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

[Ramshock]

Expected goals can feck right off.

I once taught biostatictics (multivariat mainly) and use stats in my job so I don't have an aversion to them in a general sense, but this level examination and interpretation by the majority is just fecking pointless and almost completely meaningless.

Revan is disappointed in you too probably Wibble

 

[Ramshock]

xG is just statistical wankery that has little to no bearing or affect on the outcome of games. I am genuinely amused that people are so enamoured with it that they take offense at people who feel its worthless.

 

[Pexbo]

Considering that I am working in the field, I probably know a bit or two about the technology behind it.

I also admitted that current xG models are relatively primitive, but dismissing them is still stupid.

Ha! Sorry mate I didn’t even bother to read the name but to be honest I’m surprised you’re giving it any credence whatsoever.

You more than anyone else on this forum understand the importance of data quality. As you say this is just primitive stuff and really can’t be taken seriously imo. We’re talking about a handful of scenarios taken into account, each with a relatively small number of permutations and 2 decimal place weight applied based on supposed difficulty.

The number of real permutations that aren’t taken into account make it completely pointless. Velocity of the ball, x,y,z of delievery, spin on ball and subsequent arc or swerve, goalkeeper’s starting position, goalkeeper’s momentum, numerous distraction possibilities, softness under foot, lumps on field, position of striker’s standing foot to name just a few.

It’s just far too simplistic to take seriously. As you know all about training NN’s I thought you’d find the idea of humans manually and crudely slicing a pitch up and coming up with statistics to reinforce their ideas laughable at best. Especially considering the complexities and variables involved.

I agree the principle of it is reasonable but right now we can’t and shouldn’t take anything from it.

 

[Random Task]

xG is an inherently flawed statistic due to the fact that it assumes all professional footballers are of equal ability. They are clearly not. Unpredictable events such as human error, form, weather patterns and a multitude of other factors renders the entire system inconsistent and ultimately useless. I'm surprised it gets so much 'air time' around here.

 

[Ramshock]

xG is just a starting point that serves as part of the foundation of the football analytics industry. I don't think anyone has suggested in this thread even - let alone in Man Utd's analytics dept - that xG on its own would indicate, "oh, we need to score more goals." xG is just a quick and dirty way to get a very rough picture.

But xG is now being overlaid and extended to provide much better insights by adding other data such as measurements of game states, passing in different areas of the pitch, number of opposition players bypassed by a pass, pace of the ball, skill of the individual player involved, etc. Which of these additional data will add most value is still a matter of intense discussion - and a lot of secrecy. Most teams won't talk about their own analytics efforts and where statisticians are under a team contract they typically do so under NDA.

I'm trying to think of an analogy. How about blood pressure? Nobody would only use blood pressure to diagnose someone - it's unreliable, it varies hour by hour, it's hard to get repeatable measurements over the short term, there are multiple causes, etc. But is it fundamental as an indicator of...something? Absolutely.

So, does anyone only use xG? Of course not. But it's still important.

You cannot mathematically map out the perfect game plan though can you? I feel like we need to get Dr Ian Malcolm in to talk about chaos theory.

 

[adexkola]

Tweet

— Twitter API (@user) date

Tweet

— Twitter API (@user) date

Cheers!

 

[adexkola]

xG is an inherently flawed statistic due to the fact that it assumes all professional footballers are of equal ability. They are clearly not. Unpredictable events such as human error, form, weather patterns and a multitude of other factors renders the entire system inconsistent and ultimately useless. I'm surprised it gets so much 'air time' around here.

Klopp was probably on to something...

 

[Paul_Scholes18]

xG is just a starting point that serves as part of the foundation of the football analytics industry. I don't think anyone has suggested in this thread even - let alone in Man Utd's analytics dept - that xG on its own would indicate, "oh, we need to score more goals." xG is just a quick and dirty way to get a very rough picture.

But xG is now being overlaid and extended to provide much better insights by adding other data such as measurements of game states, passing in different areas of the pitch, number of opposition players bypassed by a pass, pace of the ball, skill of the individual player involved, etc. Which of these additional data will add most value is still a matter of intense discussion - and a lot of secrecy. Most teams won't talk about their own analytics efforts and where statisticians are under a team contract they typically do so under NDA.

I'm trying to think of an analogy. How about blood pressure? Nobody would only use blood pressure to diagnose someone - it's unreliable, it varies hour by hour, it's hard to get repeatable measurements over the short term, there are multiple causes, etc. But is it fundamental as an indicator of...something? Absolutely.

So, does anyone only use xG? Of course not. But it's still important.

Yeah I think xG or chances created in certain areas of the pitch or from certain types of play would be more interesting. This could indicate how a team is getting good chances and help a manager prepare how he should stop a team/player. If you know player x has not been very good with his weaker foot you could allow him some more space on that side. Preventing movements that open up space that other team wants in order to get high xG could help teams defend better.

Although you need more than just xG in general for it to really tell you much about the bigger picture. Our xG against PSG was very high and they had a low xG so people might think we dominated the game.


I don't think xG tell us a lot about how Ole will do in the future as a United manager. The fact that he has lifted the team and given us a positive winning mentality is a big plus though.
How he manage to develop the team remains to be seen and he needs to deal with big defeats which he hasn't needed yet. Spells of lower confidence and wasting chances in games will come too.

 

[Brightonian]

Because xG is a total nonsense, probably. Based on several deeply flawed premises and highly subjective even if it wasn't.

 

[Paul_Scholes18]

Klopp was probably on to something...

Wind against Liverpool should give xG=0.

 

[rotherham_red]

Tweet

— Twitter API (@user) date

 

[adexkola]

Tweet

— Twitter API (@user) date

Click here to see whether human beings can multitask! The results will shock you!

 

[VJ1762]

I think xG is a good stat. It is a good reflection of where we need to improve. This club has a problem keeping possession in big games. In the last 12 years, the only world class midfielders we have bought are Carrick and Pogba. SAF ignored it until he retired, but he was able to wring every last drop out of the players available between 2009-2013. We seriously need world class talents like Ndombele and Veratti. Also, for a supposed big club, we concede way too many chances, which is a direct consequence of inability to retain possession. And Chris smalling is not good enough. We need another commanding CB like a VVD. I have never seen a CB who telegraphs his passes as much as Smalling does. And a right winger. With a pre-season and necessary signings made, we can see that the xG stats won't be so lopsided.

Data is the future of football. It will not be a surprise if future football scouts are data scientists. It has already revolutionised basketball, and it will do the same with the beautiful game.

 

[Moriarty]

Can xG just feck off forever nobody gives a shit

What is xG?

 

[LungiDevil]

This was my question to. I havent seen an explanation so I looked it up.



https://www.bbc.co.uk/sport/football/40699431

To pick some of the biggest outliers:

Team | offensive | defensive

Chelsea 16/17 | 37.5% | -15.3%
United 17/18 | 15.2% | 35.7%
Atletico 17/18 | 15.3% | 38%
Dortmund 14/15 | -12.7% | -25.7%

That was Dortmund's freak season in which they were dead last after 18 games. Infact after 18 games it looked like this: -24.2% | -32.8%. This is the biggest anomaly I can think of. People need to realise that 38 games is not a big enough sample and that is possible for teams to outperform their numbers over a whole season at unsustainable levels.

IMO, I would clarify the conclusion a teeny bit, because it makes the important distinction between luck and standard.

Yes 38 games look small as a sample size, but within a game you typically multiple attempts at goal and at your own goal. So that gives a sample size of 300-400 instances over which you have an average, which is pretty indicative of the standard of your team more than the luck. xG at the end of the day is a metric of conversion per chance eventhough it is aggregated and presented as per game for easier consumption.

So yes often times title winning teams will outperform that number. Part of it might be luck, but a good chunk of it will be due to better players and better form. Saying one can beat the xG for a whole season but not beyond that is just another way of saying it is difficult to carry title winning form across multiple seasons, which again we all intuitively know.

 

[99fan]

According to my understanding of Xg metric and taking in consideration the fact that it doesn't matter how many goals the team score per match, as long the team outscore the opponent to take the points.
so i think it useless to use an aggregate xg, instead each match xg should be translated to xpoints then be summed together.

Also an XG of 2.74 doesn't make sense in single match because obviously it cant resemble a real score, so i rounded every xg per match to nearest integer number.

The data taken from understat.com since Ole been in charge and i applied the previous mentioned rules to get the following results:

( unfortunatly i can't embed images so here a link to top 4 results and total points according xg)

imgur.com/a/L4VgUBc

 

[chromepaxos]

You cannot mathematically map out the perfect game plan though can you?

Er... so your argument is that you can?

I swear, I do not understand why people on here are so emotionally attached to knowing less about the game we love.

 

[Schneckerl]

Tweet

— Twitter API (@user) date

The dumbest type of tweets.

 

[chromepaxos]

xG is an inherently flawed statistic due to the fact that it assumes all professional footballers are of equal ability.

It assumes no such thing.

If you really think that when you calculate an average you're "assuming everyone is the same" then, honestly, you should probably stay away from numbers altogether.

 

[Schneckerl]

xG is an inherently flawed statistic due to the fact that it assumes all professional footballers are of equal ability. They are clearly not. Unpredictable events such as human error, form, weather patterns and a multitude of other factors renders the entire system inconsistent and ultimately useless. I'm surprised it gets so much 'air time' around here.

Would you say the same about shot/shots on goal?

 

[Ramshock]

Er... so your argument is that you can?

I swear, I do not understand why people on here are so emotionally attached to knowing less about the game we love.

No my argument is you can't. I fail to see the reason for your confusion.

 

[Ramshock]

It assumes no such thing.

If you really think that when you calculate an average you're "assuming everyone is the same" then, honestly, you should probably stay away from numbers altogether.

This is football not maths, hope this helps.

 

[Ramshock]

I think xG is a good stat. It is a good reflection of where we need to improve. This club has a problem keeping possession in big games. In the last 12 years, the only world class midfielders we have bought are Carrick and Pogba. SAF ignored it until he retired, but he was able to wring every last drop out of the players available between 2009-2013. We seriously need world class talents like Ndombele and Veratti. Also, for a supposed big club, we concede way too many chances, which is a direct consequence of inability to retain possession. And Chris smalling is not good enough. We need another commanding CB like a VVD. I have never seen a CB who telegraphs his passes as much as Smalling does. And a right winger. With a pre-season and necessary signings made, we can see that the xG stats won't be so lopsided.

Data is the future of football. It will not be a surprise if future football scouts are data scientists. It has already revolutionised basketball, and it will do the same with the beautiful game.

Yes its for the analysts but it will only work to a certain point then it relies on humanity and humanity is always unpredictable.

 

[ForestRGoinUp]

This stuff has infected all of the American sports and it seems to be inevitable in Europe now too. It’s just content filler for content creators devoid of interesting opinions. Good for gamblers I suppose but an utter waste of time otherwise.

 

[Ramshock]

This stuff has infected all of the American sports and it seems to be inevitable in Europe now too. It’s just content filler for content creators devoid of interesting opinions. Good for gamblers I suppose but an utter waste of time otherwise.

I would say it works best with baseball which is a game deep in statistical analysis and is also a stop start game that can be constantly adjusted throughout. Football is a flowing game with many unpredictable variables that makes the concept/value of xG limited in my opinion.

 

[chromepaxos]

No my argument is you can't. I fail to see the reason for your confusion.

I'm confused because you haven't actually made an argument. What is your argument exactly?

I think it is that xG can't predict the future or the result of every game, but that seems too stupid an argument for anyone to make?

 

[Snow]

If you're a top team and are below your xG you're not performing too well. You should be doing better. The problem with xG is naming it expected goals because that's not what the stat does since it's only the expected return in vaccum without consideration to other factors besides the shot iteself.

The goals we conceded against Southampton are way outside the prediction of xG. You can see that just by their virtue. Free kick goal and a young full back scoring outside the box. That's how football works. Game by game it just doesn't make sense. In basketball and baseball you can pretty much model every scenario where a point is gained so the statistical predictions work better there. Some will be happy with xG as a model based on an entire season but who honestly cares about that sort of hindsight information? You think the owner of Cardiff will care if they go down but the xG said that they should have stayed up?

 

[anant]

I'm sorry but this is just a straight misinterpretation of what the stat actually measures. Its a fallacy of projecting past outcomes into the future when the context changes.
Mourinho self imploded, and tactically left our defense both extremely exposed and while also not allowing us to attack properly. Its not about what can or cannot be sustained, but its a measure of how many clear cut shooting opportunities you create or concede. Three's no law saying that you can't be clinical with a lower xG and still win everything in sight (in fact teams playing a more pragmatic counter attacking style would).

As a counter, see how Ole coming in has put us back in the position we were in last year of outperforming our xGA. Why is our defense not conceding all these goals then? De Gea has hardly been in God mode apart from the one Spurs game.

I'm not sure if you understood the poibt but I'm talking about the difference in GA and xGA. There is no way a difference of 14 goals is sustainable! We were playing hopelessly bad last season as well and it was because of Dave being in God mode that we finished 2nd.
There is no law saying that lower xG means teams will finish lower, but I cant recall a single case where a team has consistently outperformed the xG.
Having said that, Ole case is slightly different. The difference between xG and xGA is 5 goals over 12 league games. We've outperformed our oppo or been in roughly the same xG range in nearly all games except the Spurs one which is contributing to a difference of nearly 2 goals in metric. Apart from this game, every game is contributing to a difference of 0.2-0.3 xG which can easily be explained as opposition tries to throw kitchen sink at us towards the end of the game and the quality of our GK

 

[Treble]

This is my last post for today so here goes:

It’s hard to tell because no team has actually defended their title since we last did it. But you can also interpret the fact differently. You can say that the fact they they couldn’t outperform it the next year means they were playing above themselves and just couldn’t sustain it. Or it could be they had a few very good players in key positions that either left or couldn’t reproduce that form. I’m more tempted with the latter.

I think xG is very useful for analyzing the finishing abilities of your strikers. For example if you’re consistently underperforming your xG stays and you’re losing, you need better finishers.

Similarly if you’re outperforming it, it means your forwards are in good form and are converting harder chances. You might say that they can’t do that consistently over a long period of time, but then they’ll just move on and become less prolific.

Now if a team is regularly creating easy chances for their strikers so they can pretty much be level with their xG’s you could say those chance creating stats are unsustainable (I’m not saying that cos if your system or players and good at doing that then more power to them).

My point is that xG stats measure finishing performance, but can be filled by very good midfield or complete team performances which need not be any more sustainable than outperforming the xG.

I dunno if that made sense.

Thanks. Still, your claim that there is no regression to the mean is an empirical one. But it is not based on empirical data. Nevertheless, you are cock sure that you are right. This is why you come across as, well, bigheaded, mate. You pretend to know something you don't.

 

[Beagle]

When one made up statistical metric fails to conform to what really happens out on the pitch, it's time for a new one.

 

[Treble]

Ha! Sorry mate I didn’t even bother to read the name but to be honest I’m surprised you’re giving it any credence whatsoever.

You more than anyone else on this forum understand the importance of data quality. As you say this is just primitive stuff and really can’t be taken seriously imo. We’re talking about a handful of scenarios taken into account, each with a relatively small number of permutations and 2 decimal place weight applied based on supposed difficulty.

The number of real permutations that aren’t taken into account make it completely pointless. Velocity of the ball, x,y,z of delievery, spin on ball and subsequent arc or swerve, goalkeeper’s starting position, goalkeeper’s momentum, numerous distraction possibilities, softness under foot, lumps on field, position of striker’s standing foot to name just a few.

It’s just far too simplistic to take seriously. As you know all about training NN’s I thought you’d find the idea of humans manually and crudely slicing a pitch up and coming up with statistics to reinforce their ideas laughable at best. Especially considering the complexities and variables involved.

I agree the principle of it is reasonable but right now we can’t and shouldn’t take anything from it.

That's a pretty shallow way of dismissing xG models. Idealisations, simplifications, abstractions are part and parcel of any successful modelling, no matter whether involving mental processes or artificial ones ("deep learning"). It's not like, "Shit, the model does not take into account x, y, z, it must be useless then!". That's naive at best.

The most serious limitations of statistical modelling are related to the inadequacy of its tools when it comes to identifying and tracking causal influences: what causes what. But there is already a breakthrough in this direction - Pearl, for instance, has created a mathematical language for causal reasoning ("The Book of Why"). It will take time before the new tools are successfully implemented but there is little doubt that modelling of complex pthenomena gets better and better.

Interestingly, the Understat xG picture of the top teams converges with what most fans see. It says that Liverpool are fortunate to be only a point behind City and this seems true. It says that Spurs were utterly fortunate to be only 2pts off City prior to the 27th game and this seems absolutely true. It says that Ole's United is winning at a better rate than the performances suggest and this seems true too. Sure, Ole's United seems revitalised, confident, willing to fight, capable of grinding out results. But it doesn't look yet like a team capable of winning 2,67 pts per game over the course of a whole season. Fergie's best achievement was 91pts, whereas Ole would win 102 pts if the current level of getting results continues throughout the next season. It might well happen, and it will probably happen, that we play better than now but the results are not so good.

 

[SqualorVictoria]

The dumbest type of tweets.

Ogden has some world class ability to have the worst possible take of all time, regardless of the actual topic.

 

[Revan]

xG is an inherently flawed statistic due to the fact that it assumes all professional footballers are of equal ability. They are clearly not. Unpredictable events such as human error, form, weather patterns and a multitude of other factors renders the entire system inconsistent and ultimately useless. I'm surprised it gets so much 'air time' around here.

538 considers also the quality of a player. Same chance to Messi and Welbeck gets weighted differently. In the end it is a model, there is no truth to give generalised opinions like 'xG are bad, good, etc'. xG is as good as the model and data from that company. Can be great and useful, can be bad and terrible.

 

[Sarni]

*Obama confused gif*

But it’s true. Basketball is full of stats that don’t really matter being quoted over a short period of time to prove some historical achievement or failure.

OMG player X is the first player who has got 10 ppg, 8 rpg, 5 apt and below 3.5 turnovers per game while eating chicken nuggets for dinner and drinking a soda 10 minutes before the game. This NEVER happened before.

 

[dmode]

xG has been created for teams like Liverpool who win feck all year in year out.

 

[Revan]

This is football not maths, hope this helps.

Neither is driving, hope this helps.

 

[Pexbo]

Neither is driving, hope this helps.

If a driverless car was made to the same standard as xG currently is, would you go for a spin on the motorway?

 

[Revan]

If a driverless car was made to the same standard as xG currently is, would you go for a spin on the motorway?

Nope. But I also wouldn't say that autonomous driving is bullshit, nothing beats a human driving a car. I would possibly continue that autonomous driving will be very useful, we just need to make it better.

 

[KirkDuyt]

This will have grave consequences for the hypothetical league table.

xG is dumb as feck. It's a stat that disregards the factors of luck and random chance. These factors are huge.

 

[ruddevil]

One thing for sure, xG isn't able to tell how excited I am during Solskjaer period and how bored to death I was during Mourinho's. To me, that's all that matters in football. I guess xG is useful for gamblers and fantasy leaguers.