A question about XG and how it works

[padzilla]

I have a broad idea about how it is calculated. Is it the likelihood of a player scoring a goal when they take a shot?

The XG is higher depending on the angle of the shot and how far from goal?

But who decides on the XG value of a chance? Is there like a committee assigning value to chances, which makes it sound like it's someone's opinion rather than an exact stat.

Also does opportunities like were a player only needs to make to a simple pass for an assist, when it's 3 on 2, but makes the wrong choice have an XG rating?

What about crosses that are somehow scrambled away or goalmouth drama where the defenders make last ditch clearances but an attacking player doesn't connect have an XG rating?

 

[AndersB]

It's a board consisting of the Glazers, Florentino Pérez and Mike Ashley. No wonder it's such a shit stat

 

[Tarrou]

I have a broad idea about how it is calculated. Is it the likelihood of a player scoring a goal when they take a shot?

The XG is higher depending on the angle of the shot and how far from goal?

But who decides on the XG value of a chance? Is there like a committee assigning value to chances, which makes it sound like it's someone's opinion rather than an exact stat.

Also does opportunities like were a player only needs to make to a simple pass for an assist, when it's 3 on 2, but makes the wrong choice have an XG rating?

What about crosses that are somehow scrambled away or goalmouth drama where the defenders make last ditch clearances but an attacking player doesn't connect have an XG rating?

its based on data from previous chances in similar situations

there is only an xG rating if there was a shot

 

[Jev]

I have a broad idea about how it is calculated. Is it the likelihood of a player scoring a goal when they take a shot?

The XG is higher depending on the angle of the shot and how far from goal?

Yeah.

But who decides on the XG value of a chance? Is there like a committee assigning value to chances, which makes it sound like it's someone's opinion rather than an exact stat.

The xG of a given chance is based on statistical analysis of thousands of similar situations. I.e. roughly 85 percent of penalties are scored, so a penalty is about 0.85 xG.

Also does opportunities like were a player only needs to make to a simple pass for an assist, when it's 3 on 2, but makes the wrong choice have an XG rating?

What about crosses that are somehow scrambled away or goalmouth drama where the defenders make last ditch clearances but an attacking player doesn't connect have an XG rating?

There are other measures for that but no, xG looks only at the shot that was taken.

 

[lysglimt]

At least it's not as bad as "Shots on target" - where a team can hit the woodwork 7 times and miss an open goal 4 times - and according to Shots on Target - they have none. Whereas the other team with 5 shots from 40 yards out which rolls into the hands of the goalkeeper have 5 shots on target

Why not simply use goal attempts and drop that shots on target-crap

 

[Tarrou]

At least it's not as bad as "Shots on target" - where a team can hit the woodwork 7 times and miss an open goal 4 times - and according to Shots on Target - they have none. Whereas the other team with 5 shots from 40 yards out which rolls into the hands of the goalkeeper have 5 shots on target

Why not simply use goal attempts and drop that shots on target-crap

it's just data there is nothing bad about it

if some people decide to use at as a measure of performance (similar to xG) that's not the datas fault

 

[tenpoless]

I think if you take all the shots so far and segment them based on which one ended as a goal or not, you can work out a probability based on those past data to figure out how likely a shot is going to end up in the goal. So yeah its just a probability i reckon. You can count any probability if you have enough past data.

 

[vulmik]

Also does opportunities like were a player only needs to make to a simple pass for an assist, when it's 3 on 2, but makes the wrong choice have an XG rating?

What about crosses that are somehow scrambled away or goalmouth drama where the defenders make last ditch clearances but an attacking player doesn't connect have an XG rating?

Xg only accounts for shots taken. We had a moment yesterday in the first half when Casemiro headed a free kick back across goal which probably should have been scored tbh but as the pass from Casemiro didn't find a teammate and result in a shot, the Xg for that was 0.

 

[Laurencio]

It's a statistical analysis of similar situations and shots based on historical data over x-amount of years.

Think of it as a metric that highlights how often an average side should score in a given chance. It is a chance by chance statistic. It is not an accurate calculation of how many goals a player/team should score.

A problem we have when talking about xG is that instead of looking at the quality of individual chances, we have a tendency to look at a summation of xG - i.e 10 x 0.1 chances = a goal. Which is simplistic and just plain wrong. It is an analytical tool to help determine if a team is creating good chances and to determine if a player is capable of putting away high quality chances regularly.

Some more info on it: https://statsbomb.com/soccer-metrics/expected-goals-xg-explained/

 

[Borys]

I have a broad idea about how it is calculated. Is it the likelihood of a player scoring a goal when they take a shot?

The XG is higher depending on the angle of the shot and how far from goal?

But who decides on the XG value of a chance? Is there like a committee assigning value to chances, which makes it sound like it's someone's opinion rather than an exact stat.

Any shot that has xG of 0,5 or higher has been taken thousands of times from the same place and the same angle, so that gives you a decent likelihood of scoring a goal.
Any shot that is close to 0,01 (like Bruno yesterday) doesn't really matter in the end, because how many goals like that can you score in a season?

The whole idea behind xG is that over a large sample, most players are "average" at both finishing (strikers stat) and shot stopping (goalkeepers stat). So what really is interesting are the outliers.

Another aspect to look at is the aggregated xG, best strikers tend to have very high xG over a season, that also speeks highly about the team/setup/coaching. Or low xGA (xG of the opponent). Man City have been outliers on that stat and they were winning the league.

Also does opportunities like were a player only needs to make to a simple pass for an assist, when it's 3 on 2, but makes the wrong choice have an XG rating?

What about crosses that are somehow scrambled away or goalmouth drama where the defenders make last ditch clearances but an attacking player doesn't connect have an XG rating?

That is not a scoring opportunity. Million things might have gone wrong, the pass could be bad, the stiker might have been too fast/slow/tripped himself, defender might be able to get in between.

 

[padzilla]

Good responses, thanks folks. A personal grip of mine is the premium value that assists have in the modern game, players can get huge recognition for simply passing it to a team mate, who then goes around four players and slams it in the top corner from 30 yards out.

Whereas a player can go around four players and put it on a plate for a team mate who makes a mess of it. No assist is given in this situation. I am not convinced it's not a hangover from fantasy football's integration into the life of most football supporters.

 

[Zehner]

Good responses, thanks folks. A personal grip of mine is the premium value that assists have in the modern game, players can get huge recognition for simply passing it to a team mate, who then goes around four players and slams it in the top corner from 30 yards out.

Whereas a player can go around four players and put it on a plate for a team mate who makes a mess of it. No assist is given in this situation. I am not convinced it's not a hangover from fantasy football's integration into the life of most football supporters.

There's also a stat called "expected assists" that measures how many xG you set up for your team mates. So serving it up on a plate for a team mate contributes more to that than an easy pass to a goal from 30 yards out. But still the team mate needed to take a shot - if he fails to get a shot off, it will still be 0.

I also like xGC (expected goals chain). It measures in how much xG you've been involved i some capacity, e. g. also by winning the ball and passing it to a midfielde or playingthe through ball before the cut back etc.

 

[Borys]

It's a statistical analysis of similar situations and shots based on historical data over x-amount of years.

Think of it as a metric that highlights how often an average side should score in a given chance. It is a chance by chance statistic. It is not an accurate calculation of how many goals a player/team should score.

A problem we have when talking about xG is that instead of looking at the quality of individual chances, we have a tendency to look at a summation of xG - i.e 10 x 0.1 chances = a goal. Which is simplistic and just plain wrong. It is an analytical tool to help determine if a team is creating good chances and to determine if a player is capable of putting away high quality chances regularly.

Some more info on it: https://statsbomb.com/soccer-metrics/expected-goals-xg-explained/

It's the variation that confuses people. Understanding the probability analysis is difficult if you never played with this. You might have rolled a "6" twice in a row, but still if you do it a thousand times the "6" will show only 167 times. If it occurs 150 times, than maybe in next 1000 throws you will will roll a "6" 184 times (again, this is over-simplified but you get the idea). Haaland overscored his xG last season by 6 goals, this season he underperformed by 3 goals - in the end what matters is how much xG he actually accumulates, and that stat is even more important for the team as a whole. THis is the reason many of us expected a downfall in our results a while ago and so many people actually never truly believed in us getting 5th.

 

[90 + 5min]

I have a broad idea about how it is calculated. Is it the likelihood of a player scoring a goal when they take a shot?

The XG is higher depending on the angle of the shot and how far from goal?

But who decides on the XG value of a chance? Is there like a committee assigning value to chances, which makes it sound like it's someone's opinion rather than an exact stat.

Also does opportunities like were a player only needs to make to a simple pass for an assist, when it's 3 on 2, but makes the wrong choice have an XG rating?

What about crosses that are somehow scrambled away or goalmouth drama where the defenders make last ditch clearances but an attacking player doesn't connect have an XG rating?

It is no stats to depend on in my opinion. Just something that statistic guys like. It doesn't take into account things like home, away grounds, day, afternoon or night games, weather and all those little stats that also matters. It doesn't take into account if it is defender or attacker who is taking the shot and so on. If you want it to be correct. That is why I really don't get why some people love to point out xG like it is some great thing.

 

[Pogue Mahone]

Good responses, thanks folks. A personal grip of mine is the premium value that assists have in the modern game, players can get huge recognition for simply passing it to a team mate, who then goes around four players and slams it in the top corner from 30 yards out.

Whereas a player can go around four players and put it on a plate for a team mate who makes a mess of it. No assist is given in this situation. I am not convinced it's not a hangover from fantasy football's integration into the life of most football supporters.

If that player had converted the chance put on a plate for him then his goal and the goal from the guy who beat four players would be valued identically. Does this also bother you?

The whole point of counting assists (and goals) is that over a season, there will be a mix of spawny ones and quality ones. So the players who rack up the most are invariably the best players at creating or scoring goals.

The main source of bias here is that the players who happen to be playing for the best teams have a big advantage. But they’re only playing for those teams because they’re good players, so that’s probably fair enough.

 

[Borys]

It is no stats to depend on in my opinion. Just something that statistic guys like. It doesn't take into account things like home, away grounds, day, afternoon or night games, weather and all those little stats that also matters. It doesn't take into account if it is defender or attacker who is taking the shot and so on. If you want it to be correct. That is why I really don't get why some people love to point out xG like it is some great thing.

This is not correct. xG of 0,25 means on average (in THOUSANDS of shots taken from that position, home/away, day, afternoon or night games, weather etc) you will score 1 in 4 for all attempts. All those factors you mentioned are already included in that parameter.

The only problem is if you look at individual instance and then say it doesn't really reflect the reality because you scored 2 goals from 4 shots with xG of 0,25. In the bigger sample, number of goals will often correlate with xG parameters, or will be a decent predictor for the future.

 

[tomaldinho1]

It's the variation that confuses people. Understanding the probability analysis is difficult if you never played with this. You might have rolled a "6" twice in a row, but still if you do it a thousand times the "6" will show only 167 times. If it occurs 150 times, than maybe in next 1000 throws you will will roll a "6" 184 times (again, this is over-simplified but you get the idea). Haaland overscored his xG last season by 6 goals, this season he underperformed by 3 goals - in the end what matters is how much xG he actually accumulates, and that stat is even more important for the team as a whole. THis is the reason many of us expected a downfall in our results a while ago and so many people actually never truly believed in us getting 5th.

Then add in you have multiple models and the fact the data is continuously evolving and you see a fraction of the complexity.

My issue with xG is these companies, as far as I can tell, don't release info on how they are dividing the pitch or how they factor in the player. So let's say Halaand shoots from near the penalty spot twice and the ball is in the exact same spot to the nanometre, one is him striding onto the ball with no one around him in full control and the other is him at full stretch, off balance about to be clattered by some unit like Burn. The contact on the ball is in the same exact spot but they are clearly two very different chances. Would xG would have these as equal opportunities?

Take the same example and add in the more realistic scenario that the ball isn't in the exact same spot, it's 5cm higher and a few cm wider, then Haaland also isn't in the same spot or at the same fitness level, he likely has different boots on and he grass on the pitch might be longer, the air more humid etc. etc. and again it's a slightly different chance. Do we know how any of these companies actually segment the pitch for where the contact is made.

 

[711]

I also like xGC (expected goals chain). It measures in how much xG you've been involved i some capacity, e. g. also by winning the ball and passing it to a midfielde or playingthe through ball before the cut back etc.

Not to be confused with xGC (expected goals conceded).

I know you wouldn't but it might annoy someone

 

[lysglimt]

it's just data there is nothing bad about it

if some people decide to use at as a measure of performance (similar to xG) that's not the datas fault

I am mostly referring to the stats in games showing "Shots on Goals" - more than xG

 

[Tarrou]

I am mostly referring to the stats in games showing "Shots on Goals" - more than xG

Yeah I got that. What you're (seemingly) taking issue with is people reading too much into the data, not the data itself. Right?

 

[TheRedDevil'sAdvocate]

Then add in you have multiple models and the fact the data is continuously evolving and you see a fraction of the complexity.

My issue with xG is these companies, as far as I can tell, don't release info on how they are dividing the pitch or how they factor in the player. So let's say Halaand shoots from near the penalty spot twice and the ball is in the exact same spot to the nanometre, one is him striding onto the ball with no one around him in full control and the other is him at full stretch, off balance about to be clattered by some unit like Burn. The contact on the ball is in the same exact spot but they are clearly two very different chances. Would xG would have these as equal opportunities?

Take the same example and add in the more realistic scenario that the ball isn't in the exact same spot, it's 5cm higher and a few cm wider, then Haaland also isn't in the same spot or at the same fitness level, he likely has different boots on and he grass on the pitch might be longer, the air more humid etc. etc. and again it's a slightly different chance. Do we know how any of these companies actually segment the pitch for where the contact is made.

It's an ongoing process. Lately, some models have been adding Positional Adjusted and Player Adjusted xG. There's also the Post-Shot xG already in place, which puts more emphasis on placement and keeper position. It will get more precise with time, and we still don't know the exact depth of the analysis it can offer. But it's certainly not just a gimmick, as some people think. It's being monitored by the staff of football clubs even in lower European leagues nowadays, either for team or player improvement. Going back to your question, even without the further microanalysis, the sample size (as Borys alluded to) is so vast now (and more data is added daily) that it can be deemed safe enough to work with.

 

[lysglimt]

Yeah I got that. What you're (seemingly) taking issue with is people reading too much into the data, not the data itself. Right?

A bit of both - I think it's wrong to present data which doesn't say anything about the game. In my opinion - if you present data about chances - and one team has 10 and the other 3 - it should be a clear indication that one team looked the more likely to score. With shots on target - it can be wildly inaccurate

 

[tomaldinho1]

It's an ongoing process. Lately, some models have been adding Positional Adjusted and Player Adjusted xG. There's also the Post-Shot xG already in place, which puts more emphasis on placement and keeper position. It will get more precise with time, and we still don't know the exact depth of the analysis it can offer. But it's certainly not just a gimmick, as some people think. It's being monitored by the staff of football clubs even in lower European leagues nowadays, either for team or player improvement. Going back to your question, even without the further microanalysis, the sample size (as Borys alluded to) is so vast now (and more data is added daily) that it can be deemed safe enough to work with.

I'm a fan of xG by the way but I think you need it with other stats, not just to say look, this team has a higher xG than the other and is therefore better. It's certainly more useful than shots and shots on target etc.

 

[Tarrou]

A bit of both - I think it's wrong to present data which doesn't say anything about the game. In my opinion - if you present data about chances - and one team has 10 and the other 3 - it should be a clear indication that one team looked the more likely to score. With shots on target - it can be wildly inaccurate

who decides what a chance is? Or do you just mean a shot or what?

 

[Garnacho's Shoelaces]

It's a statistical analysis of similar situations and shots based on historical data over x-amount of years.

Think of it as a metric that highlights how often an average side should score in a given chance. It is a chance by chance statistic. It is not an accurate calculation of how many goals a player/team should score.

A problem we have when talking about xG is that instead of looking at the quality of individual chances, we have a tendency to look at a summation of xG - i.e 10 x 0.1 chances = a goal. Which is simplistic and just plain wrong. It is an analytical tool to help determine if a team is creating good chances and to determine if a player is capable of putting away high quality chances regularly.

Some more info on it: https://statsbomb.com/soccer-metrics/expected-goals-xg-explained/

But 0.1xG shots are scored on average one in every ten shots. So ten shots of that quality will lead to one goal, on average, based on the historical data.

 

[Chipper]

I am mostly referring to the stats in games showing "Shots on Goals" - more than xG

Yes, I suppose shots on target is somewhat redundant now that xG exists as a measure.

It does work to a degree and was once the best we had. The top 10 teams for shots on target are, in order: City, Liverpool, Arsenal, Brighton & Spurs equal 4th, Villa, Newcastle, Chelsea, United, Bournemouth. That's not a million miles away from the league table and is definitely better than random. There's strong correlation with 9 of the top 10 for shots on target being 9 of the top 10 actual teams. They're in a different order, but even then the top 3 are top 3, Villa and Spurs roughly the same as in the league table, and United, Newcastle and Chelsea towards the lower end of the top 10 of the league table and the shots on target table. That's not even looking at defence at all.

The one team outside of the top 10 for for most shots on target (Bournemouth with 10th most), are 12th in the league so they're close to actual league position as well.

Within individual matches it's less reliable. Things get fuzzy but that's true of all stats. Having more shots on target doesn't mean you played beter or are better than your opponent over the long run, "Winning the xG" in a match doesn't tell you the whole story either, and neither does the actual scoreline. Still, as a general rule or guideline it's always been true that more shots on target = more chance of winning an individual match. Well, sort of.

No team has ever thought to just pepper the opposition with as many long range shots as possible as their main tactic to get that shots on target number up. That would be silly as the shots would include lots that didn't have much chance of going in. There might be times they want to do it a bit more than usual (dodgy 'keeper, wet conditions) but they're not making it their entire aim. More shots on target is just often a function of playing better most of the time, with the full range of all different types of shots included in there.

Now we have xG to give more context shots on target is not really needed. Still, it wasn't and isn't completely useless. With no knowledge of a league at all, armed with only a shots on target table and not even the actual league table you could get a decent idea of who the better teams are. You could then go on to predict which teams would have the most shots on target in indvidual games and go on to make far better predictions for the actual results compared to guessing completely randomly with no data at all. It's limited though and xG shits on it of course.

 

[Borys]

Then add in you have multiple models and the fact the data is continuously evolving and you see a fraction of the complexity.

My issue with xG is these companies, as far as I can tell, don't release info on how they are dividing the pitch or how they factor in the player. So let's say Halaand shoots from near the penalty spot twice and the ball is in the exact same spot to the nanometre, one is him striding onto the ball with no one around him in full control and the other is him at full stretch, off balance about to be clattered by some unit like Burn. The contact on the ball is in the same exact spot but they are clearly two very different chances. Would xG would have these as equal opportunities?

Take the same example and add in the more realistic scenario that the ball isn't in the exact same spot, it's 5cm higher and a few cm wider, then Haaland also isn't in the same spot or at the same fitness level, he likely has different boots on and he grass on the pitch might be longer, the air more humid etc. etc. and again it's a slightly different chance. Do we know how any of these companies actually segment the pitch for where the contact is made.

I don't see why you need to look at the individual cases though, nobody does that. It's just interesting to see why some teams tend to have high xG than others, and why some teams seem to limit opposition to low xG - that is a very good example because City might be doing less tackles/interceptions, but because they keep the ball so well and force opposition mistakes, they are effective at keeping the opposition at bay. This is exactly the opposite to what we're doing.

Again, nobody microanalyses individual and each shot, that's not what this stat is designed for.

Think of it like bow shooting. The further away from the target you are the less likely you score 100 points. That doesn't mean you can't score a goal from half line like Bruno did yesterday. But it's very clearly not a sustainable way to play football as teams with more xG / less xGA accumulated dominate in football and that is easy to prove on statistical level.

 

[RedfromIreland]

“Ffs, how did he miss that? That must’ve been an xg off 0.25.”
“I know Johnny, that’s twice he’s messed up xg’s more than 0.25. Do you want another pint?”

 

[Grande]

At least it's not as bad as "Shots on target" - where a team can hit the woodwork 7 times and miss an open goal 4 times - and according to Shots on Target - they have none. Whereas the other team with 5 shots from 40 yards out which rolls into the hands of the goalkeeper have 5 shots on target

Why not simply use goal attempts and drop that shots on target-crap

The only point about statistics, is making predictions beyond what results and observations give. So the most interesting stats are those that predict goals. Attempts at goal is a lot weaker predictor of goals than is Shots on target, and even more so than ExG. It’s down there with corner kicks.

The layman hysteria about attempts at goal recently is just that it is an extreme outlier and a simple narrative. Attempts stats are relevant, like possession stats, to say something about strategy. We already know that Utd this season struggles to dominate games. Even last year though, we had a strategy allowing for many shot attempts, but with second fewest goals conceded. Results have improved in 2024 after we changed to a Plan B strategy that allows even more attempts at goal as a trade off for reducing open chances. We are still not good, but alot better than in autumn.

The most interesting stats, showing that we are having a troubled season, are goals for/against, ExG for/against and shots on target for/against. We are mid table in all those counts, which rightly shows that we are not having a good season, but it’s not about shot attempts in general.

 

[Zehner]

“Ffs, how did he miss that? That must’ve been an xg off 0.25.”
“I know Johnny, that’s twice he’s messed up xg’s more than 0.25. Do you want another pint?”

Yeah, we should just ignore every statistic that becomes too complicated when you're drunk

 

[tomaldinho1]

I don't see why you need to look at the individual cases though, nobody does that. It's just interesting to see why some teams tend to have high xG than others, and why some teams seem to limit opposition to low xG - that is a very good example because City might be doing less tackles/interceptions, but because they keep the ball so well and force opposition mistakes, they are effective at keeping the opposition at bay. This is exactly the opposite to what we're doing.

Again, nobody microanalyses individual and each shot, that's not what this stat is designed for.

Think of it like bow shooting. The further away from the target you are the less likely you score 100 points. That doesn't mean you can't score a goal from half line like Bruno did yesterday. But it's very clearly not a sustainable way to play football as teams with more xG / less xGA accumulated dominate in football and that is easy to prove on statistical level.

Because that's what xG has become amongst fans - people look at individual chances and game's in isolation. I guess my issue isn't with the data it's with how people misuse it, not helped by how Sky use it now as the be all and end all stat for every game.

 

[Chipper]

What I find intersting about xG is that it's quite rare for teams or players to massively outperform or underperform their numbers over many seasons. There are a few that do it to a degree but in general the numbers are not far off and there's not many absolutely huge outliers.

Makes you think about finishing as a skill, and just how much the absolutely top strikers are better at converting their chances compared to some of the lesser players. The difference isn't nearly as big as I'd have personally thought before xG became a thing.

It shows the importance of chance creation for teams, and when it comes to forwards just how vital their movement is. The ability to find space, to make the right run so you can even have a shot to begin with etc. You can train that to an extent, and people have always commented about a striker's instinct, being able to sniff out a chance and things like that. I've always appreciated that it existed but it's bigger than I ever thought.

I never played football at any sort of high level as a kid. The best I did was 2 games for a weekend team and playing for my school. We never talked about movement at all. If you were perceived to be able to shoot well, you could be a striker.

There's actually a lot less difference between so-called good finishers and bad ones in the professional game than I ever realised.

Edit:

This season so far + last we have:

Haaland 55 league goals from 56 xG give or take.
Salah 36 from 41 xG
Watkins 33 from 35 xG

These are players paid a lot to score, to be there on the end of chances. How much xG any particular chance has includes all the data from shots made by defenders and midfielders who supposedly can't shoot very well at all from the same position, and all the "donkeys" playing up front for newly promoted teams. Of course forwards will take more shots, and those on better teams will take more shots too, so that skews the data when it comes to how much xG any individual chance is worth. Still, all those shots taken by full backs who can't hit a barn door from 6 yards are still in there and part of it.

Just gives me a better appreciation of the importance of finding the space to even have a shot. That's ultimately what gets you goals so much more than being able to finish well, which isn't really a big thing at all.

 

[RedfromIreland]

Yeah, we should just ignore every statistic that becomes too complicated when you're drunk

Aye I’m sure Fergie was fuming with xg when he was the best.
It must be very complicated when you’re sober as well, seeing there has been an in depth analysis on it’s merits on here.

 

[Zehner]

Aye I’m sure Fergie was fuming with xg when he was the best.
It must be very complicated when you’re sober as well, seeing there has been an in depth analysis on it’s merits on here.

So the best argument you can can up with is that you assume a manager who retired before xG became a thing wouldn't bother using it.

xG is used by countless top clubs. For example, an xG analyses was one of the main drivers behind Liverpool going for a certain Jürgen Klopp. The concept is completely logical and just an application of models that are proven to work perfectly fine for all kinds of usages. There really is no argument against it except for "meh, I hate everything new!"

 

[Powderfinger]

What I find intersting about xG is that it's quite rare for teams or players to massively outperform or underperform their numbers over many seasons. There are a few that do it to a degree but in general the numbers are not far off and there's not many absolutely huge outliers.

Makes you think about finishing as a skill, and just how much the absolutely top strikers are better at converting their chances compared to some of the lesser players. The difference isn't nearly as big as I'd have personally thought before xG became a thing.

It shows the importance of chance creation for teams, and when it comes to forwards just how vital their movement is. The ability to find space, to make the right run so you can even have a shot to begin with etc. You can train that to an extent, and people have always commented about a striker's instinct, being able to sniff out a chance and things like that. I've always appreciated that it existed but it's bigger than I ever thought.

I never played football at any sort of high level as a kid. The best I did was 2 games for a weekend team and playing for my school. We never talked about movement at all. If you were perceived to be able to shoot well, you could be a striker.

There's actually a lot less difference between so-called good finishers and bad ones in the professional game than I ever realised.

Edit:

This season so far + last we have:

Haaland 55 league goals from 56 xG give or take.
Salah 36 from 41 xG
Watkins 33 from 35 xG

These are players paid a lot to score, to be there on the end of chances. How much xG any particular chance has includes all the data from shots made by defenders and midfielders who supposedly can't shoot very well at all from the same position, and all the "donkeys" playing up front for newly promoted teams. Of course forwards will take more shots, and those on better teams will take more shots too, so that skews the data when it comes to how much xG any individual chance is worth. Still, all those shots taken by full backs who can't hit a barn door from 6 yards are still in there and part of it.

Just gives me a better appreciation of the importance of finding the space to even have a shot. That's ultimately what gets you goals so much more than being able to finish well, which isn't really a big thing at all.

Yes, in broad strokes this is one of the more important contributions of xG and similar metrics to our understanding of football.

Even the very best finishers in the world - Messi, Kane, and Son really - score only maybe 20-25% over their xG over large samples.

Lots of great strikers have scored right around their xG or just a little above.

Football is incredibly attached to the concept of being "clinical." People won't let that idea go because its such a compelling post-hoc narrative rationalization of outcomes (ie, you win when you're clinical...obviously). But very few players are actually clinical in a meaningful sense over large samples. A few may be like those mentioned above, and a few are sort of "anti-clinical" in that they really underperform xG over large samples (ie, Gabriel Jesus). But the vast majority are just in the messy middle, with their goal totals like 0-10% over their xG over many many seasons. And that makes the ability to generate or get on the end of chances - to anticipate where the ball will go, to beat opponents to the right spot, to get shots off in the box despite defenders being in front of you - much more important than being clinical overall.

 

[RedfromIreland]

So the best argument you can can up with is that you assume a manager who retired before xG became a thing wouldn't bother using it.

xG is used by countless top clubs. For example, an xG analyses was one of the main drivers behind Liverpool going for a certain Jürgen Klopp. The concept is completely logical and just an application of models that are proven to work perfectly fine for all kinds of usages. There really is no argument against it except for "meh, I hate everything new!"

There is an argument about it. I’m arguing about it. It’s a thing for nerds.
You either score goals or you don’t.
It may have some relevance for coaches but my point is 75,000 fans are not leaving Old Trafford wondering what the XG was.
You’ve got your opinion, I’ve got mine.
I don’t hate everything new by the way.

 

[Acheron]

It's a statistical analysis of similar situations and shots based on historical data over x-amount of years.

So it is like predicting the weather.

 

[padzilla]

I remember reading the game that people working in football data are given to study when they first take up the job is Germany beating Brazil 7-1 in the 2014 world cup semi final.

Apparently Brazil had more shots and possession but got hammered, flying in the face of the previous accepted wisdom that the number of shots and possession pointed to the superior team over the course of a match.

The new starts are encouraged to break down the game and go through it.

Apparently Germany gave away the least amount of fouls, throughout the tournament, in their own third as data showed a third of all goals are scored from set pieces.

Germany scored four goals in a 10 minute period of the first half in which they only made 25 passes during that same period.

The data showed German players would usually have the ball for more than 3 seconds before passing, it was decided to reduce to just around 1 second each, moving the ball much more quickly to keep opponents on the back foot - reducing the time in possession but making it much more effective.

 

[Zehner]

There is an argument about it. I’m arguing about it. It’s a thing for nerds.
You either score goals or you don’t.
It may have some relevance for coaches but my point is 75,000 fans are not leaving Old Trafford wondering what the XG was.
You’ve got your opinion, I’ve got mine.
I don’t hate everything new by the way.

It's not a thing for nerds, it is an objectively much better metric to judge which team had more chances. And I don't care how many fans are leaving Old Trafford not bothering about xG. It's a very easy to understand statistic and if you understand the sport, you get why it is relevant since it is just putting a number on something that is as old as football itself (which team had the better chances and deserved to win). If you aren't against it for the sake of it, that is.

 

[criticalanalysis]

At least it's not as bad as "Shots on target" - where a team can hit the woodwork 7 times and miss an open goal 4 times - and according to Shots on Target - they have none. Whereas the other team with 5 shots from 40 yards out which rolls into the hands of the goalkeeper have 5 shots on target

Why not simply use goal attempts and drop that shots on target-crap

it's just data there is nothing bad about it

if some people decide to use at as a measure of performance (similar to xG) that's not the datas fault

Don't hate the data, hate the player