XG is a terrible stat

Great thread, I didn't even read the OP but I agree so much with the title.

The fact that it's almost constantly used to highlight how it didn't correlate to the number of goals scored means that they really have no idea how to actually calculate how good chances were.

Also when discussing a player you've both watched and someone starts talking about their xG, I tend to immediately think that person is a bit of a wanker.
 
Load of bollocks. As is the notion that this thing, which seems to bear little resemblance to the game I just watched way more often than not, not to mention if you're ever bothered enough to drill down into individual chances, somehow becomes right over time due to the extra bollocks you pile on top of it.
 
xG or xA are meant to be used over large samples, it's not supposed to judge isolated actions but to give an image of how a player or a team finishes or creates compared to the average footballer/team(average in every ways possible). The best way to explain it would be that if a particular action has a 0.8 xG it means that the more attempts you have for the same action the results will tend toward 0.8, it doesn't mean that 8 of the next 10 attempts will result to a goal but people seem to think that it's the meaning of it.
 
Ole'sgunnarwin said:
If it's rubbish for one match, it's rubbish for the season. And it is absolutely awful.

It's like refereeing decisions evening out throughout the season, it's nonsense
Click to expand...

In what way?

I mean I play a bit of poker and I know if I go all in with two aces at the start of a hand I'm 80% likely to beat two kings.

In a cash game I should do this every time if I know my opponent will call with two kings. The 20% of the time I lose doesn't negate that. I could lose in such a situation 10 times in a row but it doesn't mean I wasn't doing the right thing.

xG tries to show us a similiar thing - who is doing the right thing in terms of creating the better quality chances (or who is gifting them to the oppo). Doesn't mean you'll win in a one-off game just because you had the better xG, or you can even lose a few in a row when you did. Your players might be wasteful on any given day, or the oppo goalie might do something other-worldly.

Now the xG models themselves aren't as spot on as the fixed mathematics of poker as no 2 chances are completely identical but they're not a million miles away most of the time.
 
AnotherLondonManc said:
Someone needs to post what the xg was over the entire PL season vs how many goals were scored (I cba)

If those figures match up, then it'll be fair to say that xg is a pretty reliable model. Just not in isolated incidents, it's more about averages
Click to expand...

The numbers match because it's literally data taken from the PL, there is no other way it could be.

One season is a smaller sample size than the entire data set so any deviation on a smaller sample is just that, a statistical deviation.
 
Ole'sgunnarwin said:
If it's rubbish for one match, it's rubbish for the season. And it is absolutely awful.

It's like refereeing decisions evening out throughout the season, it's nonsense
Click to expand...

you've managed to misunderstand Xg to a degree previously thought impossible
 
Ole'sgunnarwin said:
No, I've studied statistics, if there are 38 bullshit made up statistics, it means nothing at the end of the season
Click to expand...
Why would anybody call it a bullshit statistic. People are raving about the eyetest when most stats are nothing but watching by the eyetest but with a notepad counting stuff. With all due respect I struggle to believe that you studied statistics and not getting the use of this particular one.

Obviously, as with any stat, it isn't perfect and a creative user can create multiple stories using the same numbers. But that applies to every stat, heck, today was somebody posting in the post match thread for yesterdays game that a 3-1 game cant be too much of a struggle for us. But it was.

Eye test means, that you use your memories and personal experience to evaluate the probability of a shot ending up in goal. Doing this, every individual will make use of different aspects, positioning of the shooter, positioning of keeper and defenders, height of ball, speed the ball is traveling, pressure, angles whatever. You can add an infinite of number of factors if you think they need to taken into account - moon phases, zodiac signs, temperature, emotional state - if you think, you have the tools to measure those things, you can incorporate it into your model. The more factors you incorporate the more exact your probability can get. The perfect xG model would incorporate every aspect there is but obviously, we are far from that and probably will be forever because we will never be sure whether there is something we don't know about and aren't able to measure.
But that applies to football fans as well. Addtionally, many football fans wouldn't be able to articulate "their model" even though they will apply it without problems. Each individual though will apply and weight things differently. So eyetest is always subjective at the end of the day and stats are attempts to push back that subjectivity as much as possible for certain aspects of the game. Counting Goals scored doesn't leave any room for subjectivity in the question "who scored more goals", just as xG will try to answer the question how good the chances of a certain team in a specific game have been. Easy as that. And to do that taking into account actual goals scored over thousands of games, way way more than every individual could watch.

Whenever I see xG being called a stupid, or a bit as with the packing stat that was around for a while a couple of years ago, it is very shortsighted. Stats are just tools, some use of it might be stupid. Certainly some users can be stupid but the stat itself is what it is.

Also, there isn't one xG there are multiple different ones differing in attributes taking into account, numbers of games taken into consideration and so on.
 
Last edited:
I hate xg and the 'statification' of football in general.

It's just so 'trust me bro'. Does anyone know how it's calculated? Well, in very broad strokes but can you calculate it yourself? No. So how can you ever make a judgement on whether it's accurate or not?

Even the things it's calculated on, like quality of chance, positioning of other players, are subjective and not facts. Different xg models can't even agree with one another. So there isn't even a 'correct' or agreed method of calculating it.

And when the result is wildly different to the xg, just say 'oh wow they're under/over performing their underlying stats'. Or y'know....maybe the stats just aren't very reliable?

It's like religion in that you have to have faith in what you're being told is true, because you can't quantify it for yourself.
 
OverratedOpinion said:
Great thread, I didn't even read the OP but I agree so much with the title.

The fact that it's almost constantly used to highlight how it didn't correlate to the number of goals scored means that they really have no idea how to actually calculate how good chances were.

Also when discussing a player you've both watched and someone starts talking about their xG, I tend to immediately think that person is a bit of a wanker.
Click to expand...
That’s really not how probabilities work.
 
xG is an awesome stat, continuously getting better as more data is added. If you want to look at it for an individual thing, it's just as subjective as the eye test, but to show trends it is great.

I would say, specific to Antony's chance, the issue is his weird choice to go to ground likely hoodwinks the parameters. It was, essentially, a ball he could just run onto and tap in but he slid in and so that will effect the % as it will compare to other goals from the same area where a player has gone to ground to score. xG does not account for moronic decision making, it's just about how the chance is created then executed.
 
miked99 said:
I hate xg and the 'statification' of football in general.

It's just so 'trust me bro'. Does anyone know how it's calculated? Well, in very broad strokes but can you calculate it yourself? No. So how can you ever make a judgement on whether it's accurate or not?

Even the things it's calculated on, like quality of chance, positioning of other players, are subjective and not facts. Different xg models can't even agree with one another. So there isn't even a 'correct' or agreed method of calculating it.

And when the result is wildly different to the xg, just say 'oh wow they're under/over performing their underlying stats'. Or y'know....maybe the stats just aren't very reliable?

It's like religion in that you have to have faith in what you're being told is true, because you can't quantify it for yourself.
Click to expand...

It's fairly simple and in theory you can calculate it yourself, you just need to compile dozens of thousands of shots and calculate how many were scored or missed. That's the very broad description.

The only thing that creates a "new" margin of error is that you have to subjectively determine similarities between separate actions in order to group them and I guess some will use it as an argument to invalidate it.
 
NZT-One said:
Why would anybody call it a bullshit statistic. People are raving about the eyetest when most stats are nothing but watching by the eyetest but with a notepad counting stuff. With all due respect I struggle to believe that you studied statistics and not getting the use of this particular one.

Obviously, as with any stat, it isn't perfect and a creative user can create multiple stories using the same numbers. But that applies to every stat, heck, today was somebody posting in the post match thread for yesterdays game that a 3-1 game cant be too much of a struggle for us. But it was.

Eye test means, that you use your memories and personal experience to evaluate the probability of a shot ending up in goal. Doing this, every individual will make use of different aspects, positioning of the shooter, positioning of keeper and defenders, height of ball, speed the ball is traveling, pressure, angles whatever. You can add an infinite of number of factors if you think they need to taken into account - moon phases, zodiac signs, temperature, emotional state - if you think, you have the tools to measure those things, you can incorporate it into your model. The more factors you incorporate the more exact your probability can get. The perfect xG model would incorporate every aspect there is but obviously, we are far from that and probably will be forever because we will never be sure whether there is something we don't know about and aren't able to measure.
But that applies to football fans as well. Addtionally, many football fans wouldn't be able to articulate "their model" even though they will apply it without problems. Each individual though will apply and weight things differently. So eyetest is always subjective at the end of the day and stats are attempts to push back that subjectivity as much as possible for certain aspects of the game. Counting Goals scored doesn't leave any room for subjectivity in the question "who scored more goals", just as xG will try to answer the question how good the chances of a certain team in a specific game have been. Easy as that. And to do that taking into account actual goals scored over thousands of games, way way more than every individual could watch.

Whenever I see xG being called a stupid, or a bit as with the packing stat that was around for a while a couple of years ago, it is very shortsighted. Stats are just tools, some use of it might be stupid. Certainly some users can be stupid but the stat itself is what it is.

Also, there isn't one xG there are multiple different ones differing in attributes taking into account, numbers of games taken into consideration and so on.
Click to expand...
I don't think you understand me. I don't qualify it as a statistic. I think it's too opinion based to take it seriously. I understand things average out but only if it's reliable data which I don't believe this is.
 
Ole'sgunnarwin said:
I don't think you understand me. I don't qualify it as a statistic. I think it's too opinion based to take it seriously. I understand things average out but only if it's reliable data which I don't believe this is.
Click to expand...

Why is it too opinion based to take it seriously? Do you have examples showing that?
 
Ole'sgunnarwin said:
I don't think you understand me. I don't qualify it as a statistic. I think it's too opinion based to take it seriously. I understand things average out but only if it's reliable data which I don't believe this is.
Click to expand...
I can't follow that argumentation. The provider of xG took multiple aspects of a potential situation in a match into account and looked how often a shot with this particular set of attributes went in to end up with for example an 0.18 xG value for that particular shot. So when you end up in a situation where this specific set of attributes apply and you shoot, then you have a probability of 18% chance of scoring based on that particular model. It is nothing but averages, the only difference is probably the number of aspects taken into account and probably the weighting of those and the number of matches that have been processed to list out such situations.

How is that opinion based?

The easiest xG model would just look at where a player is on the field when he takes a shot. Would then check 50k matches how often a player has been shooting from this position and how many goals have been scored. If it was attempted 250 times and 50 shots ended up in goal, this xG model comes up with a 0.2 xG for that situation. There is no opinion there at all.
 
KiD MoYeS said:
I have been Googling whether clubs use xG as a metric in analysing their own players performance or transfer targets and can't find anything. That to me suggests it's not considered much.
Click to expand...
Have heard managers talk about it openly multiple times.. guys like Tuchel, even Ten Hag..
 
acnumber9 said:
My issue with it is that more teams should be outperforming it. If they aren’t, it’s not a real average. If you look at the stats here:

https://understat.com/league/EPL

Only three teams have outperformed it this season, Wolves, Brentford and Aston Villa. Surely the teams with the best players should be performing above average? It’s bollocks.
Click to expand...
Just click through different seasons. Values are collected over a longer period of time than just 1 season and don't include just Prem, but a wide range of competitions. There's of course gonna be error in the values as no 2 incidents are identical (other than penalties which is a straightforward calculation). Teams seem to be finishing poorly this season. Previous seasons sometimes you had most teams over perform it, or a 50/50 mix... It varies
 
bosnian_red said:
Just click through different seasons. Values are collected over a longer period of time than just 1 season and don't include just Prem, but a wide range of competitions. There's of course gonna be error in the values as no 2 incidents are identical (other than penalties which is a straightforward calculation). Teams seem to be finishing poorly this season. Previous seasons sometimes you had most teams over perform it, or a 50/50 mix... It varies
Click to expand...
I looked at Salah’s stats and he’s underperformed in the last three seasons. I don’t think you’d find anybody who would say he’s a below average finisher.
 
The issue might be with the name given to the stat.

Expected goal maybe sounds a bit predictive/prescriptive/opinionated, whereas if it had a name that simply described what it calculates (percentage of similar-ish chances scored) it would be more understandable.
 
acnumber9 said:
I looked at Salah’s stats and he’s underperformed in the last three seasons. I don’t think you’d find anybody who would say he’s a below average finisher.
Click to expand...

Unless your xG by some crazy coincidence is calculated as exactly x.00, you'll never hit it exactly. In the seasons you're looking at his goal output is very similar to his xG, like within a goal or two. It's so close that if you just look at a different model, with just slightly different numbers, what you're saying is no longer true.

That is average finishing, not below average.
 
NotThatSoph said:
Unless your xG by some crazy coincidence is calculated as exactly x.00, you'll never hit it exactly. In the seasons you're looking at his goal output is very similar to his xG, like within a goal or two. It's so close that if you just look at a different model, with just slightly different numbers, what you're saying is no longer true.

That is average finishing, not below average.
Click to expand...
It’s factually below the average but if it makes you happier we’ll call it average. Do you believe that Salah is an average finisher? If not, I don’t see how the xG can be considered an average.
 
bosnian_red said:
Just click through different seasons. Values are collected over a longer period of time than just 1 season and don't include just Prem, but a wide range of competitions. There's of course gonna be error in the values as no 2 incidents are identical (other than penalties which is a straightforward calculation). Teams seem to be finishing poorly this season. Previous seasons sometimes you had most teams over perform it, or a 50/50 mix... It varies
Click to expand...

One thing about xG is that it still needs to be interpreted. Some players/teams are supposed to overperform it, others are supposed to match it and others are supposed to underperform because there is a difference of quality between players when xG is based on the average outcome for all players recorded regardless of status.

So even for a penalty, there should be a different outcome depending on who takes the penalty. The average penalty taker is going to match xG, the best penalty takers will outperform it and bad penalty takers will underperform. I have seen people state that x should have scored because the xG was y but they don't consider that x is probably below average compared to the population of players recorded for the xG, as an example most shots in the box will come from strikers, the xG is highly influenced by strikers which means that if you have a CB in a similar situation it's likely that he won't match the xG.

And as far as I know few, if any, xG sources discriminate per player position. And even if they did the samples would be far smaller for certain positions and therefore inaccurate
 
Last edited:
Jotun said:
First let me say, I love stats, and I love to measure as much as possible with stats, however finishing and particularly XG is currently terrible stat and shows very little.

Why I'm saying this, well Antony's chance was yesterday evaluated as 0.42. It was an open goal from 4 yards.

Diallo's goals are all evaluated as 0.33, 0.33 and 0.39. Actually, first two goals are quite difficult finishes, the first one from a tight corner on his weaker foot. The second one from a lobbed ball first time finish. I mean they are good chances, but certainly not trivial. However the last goal that is evaluated at 0.39???? Are you telling me there is a 60% chance that a professional footballer will miss an open goal under no pressure and with no sight of goalkeeper.

Are you telling me seriously that Matheus Fernandes chance is more likely to score than Amad his last or Antony? And that chance, despite him being surrounded with two united players and a goalkeeper straight ahead of him is apparently the best chance of game. Really? Really?

Source of XG values:
https://understat.com/match/26811

What this shows is the model (at least understat's) is flawed (probably too simple) and therefor can not be relied upon to provide actual information about quality of chances over a match. That also implies it is unreliable over the course of a season. Just because something feels wrong
Click to expand...

Firstly, don't use Understat, they're regularly off. Both Fotmob and Fbref have an xG of 0.7 for that last goal.

Secondly, humans are absolutely awful at judging statistics intuitively. There are literally books written on this subject. So just because a number "feels" wrong to you means pretty much nothing.
 
quadrant said:
Firstly, don't use Understat, they're regularly off. Both Fotmob and Fbref have an xG of 0.7 for that last goal.

Secondly, humans are absolutely awful at judging statistics intuitively. There are literally books written on this subject. So just because a number "feels" wrong to you means pretty much nothing.
Click to expand...
You realise the massive contradiction there?
 
I think its okay, not great or bad for evaluating teams performances.

I think its bad for correctly evaluating one player though. Most times players have strengths finishing one type of chance and are weak with another type of chance. If they are evaluated by how often the chances are missed by all players, then their own strengths and weaknesses are not taken into account, or rather every players are and so theirs count for next to nothing.

I always think back to Saha, which was also my first thought when I first heard about the stat. We'd put him through on goal and he wouldnt finish many of the chances despite that being considered a great chance. Then he'd get the ball in the box, but towards the edge of the 18 yard box and he'd smash it into the corner. Most people would call it a harder chance, but its what Saha was good at and where he got most of his goals. But his xG would be based on all players, not just his own data. So if he goes through and misses he'll probably have a high xg and lower goals. And if he scores from his shot just inside the box he'll probably have a lower xg and higher goals.
 
quadrant said:
Firstly, don't use Understat, they're regularly off. Both Fotmob and Fbref have an xG of 0.7 for that last goal.

Secondly, humans are absolutely awful at judging statistics intuitively. There are literally books written on this subject. So just because a number "feels" wrong to you means pretty much nothing.
Click to expand...

Also the "eye test" as people often mention is another terrible one. Even in court its seen as unreliable as it has many biases and memory factors
 
It's hard to hate something as benign as statistics but their popularity does make talking about the game less fun. Like VAR it's all about being the least incorrect, any chat about a player or team risks the shadow of imaginary numbers and Jon McKenzie coming in to explain that they aren't actually doing as well as they look.
 
acnumber9 said:
I looked at Salah’s stats and he’s underperformed in the last three seasons. I don’t think you’d find anybody who would say he’s a below average finisher.
Click to expand...
Salah is a great scorer because he is constantly getting chances and "accumulating xG". If he's consistently underperforming xG then by definition he isn't a good finisher. You can be a bad finisher but still be a great goalscorer. The way I see it is finishing is just the ability the take the chance that you are in. Salah does an incredible job at always getting chances, either through his movement or creating them himself or his team. Evidently he's a bit average at taking them once there.
 
quadrant said:
Firstly, don't use Understat, they're regularly off. Both Fotmob and Fbref have an xG of 0.7 for that last goal.

Secondly, humans are absolutely awful at judging statistics intuitively. There are literally books written on this subject. So just because a number "feels" wrong to you means pretty much nothing.
Click to expand...
Also to add - both fotmob and fbref are just free to use stats services, so they won't have the best one either. Statsbomb probably has the most sophisticated data collection but they aren't free to use, but will be closer to what actual clubs use (or actual clubs do use them, but also use more advanced metrics which xG was basically the start of).
 
bosnian_red said:
Also to add - both fotmob and fbref are just free to use stats services, so they won't have the best one either. Statsbomb probably has the most sophisticated data collection but they aren't free to use.
Click to expand...
Fbref gets it stats from Statsbomb. And Fotmob get their stats from Opta.

Scratch that, they moved and now both get it from Opta.
 
acnumber9 said:
It’s factually below the average but if it makes you happier we’ll call it average. Do you believe that Salah is an average finisher? If not, I don’t see how the xG can be considered an average.
Click to expand...

Sure.

Finishing is one of the easiest skills for a footballer to train, and a much less complicated one to master compared to other attacking actions. It would be pretty scandalous if in a multi billion industry the vast majority of professionals weren't evenly clustered.

The actual hard part is getting the opportunities to finish chances in a match, and to make it worse it's a very hard and much more time consuming thing to accurately practice in a way that mirrors a real game, so that's the area you'd expect a big difference between the best and the rest.
 
JPRouve said:
On thing about xG is that it still needs to be interpreted. Some players/teams are supposed to overperform it, others are supposed to match it and others are supposed to underperform because there is a difference of quality between players when xG is based on the average outcome for all players recorded regardless of status.

So even for a penalty, there should be a different outcome depending on who takes the penalty. The average penalty taker is going to match xG, the best penalty takers will outperform it and bad penalty takers will underperform. I have seen people state that x should have scored because the xG was y but they don't consider that x is probably below average compared to the population of players recorded for the xG, as an example most shots in the box will come from strikers, the xG is highly influenced by strikers which means that if you have a CB in a similar situation it's likely that he won't match the xG.

And as far as I know few, if any, xG sources discriminate per player position. And even if they did the samples would be far smaller for certain positions and therefore inaccurate
Click to expand...
Yeah, similar situation to the Bruno goal vs Arsenal and the Maguire miss vs Liverpool. Similar chances, but you'd expect Bruno to connect sweetly with it while it's a harder chance for Maguire (even if the xG would probably be higher on Maguire's).
 
quadrant said:
Fbref gets it stats from Statsbomb. And Fotmob get their stats from Opta.

Scratch that, they moved and now both get it from Opta.
Click to expand...
Yeah fbref stopped pulling from statsbomb few years ago think it was cost related. It's still good, but not as sophisticated and you'll have more errors in it.
 
miked99 said:
I hate xg and the 'statification' of football in general.

It's just so 'trust me bro'. Does anyone know how it's calculated? Well, in very broad strokes but can you calculate it yourself? No. So how can you ever make a judgement on whether it's accurate or not?

Even the things it's calculated on, like quality of chance, positioning of other players, are subjective and not facts. Different xg models can't even agree with one another. So there isn't even a 'correct' or agreed method of calculating it.

And when the result is wildly different to the xg, just say 'oh wow they're under/over performing their underlying stats'. Or y'know....maybe the stats just aren't very reliable?

It's like religion in that you have to have faith in what you're being told is true, because you can't quantify it for yourself.
Click to expand...

So is the eye test!

At least statistics looks at every game. How many people here form their opinion on watching every single PL game this season?
 
You must log in or register to reply here.