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

Paxi

Dagestani MMA Boiled Egg Expert
Joined
Mar 4, 2017
Messages
27,678
xG needs context. Same way that 70% possession can mean you totally dominated a team, or could mean you passed it back and forth between your defenders without hurting them. Or how having made very few tackles could mean your defenders are timid, or it could mean they're excellent at reading the game.

I still prefer xG to shots on/off target, but in all cases stats should be used to illustrate a point, not be adhered to slavishly.
Great post.
 

Cait Sith

Full Member
Joined
Jun 8, 2014
Messages
1,379
How can one even say it's a pointless statistic when in fact it is SO accurate. Just by sheer statistics alone it predicts Man City to have 73 goals for, 21 against and 69 points. In reality they have 76, 20, 71. It's the most insanely accurate statistic in football that I know of and has nothing to do with "70 % possession means nothing".

Even without statistics it should be clear that this United side isn't going to keep the current win rate alive and it will normalize again. This United team isn't a "14 wins out of 17 games" side.
 

11101

Full Member
Joined
Aug 26, 2014
Messages
21,206
Just another set of stats for those who don't know what they're watching.

It's meaningless without context or understanding of where the numbers come from. We might finish more chances than expected because we're on a lucky streak, or it might be because we have Ruud Van Nistelrooy up front.

There are a bunch of different calculations for it too, lots of clubs use it but with their own formulas. I don't know where the one that's been adopted by the Caf comes from.
 

BluesJr

Owns the moral low ground
Joined
May 15, 2013
Messages
9,047
How can one even say it's a pointless statistic when in fact it is SO accurate. Just by sheer statistics alone it predicts Man City to have 73 goals for, 21 against and 69 points. In reality they have 76, 20, 71. It's the most insanely accurate statistic in football that I know of and has nothing to do with "70 % possession means nothing".

Even without statistics it should be clear that this United side isn't going to keep the current win rate alive and it will normalize again. This United team isn't a "14 wins out of 17 games" side.
We are though.
 

Sarni

nice guy, unassuming, objective United fan.
Joined
Jan 21, 2004
Messages
57,441
Location
Krakow
I blame Americans for everonye's obssesion about stats and metrics in sports. They are all shit.
 

VeevaVee

The worst "V"
Scout
Joined
Jan 3, 2009
Messages
46,248
Location
Manchester
Too much focus on stats these days. Some people just repeat them ad nauseam with total disregard to what's happened in front of their eyes.
 

kundalini

Full Member
Joined
Aug 26, 2004
Messages
5,750
We had this discussion about a week ago.

When you actually broke down the points xG awarded for each shot it started to not make much sense. I wouldn't go as far as saying its a load of shit, but it needs refining as some of the shots it was claiming were easy were in fact not easy at all when viewed. It also doesn't seem to take into consideration whether the player taking the shot has his view blocked by opponents, it also doesn't seem to take into consideration opponent pressure.

It's extremely flawed right now and anyone using it as evidence frankly doesn't understand how football works and should be laughed at. Repeatedly.
There are different models. Statsbomb include shots under pressure in their model. At the moment I believe the StatsBomb model is the most sophsticated by far. It includes height of the ball when the shot took place.

Watching the match is a far better way to assess a game than using any stats but xG is a good way to get a sense of a team's strength compared to watching occasional highlights or Youtube clips.
 

harms

Shining Star of Paektu Mountain
Staff
Joined
Apr 8, 2014
Messages
27,952
Location
Moscow
Is it unusual to have such a massive disconnect between xG and the actual points on the table? Is this is all just a complete load of bollox? Does it tell us anything useful at all?
It's definitely not a load of bollox. Mourinho's previous season saw us as incredible overachievers in terms xG, which proved to be right the season after that. De Gea in God-like form (the only thing that xG doesn't really count — quality of goalkeepers) and luck influenced it massively.

We were very clinical under Ole, and definitely scored more than we should've. We won't sustain the same form (we already turned it down a bit after the start attacking-wise), but, on the other side, our players are getting used to the system and to Ole's ideas, so we'll probably see an improvement in our xG numbers once this injury madness will end, which will balance it out.
 

RedSky

Shepherd’s Delight
Scout
Joined
Jul 27, 2006
Messages
74,145
Location
Hereford FC (Soccermanager)
There are different models. Statsbomb include shots under pressure in their model. At the moment I believe the StatsBomb model is the most sophsticated by far. It includes height of the ball when the shot took place.
You got stats for the Palace game? I don't know which website someone linked to in the Ole thread, but the numbers for that game looked very wrong when looking at each chance from a video perspective.
 

Revan

Assumptionman
Joined
Dec 19, 2011
Messages
49,419
Location
London
I'm probably being a bit self-indulgent but the Ole thread is so busy this specific topic might get lost. Why are xG painting such a grim picture of his time in charge? Is it unusual to have such a massive disconnect between xG and the actual points on the table? Is this is all just a complete load of bollox? Does it tell us anything useful at all?

Expected goals against is not super important IMO, we have De Gea in goal so it is always going to be higher than how many goals we concede. Having De Gea is essentially cheating.

Goals we scored vs expected goals is probably a bit worrisome. It might indicate clinical finishing, it might indicate the fact that we have good strikers (good strikers score more from the same chances as not good strikers). but it might partially be just good luck, and eventually, luck is going to run out. Frankly speaking, I think that we have been a bit lucky since Ole came (yesterday being a bizarre match which won't be repeated for eternity). We definitely need to inject some quality in attack.
 

FlawlessThaw

most 'know it all' poster
Joined
Oct 26, 2005
Messages
29,599
The interesting aspect of xG seems to be that all the top teams have more points than they should and all the smaller teams have less points. Clearly highlights the ruthlessness of the top teams and the fact that sometimes they take their foot off the pedal as well.

The PL stats have reassured me at least that we aren't doing that badly under Ole there. Guess we have just outperformed in Cup and CL because we have faced tough opposition there.
 

hkjack

New Member
Newbie
Joined
Oct 17, 2017
Messages
65
Supports
Man City
How do they define the expected goal?
 

harms

Shining Star of Paektu Mountain
Staff
Joined
Apr 8, 2014
Messages
27,952
Location
Moscow
We score more than 'expected' because we have some good attackers.
Over the course of few seasons it actually isn't really working (even though your logic is solid). Only a few players, like Messi, keep their outstanding G/xG rate. Best attackers just tend to get into better goalscoring positions more on a regular basis, that's where the difference lies.

De Gea does influence it though, and significantly. The "Messi of goalkeeping" tag is quite appropriate here.
 

TMDaines

Fun sponge.
Joined
Sep 1, 2014
Messages
13,906
I'm probably being a bit self-indulgent but the Ole thread is so busy this specific topic might get lost. Why are xG painting such a grim picture of his time in charge? Is it unusual to have such a massive disconnect between xG and the actual points on the table? Is this is all just a complete load of bollox? Does it tell us anything useful at all?

I don't think it is painting a grim picture, but it is just suggesting that we are getting pretty much the maximum out of our performances and winning a lot of games that could have easily been drawn or perhaps lost right now. I don't think that is a controversial thing to say when you watch our games carefully, but a lot of people seem offended by the notion.

If we want to have a good chance of winning the league next season, our performances will likely need to improve. It is unlikely we continue to score as much as we do with the chances we have, and unlikely we concede as little we do with the chances we give away.

That said, Ole came in when the club was in disarray and has built a team on the fly. I expect performances to improve when he gets more time and there's probably something in giving players confidence, which allows them to outperform xG in the short term. Long term, this may fade, but motivation and belief is a big part of football too, which xG cannot capture really.
 

calodo2003

Flaming Full Member
Joined
Feb 8, 2014
Messages
41,578
Location
Florida
I'm probably being a bit self-indulgent but the Ole thread is so busy this specific topic might get lost. Why are xG painting such a grim picture of his time in charge? Is it unusual to have such a massive disconnect between xG and the actual points on the table? Is this is all just a complete load of bollox? Does it tell us anything useful at all?

RAWK is using this as a laughing point when it comes to appointing Ole as full time. I can’t figure out why, though.

I also equate xG with fairy tales, that may cloud my judgement.
 

TMDaines

Fun sponge.
Joined
Sep 1, 2014
Messages
13,906
We score more than 'expected' because we have some good attackers. We concede less than 'expected' because we have the best goalie in the world.

xG is completely bollocks without context.
This is a fallacy. The top scorers score the most as they get the most chances, not because they are clinical particularly. It's why I am never too disappointed if a forward is missing chances, as getting them is the hardest bit.
 

RedSky

Shepherd’s Delight
Scout
Joined
Jul 27, 2006
Messages
74,145
Location
Hereford FC (Soccermanager)
So wait, if there are several different xG models floating about then why is everyone saying xG when every model is going to produce a different number. How many xG models are there? :lol:

How are we meant to know which model each user/website is using? Hopeless. At least use the model name as part of a prefix so people know which one you're using. feckin ell.
 

CM10

Full Member
Joined
Jul 24, 2014
Messages
7,275
I'm not convinced by xG as a stat. It gives you some indication of the quality of chances teams create but the measures used to decide it are still pretty arbitrary.

Basically, if we're consistently outscoring it, that can only be a good thing.
 

Gator Nate

Full Member
Joined
Jan 15, 2019
Messages
2,140
Location
Orlando, FL
There is a huge misunderstanding about Expected Goals.

Name an elite team that is performing well this season (Man City, Liverpool, Dortmund, Barcelona etc etc) and you will see that they are massively out-performing their expected goals numbers. In the table linked, green is better than expected, red is worse. https://understat.com/league/EPL/2018

Likewise, the teams at the bottom of the table will typically have been a bit unlucky, missed good chances to score in some tight games they ended up losing or drawing.

It's not just this season, it is true of most seasons. When Chelsea won the title under Conte, their expected goals for was 61.80, yet in reality they scored 85. Their expected points 75, in reality 93. Spurs finished 2nd. They too massively out-performed their expected numbers. City last season scored 14 more than expected, and got 9 more points than expected.

This may comes as a huge shock to you but elite players when performing well tend to finish chances at a better rate than an average player would. Elite keepers save shots that many other keepers wouldn't.

If you go through the individual games, Expected Goals didn't like the performances against Spurs and Leicester, matches in which United scored in the first half then sat deep and defended for the rest of the match. In contrast, the Burnley game in which we were 0-2 down, looks like it should have been an easy win for United as we created several big chances to score.

Based purely on chances, the Fulham match could have been a draw, yet anyone watching will have noticed that Fulham had several opportunities in the first 10 mins then almost nothing until the score was 0-3. Ryan Babbel's big chance after 76 mins when he hit the post from close range, was worth 0.59 xG (on average, out of 100 shots from that position, you'd score 59 times). Yet even if he had scored, Fulham would still have struggled to get a draw in the time remaining.
Thank you for putting that out there.

Folks seem to think of a tendency to revert to the mean. The problem is that xG's mean is for the entire league, from good teams with good players to bad teams with bad players (insofar as that goes in the PL). Generally speaking, a consistently good team, say, one of the big six, should never revert to the mean of the league. Likewise, a newly-promoted team won't rise above and revert to the mean of the league either.

On an individual level, in the same situation on the pitch, Aleksandar Mitrovic and Sergio Aguero have the same xG, because xG is a mathematical construct irrespective of the player. But I do expect Aguero to make that goal more often than Mitrovic, because he's a better player in general. Over time, I don't expect Aguero to fall to average, nor Mitrovic to rise to it.
 

kundalini

Full Member
Joined
Aug 26, 2004
Messages
5,750
You got stats for the Palace game? I don't know which website someone linked to in the Ole thread, but the numbers for that game looked very wrong when looking at each chance from a video perspective.
I use understat and Michael Caley, because they are free. Both models have significant flaws in them. Statsbomb provide their data to clubs and a few media outlets, no doubt at vast expense. They do however write a number of articles using their stats. https://statsbomb.com/ Here is one written about United on Feb 20th using Statsbomb's data. https://statsbomb.com/2019/02/solskjaers-manchester-united-the-real-deal-or-just-a-good-run/
 

Revan

Assumptionman
Joined
Dec 19, 2011
Messages
49,419
Location
London
How can one even say it's a pointless statistic when in fact it is SO accurate. Just by sheer statistics alone it predicts Man City to have 73 goals for, 21 against and 69 points. In reality they have 76, 20, 71. It's the most insanely accurate statistic in football that I know of and has nothing to do with "70 % possession means nothing".

Even without statistics it should be clear that this United side isn't going to keep the current win rate alive and it will normalize again. This United team isn't a "14 wins out of 17 games" side.
To be fair, it is one of the very few 'statistics' in football. Possession, the number of shots etc aren't even statistics, they are just features. And as you said, it is very accurate. Coaches would be idiots to not take this into considerations.
 

DWelbz19

Correctly predicted Portugal to win Euro 2016
Joined
Oct 31, 2012
Messages
33,745
How can one even say it's a pointless statistic when in fact it is SO accurate. Just by sheer statistics alone it predicts Man City to have 73 goals for, 21 against and 69 points. In reality they have 76, 20, 71. It's the most insanely accurate statistic in football that I know of and has nothing to do with "70 % possession means nothing".

Even without statistics it should be clear that this United side isn't going to keep the current win rate alive and it will normalize again. This United team isn't a "14 wins out of 17 games" side.
Apart from the fact that they quite literally are?
 

harms

Shining Star of Paektu Mountain
Staff
Joined
Apr 8, 2014
Messages
27,952
Location
Moscow
Like every stat it's not perfect though and requires context. People who are dismissing it completely are as wrong as those who value it over the actual results.
 

Dion

Full Member
Joined
Mar 18, 2010
Messages
4,327
That's encouraging.
I think it's undeniable that results have gone a little bit better than our performances have merited and that won't always be the case. It's worth pointing out that top teams do generally out-perform their xG over the course of a season. This season all the top 6 have, all the top 9 did the season before and we were the only top 6 team not to the season before that. In La Liga virtually every team did in 16-17 and the top 6 did in 17-18. Barcelona are doing it this year.

Atletico have conceded less than their xGA by between 4-12 every year, this doesn't happen by coincidence. xG doesn't take into account style of play, just shot position. A team that plays on the counter is less likely to concede from a shot from the same position as an expansive, open team and we've played on the counter lots under Ole.
 

RedSky

Shepherd’s Delight
Scout
Joined
Jul 27, 2006
Messages
74,145
Location
Hereford FC (Soccermanager)
I use understat and Michael Caley, because they are free. Both models have significant flaws in them. Statsbomb provide their data to clubs and a few media outlets. They do however write a number of articles using their stats. https://statsbomb.com/ Here is one written about United on Feb 20th using Statsbomb's data. https://statsbomb.com/2019/02/solskjaers-manchester-united-the-real-deal-or-just-a-good-run/
Yeah see, that's the website that I totally disagreed with as their values for the Palace game seemed wrong to me. *shrugs*
 

Cait Sith

Full Member
Joined
Jun 8, 2014
Messages
1,379
Thank you for putting that out there.

Folks seem to think of a tendency to revert to the mean. The problem is that xG's mean is for the entire league, from good teams with good players to bad teams with bad players (insofar as that goes in the PL). Generally speaking, a consistently good team, say, one of the big six, should never revert to the mean of the league. Likewise, a newly-promoted team won't rise above and revert to the mean of the league either.

On an individual level, in the same situation on the pitch, Aleksandar Mitrovic and Sergio Aguero have the same xG, because xG is a mathematical construct irrespective of the player. But I do expect Aguero to make that goal more often than Mitrovic, because he's a better player in general. Over time, I don't expect Aguero to fall to average, nor Mitrovic to rise to it.
Agüero scored 18 goals. His xGoal statistic is 17.

People really underestimate how accurate this is.
 

Revan

Assumptionman
Joined
Dec 19, 2011
Messages
49,419
Location
London
RAWK is using this as a laughing point when it comes to appointing Ole as full time. I can’t figure out why, though.

I also equate xG with fairy tales, that may cloud my judgement.
As a fairy tale as a data scientist crunching numbers which eventually are able to detect cancer and other sicknesses better than a team of top doctors?

Yes, it is happening.
 

HTG

Full Member
Joined
Apr 27, 2011
Messages
5,882
Supports
Bayern
The only reason people hate on these stats is, that they are objective and therefore don’t fit in with their personal bias.
 

Cait Sith

Full Member
Joined
Jun 8, 2014
Messages
1,379
Apart from the fact that they quite literally are?
When I flip a coin 5 times and get tails 5 times doesn't make the probability of getting tails 100 %. It should be clear what I mean. United is overperforming.

Do you think United will keep this rate up over a prolonged period of time (let's say 70 games)?
 

kundalini

Full Member
Joined
Aug 26, 2004
Messages
5,750
This is a decent article on the issue. Having watched all our matches I don't agree with all of it, but it seems reasonably well balanced. It was written on the 20th Feb so doesn't feature matches against Liverpool, Palace, Southampton or PSG away.

https://statsbomb.com/2019/02/solskjaers-manchester-united-the-real-deal-or-just-a-good-run/

One of things you don't get from Expected Goals is whether you had a full strength team out or not. I feel a lot more confident having seen Solskjaer find ways to get results when seemingly key players have been missing, than I did when he was using a small group of players.
 
Last edited:

Nick7

Full Member
Joined
Jan 24, 2011
Messages
19,231
Location
Ireland
xG is a terrible stat to use out of context. The Southampton game proves why it's terrible. It condenses the beauty and unpredictability of attacking play into monotony.
 

Dion

Full Member
Joined
Mar 18, 2010
Messages
4,327
Agüero scored 18 goals. His xGoal statistic is 17.

People really underestimate how accurate this is.
Messi's xG is 19 and his actual goals are 25.

It's accurate on batch data, no individual points.
 

Revan

Assumptionman
Joined
Dec 19, 2011
Messages
49,419
Location
London
When I flip a coin 5 times and get tails 5 times doesn't make the probability of getting tails 100 %. It should be clear what I mean. United is overperforming.

Do you think United will keep this rate up over a prolonged period of time (let's say 70 games)?
Indeed.

And of course we won't be able to keep up. Even forgetting about xG, we have won 9 away games in a row (all time United talker, not sure if it is an all time record for any English team), while if you interpolate the points we have got during Ole's era in EPL over the course of an entire season, it is an all time record for any English club, and second best ever for any club in top 5 leagues. It is totally not sustainable and it will regress to the mean. Of course, the mean is an unknown and xG do a decent (though far from great) job at approximating it.
 

rotherham_red

Full Member
Joined
Mar 12, 2005
Messages
7,407
Tbh I agree with the XG stat and it confirms to me that whilst we have undergone a renaissance under Ole under the microscope the actual quality of the football hasn’t really been that great.

It’s just that the football under Jose was that shit, we look miles better but it’s more between the ears where Ole has performed miracles rather than actually tactically/technically where I still think we fall short of Europe’s elite.

We’ve riden some luck as well but when you’re feeling positive and you fight to the last whistle then things will go in your favour. We’re getting results which the ‘quality of our football’ doesn’t merit but our heart does if that makes sense.
But that's not the question here. Much of the xG proponents are saying that it will undoubtedly regress back to the bad old days because of the numbers, but ignoring the context, which is where I have my issues with them.

A lot of them are assuming that a) the way we play won't change after Ole has an extended period of time with the players to work on tactical/structural/philosophical preparation (say, during a preseason) - look at what we managed to achieve with just a few days in Dubai, as an example; b) we won't bring in players who will fill the various holes in the squad during the summer; and c) that the players that have been in the team for the last month or so, are the first choice starting players; instead of actually looking at the context of the injuries we faced, and the strength of quality of the opposition.

A fairer comparison for the xG proponents would have been to compare the xG numbers of the likes of City, Utd, Chelsea, Liverpool, Arsenal, Spurs et al against each other during the period of time Ole has been in charge. I'd hazard a guess that their average xG numbers would not have been much different to ours on that basis. It certainly would paint a clearer picture rather than just looking at the last 17 games which Ole has been in charge of which has included a horror run of FA Cup fixtures in addition to difficult league games (whereas of the rest of the top 6, the only team who you could argue have had it as tough, have been Chelsea, and even that would be tenuous).

For instance, if you compared the xG plotmaps for City's (loss), Liverpool's (draw) and Utd's (win) games vs Leicester recently, you'd barely see any difference outside of a .1 or .2 here and there. The fact we were able to grind out a win in testing circumstances such as that is a testament to us, not something to deride Ole or the club.