Good thread this. xG is definitely of use to someone like myself who likes a bet and uses data to support or query what I think I’ve seen with my own eyes.
Half the point of xG is to not take into account who the chance falls to.
If Harry Kane and Glenn Murray both accrue chances worth a total of 20 xG over the course of a season and Harry Kane scores 26 goals while Murray scores 14 then you have evidence to back up why Harry Kane is one of the best strikers in the world because he is outperforming his xG number.
14 goals for Glenn Murray might look like a good season but if he's had the chances to score an expected 20 goals then maybe he's not as clinical as you first thought.
Maybe it's fault is in the name "expected goals" because it's not trying
to predict goals, it's more a rating system for quality of chance creation. 10 shots but
0.4 xG implies a load of long range pot shots while 5 shots and 2 xG
would suggest real guilt edge chances. It's just putting a number or a rating on something that used to be based only on gut feeling.
Tend to agree with Henry here. The Championship is an absolute killer of a league in how open it is. I’m not dismissing stats per se, just saying that I think the Championship doesn’t really abide by them. Certainly, this season has started by showing how open it is.
I struggle to reconcile how something subjective can be turned into hard data. A gilt-edged chance might have had a spinning ball that the observer couldn’t see, a bobble not taken into account or a flash of light bouncing from a watch in the crowd into the striker’s eyes. It simply might have been a lot harder than the XG hipster in the crowd could fathom.
I struggle to reconcile how something subjective can be turned into hard data. A gilt-edged chance might have had a spinning ball that the observer couldn’t see, a bobble not taken into account or a flash of light bouncing from a watch in the crowd into the striker’s eyes. It simply might have been a lot harder than the XG hipster in the crowd could fathom.
But that’s exactly how averages work, by using a large data set they smooth out the bumps and reduce the effect of outliers on the data interpretation
So this is an arbitrary stat? Who decides the ranking of each particular shot? One person may think an effort is an expected goal and another may have it at 50% chance of goal.
Its based on similar chances historically and how often they are converted.
Just seen this & great to answer to question of a few weeks back.
However, as the video says, XG takes no account who is actually taking the shot (or header)......so a bit pointless really. You could have Kane playing up front or Simon Church and the XG would show the same thing.
So this is an arbitrary stat? Who decides the ranking of each particular shot? One person may think an effort is an expected goal and another may have it at 50% chance of goal.
Its based on similar chances historically and how often they are converted.
Just seen this & great to answer to question of a few weeks back.
However, as the video says, XG takes no account who is actually taking the shot (or header)......so a bit pointless really. You could have Kane playing up front or Simon Church and the XG would show the same thing.
I'll leave you statto's & saddo's to it.....
As @Callumcafc said that's the point. It wouldn't work if each player had their own XG.
It's just another way to compare clubs that can indicate things such as which teams regularly create good chances or whether a team on a good run might not keep it up.
No one is saying stats are everything or that XG doesn't have any flaws.
Surely it it does matter who is having the shot. I would expect a certain player to score in certain situations and not expect another to score in exactly the same situation.
Funny how on the Nathan Jones thread there is a league table that compares the seasons results with the results from the expected goals and only 3 or 4 teams have had the same results after 5 games.
Low-budget team with idiot owner finds itself at the top of the Championship... I mean... I just don’t need hipster stats to tell me that’s a surprise we can’t expect to maintain!
You forgot - with the best manager and managment team in the league.
Surely it it does matter who is having the shot. I would expect a certain player to score in certain situations and not expect another to score in exactly the same situation.
Funny how on the Nathan Jones thread there is a league table that compares the seasons results with the results from the expected goals and only 3 or 4 teams have had the same results after 5 games.
Suggests the stat is not that accurate to me.
It's a shitty name but have a look at how its performed historically.... It's pretty damn accurate.
Half the point of xG is to not take into account who the chance falls to.
If Harry Kane and Glenn Murray both accrue chances worth a total of 20 xG over the course of a season and Harry Kane scores 26 goals while Murray scores 14 then you have evidence to back up why Harry Kane is one of the best strikers in the world because he is outperforming his xG number.
14 goals for Glenn Murray might look like a good season but if he's had the chances to score an expected 20 goals then maybe he's not as clinical as you first thought.
Maybe it's fault is in the name "expected goals" because it's not trying
to predict goals, it's more a rating system for quality of chance creation. 10 shots but
0.4 xG implies a load of long range pot shots while 5 shots and 2 xG
would suggest real guilt edge chances. It's just putting a number or a rating on something that used to be based only on gut feeling.
I struggle to reconcile how something subjective can be turned into hard data. A gilt-edged chance might have had a spinning ball that the observer couldn’t see, a bobble not taken into account or a flash of light bouncing from a watch in the crowd into the striker’s eyes. It simply might have been a lot harder than the XG hipster in the crowd could fathom.
I don't really know what point you're making? The alternative to xG is just people saying "They may have had more shots than us, but we created better chances." xG puts a number on it.
Trying to imagine your conversation in the pub after games?
"5 yards out, open goal and he misses - dreadful striker, we need to get rid of him."
"Yeah, but the sun was reflecting off someone's watch in the crowd and it got in his eyes, so y'know, can't really blame him."
I struggle to reconcile how something subjective can be turned into hard data. A gilt-edged chance might have had a spinning ball that the observer couldn’t see, a bobble not taken into account or a flash of light bouncing from a watch in the crowd into the striker’s eyes. It simply might have been a lot harder than the XG hipster in the crowd could fathom.
I don't really know what point you're making? The alternative to xG is just people saying "They may have had more shots than us, but we created better chances." xG puts a number on it.
Trying to imagine your conversation in the pub after games?
"5 yards out, open goal and he misses - dreadful striker, we need to get rid of him."
"Yeah, but the sun was reflecting off someone's watch in the crowd and it got in his eyes, so y'know, can't really blame him."
??!??!?!?!
You took the light example and made it literal, which is disingenuous at best, rather than an exaggeration intended to highlight that, defining the quality of a chance from the sidelines is far from a perfect way to come up with an indisputable mathematical system.
Is that the Brentford owner who sacked his manager when they were near the top of the division only to then see them flop?
Reading did well and then did badly. XG doesn't prove or disprove that, it was already a proven fact.
Stats are fine but they just stats, it's the almost religious faith placed in them by somevthat is illogical.
Not everything, or even most things, in life fit neatly in statistical or other boxes.
People seek order in a chaotic world hence religion, conspiracy theories and over reliance on statistical models.
Football, far more than other sports, is random. That is its beauty.
With techniques that have emerged in the last 5-10 years most things now can be modelled using a machine learning algorithm.
Really,?
Read the thread on here on mental health and tell me a machine can model it.
Love,hate,war, all algorithms or is it just that people want certainty in an uncertain world?
By the way Coates, like Bloom, is a bookie, a professional that existed, and often thrived, long before Xg came along. My uncle taught me how to calculate the 100 as a kid.
Seems like some of us are interested in this sort of analysis, so will continue to track/discuss it going forward.
Looks like we were deserving winners on Saturday (which tallies up with the view of those there) with a 1.5-2.2 win. As mentioned above, a penalty goes in for around 0.75.
Whilst we are no doubt punching above our weight somewhat, the "fair table" still sees us in 7th. If we emerge relatively in touch after the Leeds, Fulham, Swansea etc... run then it may be time to move the goalposts.
Of the more eyecatching performances of the weekend.... Brentford were full value for their defeat of Derby (2.6-0.2), and Leeds were spectacularly unfortunate to lose to Swansea, with the data giving them the match 2.7-0.7.
Interestingly, top of the table Swansea are the side most outperforming their data. They should be down in 18th with a goal difference of -2.3, yet sit clear at the top of the table on 16 points and a goal difference of +8...... 3 goals of that 10.3 goal discrepancy come from the Leeds game at the weekend, but even that aside it seems they are potentially vulnerable
I struggle to reconcile how something subjective can be turned into hard data. A gilt-edged chance might have had a spinning ball that the observer couldn’t see, a bobble not taken into account or a flash of light bouncing from a watch in the crowd into the striker’s eyes. It simply might have been a lot harder than the XG hipster in the crowd could fathom.
I don't really know what point you're making? The alternative to xG is just people saying "They may have had more shots than us, but we created better chances." xG puts a number on it.
Trying to imagine your conversation in the pub after games?
"5 yards out, open goal and he misses - dreadful striker, we need to get rid of him."
"Yeah, but the sun was reflecting off someone's watch in the crowd and it got in his eyes, so y'know, can't really blame him."
??!??!?!?!
You took the light example and made it literal, which is disingenuous at best, rather than an exaggeration intended to highlight that, defining the quality of a chance from the sidelines is far from a perfect way to come up with an indisputable mathematical system.
The whole point of xG is that it's balanced out over a huge amount of data and takes account of these one-off discrepancies - there's no such thing as a chance that will definitely result in a goal, or one that definitely won't.
What's interesting is how Swansea is far over-performing and xG shows they may be sub-average for the division, even after their defeat of Leeds. Also, our xG was 10th a week ago and is 7th now. Stoke probably should not panic yet. Table through yesterday...
PL. Most under-performing? United. They might be better than they look. Happy to say that Palace are the most over-performing. Notice also how xG is not as far off from results in the PL as The Championship. That is probably because the best players perform closer to what xG says they should and the variable nature of quality in the lower leagues means more wiggle room to over and under-perform.
Seems like some of us are interested in this sort of analysis, so will continue to track/discuss it going forward.
Looks like we were deserving winners on Saturday (which tallies up with the view of those there) with a 1.5-2.2 win. As mentioned above, a penalty goes in for around 0.75.
Whilst we are no doubt punching above our weight somewhat, the "fair table" still sees us in 7th. If we emerge relatively in touch after the Leeds, Fulham, Swansea etc... run then it may be time to move the goalposts.
Of the more eyecatching performances of the weekend.... Brentford were full value for their defeat of Derby (2.6-0.2), and Leeds were spectacularly unfortunate to lose to Swansea, with the data giving them the match 2.7-0.7.
Interestingly, top of the table Swansea are the side most outperforming their data. They should be down in 18th with a goal difference of -2.3, yet sit clear at the top of the table on 16 points and a goal difference of +8...... 3 goals of that 10.3 goal discrepancy come from the Leeds game at the weekend, but even that aside it seems they are potentially vulnerable
All sounds good but as you have said it is totally meaningless as in real life Swansea are top, whereas according to you & your stats they should be 18th. Brian Clough would be having a field day with this & to paraphrase him......football is played on grass & not on a screen or bits of paper.
The stats aren’t telling anyone where any team ‘should’ be but when it’s presented in a league table format I can see why it’s easy to believe that.
In my opinion, making a league table out of xG stats is a bit of a stretch. Ben Mayhew says himself that any xG results where the teams are within a third of a goal of each other called a draw.
Why a third? I could create a league table from shots on target and any team who are within two SoT of their opponents are awarded a draw.
I think it’s just a way to present xG data in a way that’s accessible for your average follower of football.
I much prefer ratio rankings myself.
For example if a game between Charlton and Cardiff were to end 2.5-1.5 according to xG, Charlton are given a percentage of 62.5% (2.5 divided by 4) and Cardiff given a percentage of 37.5% (1.5 divided by 4).
This way, everyone is ranked accordingly without trying to predict what results ‘should’ have been.
xG is not an opinion and not a "should." It is simply the factual ratio of all shot data showing what percentage of shots from that location go in, or do not. And then whether a player (or group of players) is over or under-performing the average of... everyone else.
If 6 times in 10 a shot from X location goes in and you are averaging 9 of 10, odds are that over time you will regress to the mean and vice-versa.
xG is probably being used by every club in Europe. It is OPTA's single most quoted stat and they are making a lot of money. For a reason. I saw an article showing that net transfer spends are dropping in the Europe over the last two years because more and more clubs are evaluating attacking players based on xG and not their raw scoring stats. Thus, instead of over-paying for goal scorers who simply out-performed their xG by chance, they are offering more reasonable sums based on what the expectation is, over the long-term. This all makes total sense to me.
Simple fact is that clubs don't waste money on analysis that fails over the long-term. Clubs, especially those outside England, are hiring more new data analysts than ones in the field. xG is here to stay because it explains what is happening in a way that pure opinion and eyeballs do not. It is not the be-all, end-all, and to try to manage a team by it would be ridiculous, but it is an effective new tool to get an edge. Or at least a new way to look in the rear-view mirror.
Yeah Experimental361 is the best free site out there I think.
You can pay thousands for some crazily accurate data, but as there are so many people looking at this sort of thing now I don't think it gives you that much of an edge punting anymore.
Just interested to see if it backs up our position in the table as we go through the season!
Comments
Reading 1.46 - 1.44 Charlton
a shot from Aguero is more likely to lead to a goal than the exact same shot from a lumbering centre back.
Does a 60th minute penalty when three up ha e a higher xG than an 89th minute one when 0-0?
However, as the video says, XG takes no account who is actually taking the shot (or header)......so a bit pointless really. You could have Kane playing up front or Simon Church and the XG would show the same thing.
I'll leave you statto's & saddo's to it.....
It's just another way to compare clubs that can indicate things such as which teams regularly create good chances or whether a team on a good run might not keep it up.
No one is saying stats are everything or that XG doesn't have any flaws.
Surely it it does matter who is having the shot.
I would expect a certain player to score in certain situations and not expect another to score in exactly the same situation.
Funny how on the Nathan Jones thread there is a league table that compares the seasons results with the results from the expected goals and only 3 or 4 teams have had the same results after 5 games.
Suggests the stat is not that accurate to me.
As I said, they're recent techniques but the beauty of the tools is that they can be applied to nearly every question that needs to be answered.
Looks like we were deserving winners on Saturday (which tallies up with the view of those there) with a 1.5-2.2 win. As mentioned above, a penalty goes in for around 0.75.
Whilst we are no doubt punching above our weight somewhat, the "fair table" still sees us in 7th. If we emerge relatively in touch after the Leeds, Fulham, Swansea etc... run then it may be time to move the goalposts.
Of the more eyecatching performances of the weekend.... Brentford were full value for their defeat of Derby (2.6-0.2), and Leeds were spectacularly unfortunate to lose to Swansea, with the data giving them the match 2.7-0.7.
Interestingly, top of the table Swansea are the side most outperforming their data. They should be down in 18th with a goal difference of -2.3, yet sit clear at the top of the table on 16 points and a goal difference of +8...... 3 goals of that 10.3 goal discrepancy come from the Leeds game at the weekend, but even that aside it seems they are potentially vulnerable
In my opinion, making a league table out of xG stats is a bit of a stretch. Ben Mayhew says himself that any xG results where the teams are within a third of a goal of each other called a draw.
Why a third? I could create a league table from shots on target and any team who are within two SoT of their opponents are awarded a draw.
I think it’s just a way to present xG data in a way that’s accessible for your average follower of football.
I much prefer ratio rankings myself.
For example if a game between Charlton and Cardiff were to end 2.5-1.5 according to xG, Charlton are given a percentage of 62.5% (2.5 divided by 4) and Cardiff given a percentage of 37.5% (1.5 divided by 4).
This way, everyone is ranked accordingly without trying to predict what results ‘should’ have been.
You can pay thousands for some crazily accurate data, but as there are so many people looking at this sort of thing now I don't think it gives you that much of an edge punting anymore.
Just interested to see if it backs up our position in the table as we go through the season!