck.
 

adexkola

Doesn't understand sportswashing.
Joined
Mar 17, 2008
Messages
48,031
Location
The CL is a glorified FA Cup set to music
Supports
orderly disembarking on planes
I think people are a bit confused regarding the reliability of xG when applied to a single game.

Yes, there is more variability around a single game statistic. But the idea that xG is garbage until you can analyze it over an entire season is contradictory. You cannot, statistically speaking, collate 38 garbage statistics and have the sum be a sound indicator of anything.

If you believe that over 38 games xG is very useful, you are not allowed to say it is pointless or nearly pointless in the context of a single game. A statistic can have a fat confidence band around it and still be useful in most situations.
 

emperortan

New Member
Newbie
Joined
Mar 22, 2015
Messages
314
Great player and a good manager, but by no means a great latter, which is precisely what we need to break our duck.
"Good manager" has ability to make their career in mid-table EPL clubs, which I don't see in Ole. No midtable EPL club would hire him before and now. Ironically, in EPL, only MU would dare to do the absurd move like that.
 

anant

Correctly predicted Italy to win Euro 2020
Joined
Feb 28, 2015
Messages
8,258
I think people are a bit confused regarding the reliability of xG when applied to a single game.

Yes, there is more variability around a single game statistic. But the idea that xG is garbage until you can analyze it over an entire season is contradictory. You cannot, statistically speaking, collate 38 garbage statistics and have the sum be a sound indicator of anything.

If you believe that over 38 games xG is very useful, you are not allowed to say it is pointless or nearly pointless in the context of a single game. A statistic can have a fat confidence band around it and still be useful in most situations.
No one's saying entire season. Just over a big enough sample - atleast 10-12 games. And it's not me, if you follow any person on twitter who is an analyst, they'd say the same.

At the end of the day, xG is sort of a regression model. So, there is going to be an error term associated with each shot, which in most cases would only cancel out over a big enough sample.
 

steve.crowford

New Member
Newbie
Joined
Aug 2, 2019
Messages
329
Location
Ukraine
Its funny how Ole's friends in the media keep moving the goal posts. At the start of the season after signing Ronaldo, they were all saying Ole MUST challenge for the title and get a deep run in the CL with this squad. They're now saying he has to win a trophy. It's honestly pathetic.
Give any other world class manager this squad and the expectation will ALWAYS be to challenge for the title AND win a trophy. The fact that expectations are so low under Ole just goes to show that nobody actually rates him as a manager. It's just vibes and good feelings and quite frankly, even that is just absolute nonsense and embarrasing to say the least.
Then there's always the caveat of him building a great squad. Who cares? He's not the director of football! It's not that difficult jotting down a list of the best players in world football. All that matters is what you do with that group of players and Ole is failing at absolutely everything.
The standards are so low at this football club, it's honestly sad to see. You have white flags waving at the Bernabeu after just losing to Sherif, meanwhile were sat there singing "viva Ronaldo" and "Oles at the wheel" after getting a last second winner against Villerreal. Theres levels to this.
 

sullydnl

Ross Kemp's caf ID
Joined
Sep 13, 2012
Messages
34,063
Not xG but interesting nonetheless:





Coincidentally or not the three teams with the biggest reputation for investing in analytics on top.
 

FatTails

New Member
Joined
Mar 21, 2021
Messages
1,859
I think people are a bit confused regarding the reliability of xG when applied to a single game.

Yes, there is more variability around a single game statistic. But the idea that xG is garbage until you can analyze it over an entire season is contradictory. You cannot, statistically speaking, collate 38 garbage statistics and have the sum be a sound indicator of anything.

If you believe that over 38 games xG is very useful, you are not allowed to say it is pointless or nearly pointless in the context of a single game. A statistic can have a fat confidence band around it and still be useful in most situations.
It’s not garbage in one game, no. I think most people mean it is not predictive, but duh, no single datapoint can be predictive on its own, regardless of quality.

If our xG is lower than our goals in one game, that doesn’t mean much about us as a team but it means something about that particular game. Now if we have a decent size sample, we can start making some inferences about us as a team, and possibly some predictions on what comes next.
 

FatTails

New Member
Joined
Mar 21, 2021
Messages
1,859
Not xG but interesting nonetheless:





Coincidentally or not the three teams with the biggest reputation for investing in analytics on top.
Surprised we’re not even lower. Last year we were 17th at % of shots from the 6 yards box, and 10th at % of shots from within the 18 yards box.
 

tomaldinho1

Full Member
Joined
Nov 26, 2015
Messages
17,374
Surprised we’re not even lower. Last year we were 17th at % of shots from the 6 yards box, and 10th at % of shots from within the 18 yards box.
Honestly, the best thing about all the stats pouring into football now is managers can’t hide. Fans have a lot more basic knowledge and means to check it now.
 

anant

Correctly predicted Italy to win Euro 2020
Joined
Feb 28, 2015
Messages
8,258
Surprised we’re not even lower. Last year we were 17th at % of shots from the 6 yards box, and 10th at % of shots from within the 18 yards box.
Shot distance without the scenario is a pointless stat. For example, in terms of shot distance we were 14th best in the league, Spurs 13th best, Chelsea 4th best however in terms of npxG/shot, we were tied 8th with Chelsea, Spurs were T-1. What that makes you believe is that when Spurs had a shot, they were probably 1v1, whereas in case of Chelsea, there were multiple players between the attacker and the goal
 

FatTails

New Member
Joined
Mar 21, 2021
Messages
1,859
Shot distance without the scenario is a pointless stat. For example, in terms of shot distance we were 14th best in the league, Spurs 13th best, Chelsea 4th best however in terms of npxG/shot, we were tied 8th with Chelsea, Spurs were T-1. What that makes you believe is that when Spurs had a shot, they were probably 1v1, whereas in case of Chelsea, there were multiple players between the attacker and the goal
It’s not pointless at all. xG is heavily affected by shot distance, even without context, and xG is pretty predictive of actual score.

There are anomalies like what you mentioned but overall the closer to a goal a shot is taken, the higher the odds of scoring. If you want to focus on anomalies (like when a shot from distance has higher odds than one from inside the 6 yard box), every metric will be useless.
 

adexkola

Doesn't understand sportswashing.
Joined
Mar 17, 2008
Messages
48,031
Location
The CL is a glorified FA Cup set to music
Supports
orderly disembarking on planes
It’s not garbage in one game, no. I think most people mean it is not predictive, but duh, no single datapoint can be predictive on its own, regardless of quality.

If our xG is lower than our goals in one game, that doesn’t mean much about us as a team but it means something about that particular game. Now if we have a decent size sample, we can start making some inferences about us as a team, and possibly some predictions on what comes next.
Of course you can't make reliable season predictions using one game. If that was the bone of contention then this thread would be shorter.

Regarding the bolded? Depends. Off by a little? Then sure, noise. Off by a lot? Then questions should be asked about whether our performance correlated with the final score. Doesn't mean the answer will always be negative (we may have an ultra clinical striker, after the goals we locked up shop)... But the stat can be a useful indicator for a single game without it being predictive of our performance 10 games down the line.
 

anant

Correctly predicted Italy to win Euro 2020
Joined
Feb 28, 2015
Messages
8,258
It’s not pointless at all. xG is heavily affected by shot distance, even without context, and xG is pretty predictive of actual score.

There are anomalies like what you mentioned but overall the closer to a goal a shot is taken, the higher the odds of scoring. If you want to focus on anomalies (like when a shot from distance has higher odds than one from inside the 6 yard box), every metric will be useless.
I'm not saying otherwise. But if my xG/shot is higher because of the style of play, do players give a feck?

The closer you are to goal, higher the xG like you said. However, number of players between you and the goal, what height are you striking the ball at, the body part used also play a role. In this case, it's more likely that teams that rely more on direct attacks - like we do and what Spurs did last season - they wouldn't mind taking a shot from 18 yards as in most likelihood, they're already 1v1 vs the GK. Teams like City rely on stringing passes together to break the defence down and then using a cutback - which would result in higher number of shots from 12 yards out.

And hence this metric. If anything, xG/shot is a better metric as it takes nto account all such nuances that you're trying to highlight by using just the shot distance
 

JB7

Full Member
Joined
Jun 18, 2008
Messages
8,826
Honestly, the best thing about all the stats pouring into football now is managers can’t hide. Fans have a lot more basic knowledge and means to check it now.
I tend to go the other way if I'm honest. Fans tend to think because they have a few statistics to back up an argument that they genuinely know better than qualified people who see the players day in day out. They fail to realise that similar statistics can usually be used to back up the other side of the argument and they also don't use the most vital thing of all to formulate their own opinion - their own eyes.
 

adexkola

Doesn't understand sportswashing.
Joined
Mar 17, 2008
Messages
48,031
Location
The CL is a glorified FA Cup set to music
Supports
orderly disembarking on planes
I tend to go the other way if I'm honest. Fans tend to think because they have a few statistics to back up an argument that they genuinely know better than qualified people who see the players day in day out. They fail to realise that similar statistics can usually be used to back up the other side of the argument and they also don't use the most vital thing of all to formulate their own opinion - their own eyes.
Your eyes are trash. It's not an insult, mine are too. So are everyone's. Otherwise, clubs wouldn't bother with hiring geniuses to look at numbers. They would just use their eyes.
 

tomaldinho1

Full Member
Joined
Nov 26, 2015
Messages
17,374
I tend to go the other way if I'm honest. Fans tend to think because they have a few statistics to back up an argument that they genuinely know better than qualified people who see the players day in day out. They fail to realise that similar statistics can usually be used to back up the other side of the argument and they also don't use the most vital thing of all to formulate their own opinion - their own eyes.
Just as an example, xG he become a better (and unbiased) tool than the ‘eye test’. Even now, in its infancy, it’s already broadly accurate at a high level. Human beings make errors, lots of them, particularly human beings who want to see something in manner that pleases them. It might kill the sport but a fan now knows far more about the game then a fan of 10 years ago because there is so much data and data is simply what is it, there is no bias.
 

JB7

Full Member
Joined
Jun 18, 2008
Messages
8,826
Your eyes are trash. It's not an insult, mine are too. So are everyone's. Otherwise, clubs wouldn't bother with hiring geniuses to look at numbers. They would just use their eyes.
Just as an example, xG he become a better (and unbiased) tool than the ‘eye test’. Even now, in its infancy, it’s already broadly accurate at a high level. Human beings make errors, lots of them, particularly human beings who want to see something in manner that pleases them. It might kill the sport but a fan now knows far more about the game then a fan of 10 years ago because there is so much data and data is simply what is it, there is no bias.
Both completely miss my point.

I've not said that statistics don't have their place within the game, of course they do and that is why clubs hire people to evaluate all different types of data.

What I am saying is the amount of people that now think they have a better understanding of the game than supporters in years gone by because they can quote a series of statistics that they barely understand is ridiculous. Fans don't know more now than they did ten years ago, they have access to more data correct but they don't have any more understanding of the game than they did 10, 20, 30 years ago.

As for the belittling of the eye test, if what you both appear to be suggesting is true then there is no need to fans to actually watch the matches to formulate their own opinions, which again, is completely ridiculous. If the eye test was redundant then every scout in the game would be out of a job.
 

Daengophile

Full Member
Joined
Jan 12, 2015
Messages
504
It's all just the end product of computed gobbledygook.

Some kind of juvenile algorithm based on rudimentary player statistics that in themselves are as worthless as chewed gum

If you want to see the truth, read below:

The meaning of life is "Manchester United"

Computer says the meaning of life is 42

Liverpool, for the record, is a big, fat zero
 

tomaldinho1

Full Member
Joined
Nov 26, 2015
Messages
17,374
Both completely miss my point.

I've not said that statistics don't have their place within the game, of course they do and that is why clubs hire people to evaluate all different types of data.

What I am saying is the amount of people that now think they have a better understanding of the game than supporters in years gone by because they can quote a series of statistics that they barely understand is ridiculous. Fans don't know more now than they did ten years ago, they have access to more data correct but they don't have any more understanding of the game than they did 10, 20, 30 years ago.

As for the belittling of the eye test, if what you both appear to be suggesting is true then there is no need to fans to actually watch the matches to formulate their own opinions, which again, is completely ridiculous. If the eye test was redundant then every scout in the game would be out of a job.
For me it’s simply logical that fans would know more now, on average, because there is more coverage, more games, more data, more everything right now. I’m middle aged for the record but it’s not hard for me to admit and is all part of football’s growth.

Re the eye test part, sadly most seem to have forgotten football was created for entertainment. You watch a game for entertainment. I’ve played all my life, played to a decent level (which again ties into why fans now on average know more as participation is much higher than 10,20,30 years ago). I, and I’m sure many on this forum, like to think we have a good idea of what’s happening in a game and probably, we do but stats definitely do. That’s the difference here.

On scouting, that is actually true. The role of scouts is changing dramatically - data again is taking centre stage. Look at FM, they built a huge database by employing their own scouts and now PL clubs, instead of signing their own scouts, pay them for their data. Everything is become data driven and every league will be mapped and, so long as there’s money in the sports, scouts will be predominantly ‘scouting’ on computer screens, not traveling to watch matches and so will every club still have as many on their books? Sadly, probably not.
 

NZT-One

Full Member
Joined
Mar 8, 2021
Messages
2,342
Location
Berlin
For me it’s simply logical that fans would know more now, on average, because there is more coverage, more games, more data, more everything right now. I’m middle aged for the record but it’s not hard for me to admit and is all part of football’s growth.

Re the eye test part, sadly most seem to have forgotten football was created for entertainment. You watch a game for entertainment. I’ve played all my life, played to a decent level (which again ties into why fans now on average know more as participation is much higher than 10,20,30 years ago). I, and I’m sure many on this forum, like to think we have a good idea of what’s happening in a game and probably, we do but stats definitely do. That’s the difference here.

On scouting, that is actually true. The role of scouts is changing dramatically - data again is taking centre stage. Look at FM, they built a huge database by employing their own scouts and now PL clubs, instead of signing their own scouts, pay them for their data. Everything is become data driven and every league will be mapped and, so long as there’s money in the sports, scouts will be predominantly ‘scouting’ on computer screens, not traveling to watch matches and so will every club still have as many on their books? Sadly, probably not.
Agreed. I think, it is part of modern times and a somewhat trend to focus on data instead of humans for specific things. I mean, I am sure, that the stats we have today are nothing more than a verbalized form of stuff, some people were able to see 30 years ago as well. Back then, it would have been called "instinct", "experience" and "he/she has an eye". Now some of the things that were looked at are identified and defined, people just know what to look for. And this occurs with everything. The more we understand, the more stats are available.

In terms of scouts, I think, it is very understandable, that clubs try to shift to models that aren't entirely based on specific individuals with "the right instinct or experience", therefor they invest in data pools. They rely on services who are able to provide data independent of employees, all you need is a data analysist who takes the data and comes to conclusions. That being said - I don't think, the times of scouts are gone - absolutely not. For a simple reason: all the current data coverage is about the measurable side of things, but reality is more than just its measurable side. Therefor scouts are needed, their instincts and experience is needed because they round of the whole idea of gathering information. Who knows, with more technology, more sensors, more data creation, AI based support systems to deal with the huge piles of data that is already created, at some point scounts won't be needed. But this is far future.

I am very happy to hear United investing in great talents to boost their usage of modern technology, it isn't a silver bullet at all, but not making use of it, would be a wasted chance.
 

sullydnl

Ross Kemp's caf ID
Joined
Sep 13, 2012
Messages
34,063


*insert all necessary qualifiers for looking at xG in relation to a one-off game*
 

sullydnl

Ross Kemp's caf ID
Joined
Sep 13, 2012
Messages
34,063


Different way of calculating xG than above, though again Leicester do better as you'd expect from having watched the game.

The more telling thing on this graphic is the shot map. Taking all those low xG shots....
 

NZT-One

Full Member
Joined
Mar 8, 2021
Messages
2,342
Location
Berlin


Different way of calculating xG than above, though again Leicester do better as you'd expect from having watched the game.

The more telling thing on this graphic is the shot map. Taking all those low xG shots....
What do you do if you have nothing up your sleeve...Obviously it isn't ok from Leicester to no give us any room...
 

RazorOz

Full Member
Joined
Feb 10, 2014
Messages
252
The thing is people like to discredit xG as irrelevant when the result goes against the xG, but reality is you won't continually out perform expectation. If you keep performing crap, eventually it will catch up to you. Is there a single game you can point to this season where we actually performed well? Individual brilliance won't bail you out every game.
 
  • Like
Reactions: Cheimoon

Champ

Refuses to acknowledge existence of Ukraine
Joined
Jun 17, 2017
Messages
9,888


*insert all necessary qualifiers for looking at xG in relation to a one-off game*
So expected goals state it was a close game up until 75 mins plus, which was when we brought Matic off....

Looks to me that Leicester took the chances they had, United didn't, however as you correctly state, expected goals aren't meant to be looked at in a single game format.
 

TMDaines

Fun sponge.
Joined
Sep 1, 2014
Messages
13,906
Anyone got an updated one of those charts that plot xG for and against match by match? That will show the funk that United now find themselves in. We’re now on a run of deservedly drawing and losing games.
 

TheMagicFoolBus

Full Member
Joined
Nov 5, 2016
Messages
6,502
Location
Lisboa, Portugal
Supports
Chelsea
Anyone got an updated one of those charts that plot xG for and against match by match? That will show the funk that United now find themselves in. We’re now on a run of deservedly drawing and losing games.

This is from The Athletic - rolling 10 game average of xG for and against:
https://theathletic.com/2884251/202...righton-what-the-numbers-say-about-your-team/

Here's what they said about United specifically:
United came just short of a genuine title challenge last season while still being right up there near the top of the table for most of the year. Judging by the start they have had, it does look like it will be more of the same.

Ole Gunnar Solskjaer has brought in star names but not all of them have hit the ground running. Looking at United’s attack, their xG has not necessarily gone up a gear, registering a non-penalty xG of 1.5 per game — an improvement of only 0.1 compared with last season and good enough for the sixth-best in the league.

At the other end, they look vulnerable on the counter and teams can get a shot away on the break too easily — most recently highlighted by the goal they conceded against Everton.

Their xG-against of 1.1 per game is nearly double that of Manchester City, and the trend looks to be creeping up further despite not playing any of their league rivals. It feels like United haven’t quite clicked.
 

Classical Mechanic

Full Member
Joined
Aug 25, 2014
Messages
35,216
Location
xG Zombie Nation
expected goals aren't meant to be looked at in a single game format.
It can't be used as a predictor of future form from a single game but it can be used to look at the balance of the game, which side created and conceded the bigger chances and suggests which side was the 'deserving' winner or loser.
 

Champ

Refuses to acknowledge existence of Ukraine
Joined
Jun 17, 2017
Messages
9,888
It can't be used as a predictor of future form from a single game but it can be used to look at the balance of the game, which side created and conceded the bigger chances and suggests which side was the 'deserving' winner or loser.
The issue being things like penalties, they can skew expected goals massively be it a miss or a goal.
Its a metric which really cannot be relied upon for a single game as pointed out, but more for a season,