League Two - xG Data

Some interesting data coming out of the ever-intriguing Experimental 3-6-1 site. First up, the League Two table based on xG (or expected goals).

In a nutshell, it estimates that we should have been expected to score a further 10 goals, carrying us up 11 places from 16th to 5th, above both Leyton Orient (currently top) and Northampton (currently 3rd).

[img]https://experimental361.files.wordpress.com/2023/01/2023-01-02-103.png[/img]

Some understandably question how meaningful xG actually is, but taken at face value it really does underline how season-defining our ability to recruit a top striker or two in the window could be.

Perhaps even more surprising is the xG progression of the Salford game, which suggests we should have run out comfortable winners.

[img]https://experimental361.files.wordpress.com/2023/01/2022-12-29-grimsby-salford.png[/img]

That’s possibly a questionable outlook, and the data is no doubt influenced by the fact that they sat back in the second half, but does perhaps suggest the overall performance was better than the result, which was defined by a handful of great finishes, individual mistakes and flukey deflections.

Do with it what you will, I just thought it was mildly interesting.

From what I've seen, all the, what statisticians call, big indicators, have us much further down the league than we should be. Those include shots on target, big chances conceded, progressions up the pitch, time spent in opposition half, passes into the final third and touches in the box. We're well in the top half of most of those I think, suggesting we've underperformed in terms of points relative to our actual performances.

As Poojah says, merely indicates that signing the right striker this window could really make or break our season. 5 points of the playoffs with a game in hand. Imagine if the table looked like that one!

Not into all this xG crap.

The only statistic that matters is how many times you put the ball into the net!

Quoted from Poojah

Some interesting data coming out of the ever-intriguing Experimental 3-6-1 site. First up, the League Two table based on xG (or expected goals).

In a nutshell, it estimates that we should have been expected to score a further 10 goals, carrying us up 11 places from 16th to 5th, above both Leyton Orient (currently top) and Northampton (currently 3rd).

[img]https://experimental361.files.wordpress.com/2023/01/2023-01-02-103.png[/img]

Some understandably question how meaningful xG actually is, but taken at face value it really does underline how season-defining our ability to recruit a top striker or two in the window could be.

Perhaps even more surprising is the xG progression of the Salford game, which suggests we should have run out comfortable winners.

[img]https://experimental361.files.wordpress.com/2023/01/2022-12-29-grimsby-salford.png[/img]

That’s possibly a questionable outlook, and the data is no doubt influenced by the fact that they sat back in the second half, but does perhaps suggest the overall performance was better than the result, which was defined by a handful of great finishes, individual mistakes and flukey deflections.

Do with it what you will, I just thought it was mildly interesting.

Reflects the two-goal salvo, a schoolboy error followed by a freak goal, that sank us when we were on top after the equaliser.

Could someone please explain to a mere simpleton what xG is and how they work it out?
Looks like something that a statistition with no knowledge of football has come up with.

Quoted from mimma

Could someone please explain to a mere simpleton what xG is and how they work it out?
Looks like something that a statistition with no knowledge of football has come up with.

The expected number of goals based off quality of the shots you take, based on factors like where the shot is taken etc. a penalty or shot from 6 yards out has a far higher xG than a popshot from 30 yards.

Quoted from mimma

Could someone please explain to a mere simpleton what xG is and how they work it out?
Looks like something that a statistition with no knowledge of football has come up with.

Some nerds basically determine the statistical likelihood of scoring a goal from each attempt, based on things like distance from goal, angle, body part used and some subjective shite like if it's a 'big chance' and the 'pattern of play', and then give it a score from 0 (no chance of ever scoring a goal) to 1 (100% chance of scoring a goal). Basically if a chance has an xG of 0.79 (such as a penalty), then it should result in a goal 79 times out of 100 (statistically 79% of penalties are converted into goals). Someone hitting a backheeled shot from 30 yards at an acute angle would probably be given an xG on the effort of like 0.02. I don't massively rate it as a method for analysing the performance of a team within a match or over a season - I think really all it does is tell you is if you've got some good or poor finishers in your team (and it seems our finishing this season has been poor based on xG).

If there had been xG data about 40 years ago then it would have been saying I would have got laid every night, but as always in the real world it wasn’t quite like that.

Quoted from Les Brechin

Not into all this xG crap.

The only statistic that matters is how many times you put the ball into the net!

In terms of points, of course, but there’s a reason club’s employ data analysts; it helps you understand the where and the why behind the bottom-line results you’re getting.

For me, this helps reaffirm what we already know - that we’re light in quality when it comes to what happens in the box. If we can solve that in the window, I honestly believe we have an exciting rest of the season ahead of us.

So, us and Stockport are the biggest movers upwards, may be worth a punt for us both to reach the play-offs (if we sign that striker).

I recall looking at something similar for our 2016 promotion season - we were way out at the top.

There are many ways to win a football match but all the statistics,  all the data mean diddly squat.

A long ball team can have a winning formula and an ultra passing team can have a winning formula, which is one of football's many joys.

The only thing that matters is how many goals you score, and any near misses, goals you should have scored or chances missed are irrelevant.

Those dismissing data are missing the point. It helps enormously if you analyse it correctly. Just ask Brentford.

It's the same principle as highlighted in the film moneyball.

All this data has done, as Poojah says, is confirm what we all know and PH knows. We don't score enough when on top in games and we need players who statistically score more of their chances.

In terms of performances this season, I think we have been generally good and I haven't seen many teams who are demonstrably better than us. They just score more of the chances that come their way.

Fix that conundrum and we will rise up the table.

Quoted from ginnywings

Those dismissing data are missing the point. It helps enormously if you analyse it correctly. Just ask Brentford.

It's the same principle as highlighted in the film moneyball.

All this data has done, as Poojah says, is confirm what we all know and PH knows. We don't score enough when on top in games and we need players who statistically score more of their chances.

In terms of performances this season, I think we have been generally good and I haven't seen many teams who are demonstrably better than us. They just score more of the chances that come their way.

Fix that conundrum and we will rise up the table.

As you say you don't need data analysis to tell you that?!

Funny thing is adding a player who "scores more goals" upsets the balance in other areas making the previous data meaningless.

Quoted from ginnywings

Those dismissing data are missing the point. It helps enormously if you analyse it correctly. Just ask Brentford.

It's the same principle as highlighted in the film moneyball.

All this data has done, as Poojah says, is confirm what we all know and PH knows. We don't score enough when on top in games and we need players who statistically score more of their chances.

In terms of performances this season, I think we have been generally good and I haven't seen many teams who are demonstrably better than us. They just score more of the chances that come their way.

Fix that conundrum and we will rise up the table.

We do seem to give away more daft goals than we score. Not many come to mind where the opposition has gifted us a goal, like Salford's second, and I can recall several quality goals scored from outside the box that will have an xG of about 0.05. These things tend to balance themselves out over the course of a season though.

I do find stuff like this interesting, but it baffles me sometimes: how was Matt Smith's total xg against us 0.6 when one of his goals was a header from about 5 yards out? (surely that would be at least that much on its own?)

Quoted from Abdul19

I do find stuff like this interesting, but it baffles me sometimes: how was Matt Smith's total xg against us 0.6 when one of his goals was a header from about 5 yards out? (surely that would be at least that much on its own?)

Good point, but the other two will be worth very little between them, one was a scuff into the ground and it's hit his standing leg, the other is one you'd actually expect Max to save, bet they're barely worth 0.1 between them.

Quoted from Mariner_09

Good point, but the other two will be worth very little between them, one was a scuff into the ground and it's hit his standing leg, the other is one you'd actually expect Max to save, bet they're barely worth 0.1 between them.

I can't remember which team it was but was watching a game on MOTD and one team scored 4 but their xG was something like 1.75?

Quoted from Chrisblor

Some nerds basically determine the statistical likelihood of scoring a goal from each attempt, based on things like distance from goal, angle, body part used and some subjective shite like if it's a 'big chance' and the 'pattern of play', and then give it a score from 0 (no chance of ever scoring a goal) to 1 (100% chance of scoring a goal). Basically if a chance has an xG of 0.79 (such as a penalty), then it should result in a goal 79 times out of 100 (statistically 79% of penalties are converted into goals). Someone hitting a backheeled shot from 30 yards at an acute angle would probably be given an xG on the effort of like 0.02. I don't massively rate it as a method for analysing the performance of a team within a match or over a season - I think really all it does is tell you is if you've got some good or poor finishers in your team (and it seems our finishing this season has been poor based on xG).

If you're not sure what xG is and you're only going to take notice of one post in this thread then this is the one.

Quoted from Les Brechin

Not into all this xG crap.

The only statistic that matters is how many times you put the ball into the net!

Do you write your team name on the back of a cigarette packet too?  - I'm only joking btw

Quoted from lew chaterleys lover

There are many ways to win a football match but all the statistics,  all the data mean diddly squat.

A long ball team can have a winning formula and an ultra passing team can have a winning formula, which is one of football's many joys.

The only thing that matters is how many goals you score, and any near misses, goals you should have scored or chances missed are irrelevant.

Who was the first chap to really analyse football games?  Can't remember off the top of my head but anyway he discovered that the vast majority of goals were scored in 3 passes or less from his data, and everyone immediately decided that this meant long ball was the best way to score goals without thinking about the fact a high press to turnover ball near the opposition box would also result in a '3 passes or less' goal etc.   The data is only as good as the person reading it really

Quoted from OddShapedBalls

Do you write your team name on the back of a cigarette packet too?  - I'm only joking btw

[youtube]6PbP_pIgR4s[/youtube]

Quoted from OddShapedBalls

Who was the first chap to really analyse football games?  Can't remember off the top of my head but anyway he discovered that the vast majority of goals were scored in 3 passes or less from his data, and everyone immediately decided that this meant long ball was the best way to score goals without thinking about the fact a high press to turnover ball near the opposition box would also result in a '3 passes or less' goal etc.   The data is only as good as the person reading it really

John Beck?

Quoted from OddShapedBalls

Who was the first chap to really analyse football games?  Can't remember off the top of my head but anyway he discovered that the vast majority of goals were scored in 3 passes or less from his data, and everyone immediately decided that this meant long ball was the best way to score goals without thinking about the fact a high press to turnover ball near the opposition box would also result in a '3 passes or less' goal etc.   The data is only as good as the person reading it really

Like him or loathe him I believe the late Don Revie was one of the first to keep an analysis of players. Whether he included data on the actual games or not I don't know.

Quoted from OddShapedBalls

Who was the first chap to really analyse football games?  Can't remember off the top of my head but anyway he discovered that the vast majority of goals were scored in 3 passes or less from his data, and everyone immediately decided that this meant long ball was the best way to score goals without thinking about the fact a high press to turnover ball near the opposition box would also result in a '3 passes or less' goal etc.   The data is only as good as the person reading it really

Howard Wilkinson maybe?

You can often reach the wrong conclusions when reading data, and what might be a solution for one team, would not work for another with the same data.

My general issue with xG is two fold:

a) I feel it's very on-trend, and far too many kids are crying over it as if it's more worthy than the goals column

b) who questions the validity of how it's calculated? For example, looking at the data provided during that Salford game, it looks like Khan's chance (which he scored) was about 0.4, so 40% of the time from that position, in those exact circumstances, he'll score, 60% of the time he'd be expected to miss.

But then you look at Salford's fourth. It jumps by about 0.3, yet the bloke heads it in, relatively unchallenged, three yards out. Is the data suggesting he'd only likely convert that chance 30% of the time? Because most footballers will bury a header from that range. I'd personally put it at 0.8 or 0.9.

So who is making these decisions? Obviously someone is watching; something (programmatic or robotic) is measuring. I'd question the quality of that programming.

We might still be forming opinions on an unreliable methodology. People have bought into it, almost without question or scrutiny.

It's trying to make something subjective, objective. It's not as black and white as 'did the ball cross the line' or 'was he offside'? Attaching value to a unique set of circumstances is incredibly difficult to put a number to, so there's masses of mistrust from me, and mistrust towards people who put too much faith/value in it (for now).

Maybe I need educating more, but I think it's healthy to question the validity of most things.

All teams do data analysis now, suggesting there is little to be gained by doing so, because everyone else is.

But if you don't do it, then you would be at a disadvantage compared to others, so it's become a big part of the game.

I think that is why some fans are baffled by team selections and the use of substitutes during games. The management have training data and up to the minute match data that we are not privy to.

For instance, injuries are more likely to occur when a players recovery rate has dropped below a certain parameter and can be the reason he's replaced, even though he appears to be having a good game.

We often hear fans say how someone is blowing out of his ar$e and they couldn't understand why he was kept on, but the data is probably telling a different story.

Quoted from OddShapedBalls

Who was the first chap to really analyse football games?  Can't remember off the top of my head but anyway he discovered that the vast majority of goals were scored in 3 passes or less from his data, and everyone immediately decided that this meant long ball was the best way to score goals without thinking about the fact a high press to turnover ball near the opposition box would also result in a '3 passes or less' goal etc.   The data is only as good as the person reading it really

Charles Hughes

Quoted from Les Brechin

Not into all this xG crap.

The only statistic that matters is how many times you put the ball into the net!

And how many the ball ends up In yours !

Data can be selective and used to support any point of view needed.

  I saw a stat this week that Leeds had faced 55 good chances against them this season.  The most in the PL. The number includes goals against and chances that should have conceded a goal.

Despite this Leeds are not in the bottom 3 and will probably stay up.  

Quoted from RichMariner

My general issue with xG is two fold:

a) I feel it's very on-trend, and far too many kids are crying over it as if it's more worthy than the goals column

b) who questions the validity of how it's calculated? For example, looking at the data provided during that Salford game, it looks like Khan's chance (which he scored) was about 0.4, so 40% of the time from that position, in those exact circumstances, he'll score, 60% of the time he'd be expected to miss.

But then you look at Salford's fourth. It jumps by about 0.3, yet the bloke heads it in, relatively unchallenged, three yards out. Is the data suggesting he'd only likely convert that chance 30% of the time? Because most footballers will bury a header from that range. I'd personally put it at 0.8 or 0.9.

So who is making these decisions? Obviously someone is watching; something (programmatic or robotic) is measuring. I'd question the quality of that programming.

We might still be forming opinions on an unreliable methodology. People have bought into it, almost without question or scrutiny.

It's trying to make something subjective, objective. It's not as black and white as 'did the ball cross the line' or 'was he offside'? Attaching value to a unique set of circumstances is incredibly difficult to put a number to, so there's masses of mistrust from me, and mistrust towards people who put too much faith/value in it (for now).

Maybe I need educating more, but I think it's healthy to question the validity of most things.

It's very complex, Smiths header for Salford may be lower because you would also expect a defender or the keeper to collect the ball.

Playing devils advocate here as I've seen people saying our xG highlights our need for a striker to convert them. What if our xG is actually coming from midfielders and defenders at set pieces?

Quoted from LN8Mariner

Charles Hughes

Nope, Wing Commander Reep:

https://en.wikipedia.org/wiki/Charles_Reep

As for xG, it's so arbitrary in many regards, that I don't think it really provides much useful information, however much the sports scientists like it. If I understand it correctly, it doesn't take into consideration anything to do with the actual abilities of the players, just an analysis of factors which somehow decide how 'good' a chance is. However, you can't ignore the individual aspect because, fundamentally, some players are good finishers and others are not. A chance from 1 yard might be expected to be converted 9 times out of 10 but for some players, you're talking perhaps 3 or 4 of those being scored!

I also wouldn't necessarily agree that all we're lacking is a goalscorer to turn our 'xG' into real goals. We do manage some pretty good approach play on occasion but a lot of the reason for a lack of goals is as much inaccurate crossing as poor finishing. If the final cross is just behind the player or just a little too high, it's not a great chance whatever the angle or distance from goal. No doubt a better striker would still manage to convert more of these difficult chances, but I don't think anybody can seriously say that we've played well enough to be in the play-off spots for most of the season.

Quoted from Maringer

Nope, Wing Commander Reep:

https://en.wikipedia.org/wiki/Charles_Reep

As for xG, it's so arbitrary in many regards, that I don't think it really provides much useful information, however much the sports scientists like it. If I understand it correctly, it doesn't take into consideration anything to do with the actual abilities of the players, just an analysis of factors which somehow decide how 'good' a chance is. However, you can't ignore the individual aspect because, fundamentally, some players are good finishers and others are not. A chance from 1 yard might be expected to be converted 9 times out of 10 but for some players, you're talking perhaps 3 or 4 of those being scored!

I also wouldn't necessarily agree that all we're lacking is a goalscorer to turn our 'xG' into real goals. We do manage some pretty good approach play on occasion but a lot of the reason for a lack of goals is as much inaccurate crossing as poor finishing. If the final cross is just behind the player or just a little too high, it's not a great chance whatever the angle or distance from goal. No doubt a better striker would still manage to convert more of these difficult chances, but I don't think anybody can seriously say that we've played well enough to be in the play-off spots for most of the season.

That's the chap!  I remember him doing an interview years ago where he said he preferred passing, attacking football but his name had become synonymous with 'long ball'.

What I would like to see from the League is a stat pack like this one that the Union has, very interesting reading to kill a few minutes:  https://www.unitedrugby.com/statistics

My other team being the Scarlets it's something I like to look at occasionally but again shows how stats alone don't tell the full story.  For example we have the 4th best attack out of 16 teams and the 7th best defence according to the measures used....but are 15th out of 16 teams in the league with only 2 wins all season.  

Quoted from jamesgtfc

It's very complex, Smiths header for Salford may be lower because you would also expect a defender or the keeper to collect the ball.

Playing devils advocate here as I've seen people saying our xG highlights our need for a striker to convert them. What if our xG is actually coming from midfielders and defenders at set pieces?

I didn't think it has anything to do with a defender would normally get it, it's to do with once the 'shot' has been struck, how likely that is to result in a goal.

I don't think the statistics are as good at this level as in the premier league, there's not as much money for the people to do the analysis and the quality of player is different.

It's also complicated, because it only says that we should have scored more goals than we have, which suggests having a fox in the box type would bring us more goals. But it's not so simple because if we had say an Amond that scored these chances, we'd be losing the build up part that Taylor contributes and likely have fewer chances.

Has anyone told Johnny Mc that 79% of penalties end in goals??
Charles Hughes was the guy that influenced Graham Taylor when he took Watford from division four to one. Not pretty but very effective

I'm all for statistics and analysis but this Xg thingy is one statistic too far.
Put simply, it's a load of sh1te.

Quoted from pen penfras

I didn't think it has anything to do with a defender would normally get it, it's to do with once the 'shot' has been struck, how likely that is to result in a goal.

I don't think the statistics are as good at this level as in the premier league, there's not as much money for the people to do the analysis and the quality of player is different.

It's also complicated, because it only says that we should have scored more goals than we have, which suggests having a fox in the box type would bring us more goals. But it's not so simple because if we had say an Amond that scored these chances, we'd be losing the build up part that Taylor contributes and likely have fewer chances.

I made the point in your last paragraph earlier - it would be OK if you can add a striker to the existing team to covert the chances but you are not allowed 12 players! The whole dynamic changes making the previous set of statistics worthless.

Hurst strikes me as the sort of fella who would be level headed about the use of data and take the more pertinent aspects of it and take this fantasy xg with a pinch of salt.

Having said that I suppose statistics play a part in who we let go in terms of how much running they do, say somebody like Maguire Drew. You could make an argument against that saying the only stats that really matter is how many chances and goals he makes and scores.

It's all too subjective for my liking, as highlighted by many posts here. I also say that as someone who spends a lot of time analysing and interpreting data.

For all the xG data around Town, I'm not convinced we're creating that many chances we should be burying, I think the issue is further back in the team that we're not creating enough chances to start with, be that through a creative midfielder or someone to lead the line in the first place.   Which massively brings into question further the subjective nature of what is a 'good chance'.  Far, far too many variables at play to say Shot X has this chance of being a goal and Header Y this.  I think that's the beauty of football too, it's not just played on a spreadsheet.

Data is huge in football these days.  Also pleasing to see Town are using it far more than we ever have done before.  But I don't think using metrics like xG alone is enough.

As a sidenote, data in football has largely emerged through the gambling industry.  A lot of the firms who provide that data are happy to share to a wider audience.   You have to question why they would want that to be out in the open if it's almost like a cheat code...

Ultimately the only stats that count are goals scored, goals conceded and points gained.  Whatever you can do to improve them is of course important, but sometimes you can't explain football with statistics.  

Quoted from Poojah

[img]https://experimental361.files.wordpress.com/2023/01/2022-12-29-grimsby-salford.png[/img]

These pictures didn't show up for me before. That is so flawed, it's ridiculous. The 2nd and 3rd are deemed as similar chances because they're the same distance out, but the 2nd had a player driving through into space with loads of time to pick his spot and hit it true. The 3rd had 2 defenders on the turn and fluked it in. I'd back the 2nd chance to be a goal 1/3 of the time and the 3rd I doubt 1/20.

And to think that we deserved to win that game because we missed some decent chances when the game was out of sight is stupid. If they weren't 3 goals in front, they wouldn't have sat back like that.

Quoted from pen penfras

These pictures didn't show up for me before. That is so flawed, it's ridiculous. The 2nd and 3rd are deemed as similar chances because they're the same distance out, but the 2nd had a player driving through into space with loads of time to pick his spot and hit it true. The 3rd had 2 defenders on the turn and fluked it in. I'd back the 2nd chance to be a goal 1/3 of the time and the 3rd I doubt 1/20.

And to think that we deserved to win that game because we missed some decent chances when the game was out of sight is stupid. If they weren't 3 goals in front, they wouldn't have sat back like that.

Never mind all that..answer the questions on the thread about office staff taking club to account for them leaving which you’ve avoided for over a week..you were the one who brought it up..

Quoted from marinerjase

Never mind all that..answer the questions on the thread about office staff taking club to account for them leaving which you’ve avoided for over a week..you were the one who brought it up..

Ask Jason Stockwood at the fans forum. I'm not messing up people's legal claims by posting details on a forum, but there are several former employees that are taking action against the club

Quoted from OddShapedBalls

Who was the first chap to really analyse football games?  Can't remember off the top of my head but anyway he discovered that the vast majority of goals were scored in 3 passes or less from his data, and everyone immediately decided that this meant long ball was the best way to score goals without thinking about the fact a high press to turnover ball near the opposition box would also result in a '3 passes or less' goal etc.   The data is only as good as the person reading it really

Charles Reep:
https://en.wikipedia.org/wiki/Charles_Reep