xG table

Redwurzel

Well-known member
There are no terrible teams in the Championship, Rotherham will be paying £10k to £12k a week to attract players like Peltier and pay Hall's wages. They were never out of the top 3 last season in Div1 and have a very shrewd manager.

We still should be beating them, but there is not as big a gap resources wise as they used to be. Therre will be a much bigger resources gap between us and the parachute clubs (Norwich, Watford, Burnley, WBA, Sheffield United).

Stats were 13 to 3 on goal attempts to Boro.
 

Laughing

Well-known member
It's good at gauging chance creation, and chances are generally what matter, as the better the chances, the higher the probability to score, all things (ability wise) assumed being equal, but of course this varies.

It means more over more games, but if we finish 21st and top the xG stats then it will probably be the first time there's even been such a disparity. It's not an exact science, far from it, but it's more reliable than feel or guesswork. The largest disparity over a teams position is normally around 8-10 places over the course of a season, most teams are within 3/4 places. We won't finish below mid table doing what we are, it's highly improbable over that many games. If it does end up that way, then I'll just accept it, providing what I'm seeing matches the stats (which for me it does up to now). I have pretty much agreed or predicted the xG range for most games quite accurately.

International penalties for stats are not worth a grain of salt for standard deviation, not considering all games at least, as the difference in standard and pressure is absolutely massive. Over a 46 game championship season the levels of players and keepers are a hell of a lot closer.

Yes, they have updated it (now consider about 30 metrics and dumped "big chances"), so it will certainly be more accurate (as they will have done a back analysis), but even over former years, team position and results correlated pretty well with xG, a lot better than any other metric.

I'd expect Opta Analysts have good knowledge, probably more than or similar to a weekly football fan, and not subject to the same bias or pessimism.

I wouldn't use xG as a prediction tool, it won't factor for who is playing, injuries, form or one-off events, but looking back it works well and, better than someone saying we lost, so we played crap, it's not that simple. Prediction of individual games is tough enough as it is, but things like xG and other stats become more and more accurate over time.

Similar to crosses and the like, we've had 3rd most crosses I think (missed crosses don't count as chances), and conceded the 5th least.

Save % is 3rd lowest, shots on target % second lowest, I'd expect a lot better from our players in those areas (the crucial ones).

We're not giving much away (even with the individual errors), but conceding too much, and we're not putting enough away. I suppose the "playing well" or as expected part is the bit in the middle, stopping chances and crosses and creating them, and we're doing these bits well, especially in the earlier games.
Oh dear Andy.

Firstly it is quite good at indicating how a game went, in so far as it's probably the best we have, however that isn't saying much.

Your assertion that it gets more accurate over more games, of course it does, it could be that xG drops to reflect the position of the club or the club climb the table to match their xG. Again this says nothing of any statistical relevance.

Stats for international penalties don't matter for SD- OK, whatever. Your reasoning for this statement is also nonsense. It assumes there is more pressure on in an international game than a game with 2 sides facing relegation. It's nonsense andy and has nothing but a simple conjecture from you to back the opinion up.

I have no idea whether the update will make the stat more or less accurate. Time will tell. Back-analysis can be very poor analysis, depending on how it was performed. I have no idea how it was performed, but presumably the same as for version 1 of xG which was poor. Again by opta's own admission.

I am sure Opta have a huge amount of expertise, they had the same level of expertise when they released version 1 of xG which again was poor, so not sure what you are saying here. If you are ssaying htey must be right and I am wrong, then no that may be the case, but the stats I gave which are better indicators is a fact, not an opinion.

I agree I wouldn't use xG as a predictor or even as part of other metrics, because it hasn't proved itself. It may with version 2 and I'll do the analysis at the end of the season.

I agree we are not giving too much away and we are creating chances. No issues with that assertion, but I certainly wouldn't form that opinion from stats alone, that would be ridiculous. Ultimately there is, probably, a reason why our stats look good and our position doesn't. It may be bad luck in-running, which is one of the reasons why xG tells only a fraction of the story, and not very well at that. It could also be that the xG stat isn't very good at assessing chances and attributing a value to the quality of the chance.

We have had this argument before, you can take what I say at face value, or not. I spent several hundred hours trying to use xG in a machine learning capacity. The BigML algorithm, the Azure ML platform and my own. All of them told me that xG was not very influential. The algorithms ran over 10k games over 20 odd leagues and 3 seasons and I removed it because it didn't work. I then looked at why it didn't work. That's because the formula wasn't very good.

Oh and I paid a lot of money for access to the stat from stats perform, opta's parent company.

Id it a good indicator of how a game played out? Yup it is, because it's the best we have, which again isn't saying much.
 

Laughing

Well-known member
There are no terrible teams in the Championship, Rotherham will be paying £10k to £12k a week to attract players like Peltier and pay Hall's wages. They were never out of the top 3 last season in Div1 and have a very shrewd manager.

We still should be beating them, but there is not as big a gap resources wise as they used to be. Therre will be a much bigger resources gap between us and the parachute clubs (Norwich, Watford, Burnley, WBA, Sheffield United).

Stats were 13 to 3 on goal attempts to Boro.
Rotherham were fairly terrible in that particular game though. However, was it because we stopped them playing? I don't know. They could equally be great for a run of games, I doubt it, but I don't know. It's why football performance is difficult to analyse. Team invasion sports always are.
 

Otto42

Well-known member
It's a good indicator of performance, but the measure (Key Performance Indicator) we use is points won in games. I get we can use this as evidence to say Wilder's still the man for the job and we should expect things to get better, but that's all.
 

Andy_W

Well-known member
Oh dear Andy.

Firstly it is quite good at indicating how a game went, in so far as it's probably the best we have, however that isn't saying much.

Your assertion that it gets more accurate over more games, of course it does, it could be that xG drops to reflect the position of the club or the club climb the table to match their xG. Again this says nothing of any statistical relevance.

Stats for international penalties don't matter for SD- OK, whatever. Your reasoning for this statement is also nonsense. It assumes there is more pressure on in an international game than a game with 2 sides facing relegation. It's nonsense andy and has nothing but a simple conjecture from you to back the opinion up.

I have no idea whether the update will make the stat more or less accurate. Time will tell. Back-analysis can be very poor analysis, depending on how it was performed. I have no idea how it was performed, but presumably the same as for version 1 of xG which was poor. Again by opta's own admission.

I am sure Opta have a huge amount of expertise, they had the same level of expertise when they released version 1 of xG which again was poor, so not sure what you are saying here. If you are ssaying htey must be right and I am wrong, then no that may be the case, but the stats I gave which are better indicators is a fact, not an opinion.

I agree I wouldn't use xG as a predictor or even as part of other metrics, because it hasn't proved itself. It may with version 2 and I'll do the analysis at the end of the season.

I agree we are not giving too much away and we are creating chances. No issues with that assertion, but I certainly wouldn't form that opinion from stats alone, that would be ridiculous. Ultimately there is, probably, a reason why our stats look good and our position doesn't. It may be bad luck in-running, which is one of the reasons why xG tells only a fraction of the story, and not very well at that. It could also be that the xG stat isn't very good at assessing chances and attributing a value to the quality of the chance.

We have had this argument before, you can take what I say at face value, or not. I spent several hundred hours trying to use xG in a machine learning capacity. The BigML algorithm, the Azure ML platform and my own. All of them told me that xG was not very influential. The algorithms ran over 10k games over 20 odd leagues and 3 seasons and I removed it because it didn't work. I then looked at why it didn't work. That's because the formula wasn't very good.

Oh and I paid a lot of money for access to the stat from stats perform, opta's parent company.

Id it a good indicator of how a game played out? Yup it is, because it's the best we have, which again isn't saying much.
The "oh dear" isn't necessary, your opinion is no more valid than mine. Professional clubs can see worth in it (which is largely limited to post analysis), which is good enough for me (and should be for you), and I've largely agreed with it, this year especially.

It's relevant when people are saying we're playing crap or deserve to be where we are, or think a manager change would bring a better difference between chance creation and chances conceded. It would still be the same squad etc.

The only stats that matter are the ones which make up the table, and this is fine if people say this, as long as they're saying the stats which make up the table aren't always a reflection of expected outcomes from post-analysis. I.e if you beat Swansea 2-1 it doesn't mean that result would be more replicable than say the Reading, Sheff Utd, Stoke games, if we repeated the chances (performance) over and over again. It's fine to not get what you deserve some games, and get more in others, that's football, and sometimes you will have a string which is away from the norm, just like having 11 reds in a row on roulette.

My main reasoning on penalties (or anything international-related) is for example, 90% of England's games, we're not playing teams of a comparable level (not like as close as any individual English league is), you would need to break the stats into various sections where they are comparable, i.e maybe compare knockout stages of tournaments, or even better quarter-finals and onwards. There's nothing to gain by including the Faeroe islands taking a penalty against France, and vice versa, not when trying to gauge the chance Harry Kane has of scoring against a top 10 side. Boro wouldn't be taking penalties against a guy who plays for Norton and works a the chip shop full time.

In order to analyse correctly (most accurately), you need to remove anomalies/ non comparable situations from the dataset, and when using for predictions you need to drill down even further. It's like lumping covid survival rates altogether, it's pointless and lacks detail, it needs to be looked at by age group and then level of health if you want to go one further, and want to get the most out of it.

Like I say, my guess is they back analysed it already, or beta tested it, more than any of us are capable of full time, and certainly in our spare time.

I take your comment at face value, as in I don't think you're lying, but it doesn't mean that I think your analysis is correct, and I've no reason to think you're going to be able to come up with something better than Opta. How long do you think Opta have spent on it, and how much money? xG isn't influential on future outcomes, it's not designed to be, it's a post analysis tool, and that's it.

Paying for stats doesn't mean they're being used correctly, just like me paying for a financial times subscription does not make me Warren Buffet, Charlie Munger, Nick Leeson or Jordan Belfort. Just like someone watching a youtube video on my industry doesn't make them an expert, despite them thinking they are.

Like I say, it's the best we have, which is a good indicator when used correctly, and far better than opinions, especially those influenced by peoples feelings, especially on subjects they care about and may have a bias or pessimism about. Some people just can't look passed the results.
 

BoroMart

Well-known member
It's a good indicator of performance, but the measure (Key Performance Indicator) we use is points won in games. I get we can use this as evidence to say Wilder's still the man for the job and we should expect things to get better, but that's all.
nah mate, that's the critical success factor
 

Andy_W

Well-known member
Honestly can’t stand that stat! Does my head in. Lost all respect for Chris Wilder when he brought it up in a presser.. in defence of us losing games.
Why don't you like it, don't you believe chances lead to goals? If not, why?

Is the aim of the game not to create more and better chances than the opposition, and hope your players are good enough to score/ not conceded them? What else can we do, not create, and give more chances away, how would that work out? Better or worse?

It's not an excuse, it's a partial explanation of a stat that few seem to understand, I'm happy that he's been quite open about things, where he can be.

Some other explanations are:
We're missing chances we should be scoring. If we aren't good enough to score them, then we have no right to complain we're not scoring, we're certainly creating enough, and putting in enough largely good-quality crosses.
The vast majority of errors, where we have missed chances, have been from players we had last year.
We're making silly individual errors, if we aren't good enough to cut these out, then don't complain we're conceding.
The vast majority of defensive errors, which have cost goals, have been from players we had last year.
We had some shocking decisions go against us in the first few games (which cost a fair few goals)
We've not really had any shockers go for us.
We played some pretty good sides in the first few games
We've been worse without Lenihan and Steffen, as far as "playing well", which I would categorise as creating more than the opposition, especially against good sides.
We bought two strikers who seemingly aren't as good as what we had last year (in the opinion of the manager), and one of the players getting a game ahead of them was a midfielder, or they're not up to speed yet.
We sold our best midfielder, and the only one likely to improve, and another £5m wasted on another who we gave away for nothing. We did this before we had replacements, which was lunacy.
None of the permanent signings we have spent money on were first choice, or are guaranteed starters.
It's a game of fine margins and nobody has battered us yet. Yes we've been battered first half in some games, but reacted well, which has been a surprise.
Three probable starting players have missed a lot of the games, I know this can happen, but it has happened, lets not pretend it hasn't.
High turnover of players, the main striker has only been here a few games
We were meant to get in two main strikers
We get dominated in the middle, more than anywhere else on the pitch, even though we're playing three in there. This pretty much eliminates all options of playing less than three players in there.

I don't think any of those other explanations can largely be blamed at the manager.

What I suppose the manager can maybe do better at:
We seem to lack concentration/ awareness, especially at the back, I'm not sure if there's any specific training we can do for this
Some defenders and Roberts were launching it too much (and don't have long ball ability), but they've got better at this (keeping it on the deck), they still need to work on it mind.
Improving confidence, although the players need to dig themselves out of this hole I think.
Kept McGree for the middle, when it became apparent that the free midfield signings were not going to cut it, although saying that the midfield have done ok in possession, and only been under the cosh when we've given it away too easily. Plus, McGree seemed a better finisher than Watmore, and that's saying something.
McNair is not performing like last year (or any other year), despite playing in a position which should suit him better, maybe last year he looked better as we didn't attack down the left, and he had Tav on his side.
Dijksteele has looked the best going forward, made the least mistakes at the back, and is also the quickest at the back, he has to play.
Dijksteele wasn't on the pitch when we shipped 3 in against QPR, but was when we scored two and looked far better, second half.
Dijksteele wasn't on the pitch when we shipped 3 in against Cardiff, but was when we scored two and looked far better, second half.
I would say we could do with another formation option, but keeping the same formation has got us back into games when we were behind, so not sure we need this yet.
 

newyddion

Well-known member
To be fair with the reality of the situation, with it being a striker and the keeper so far off his line, it would have been a higher probability I expect.

There's a reason not many goals are scored from that far out, as people don't attempt them, as from that position they may be able to work a better chance and as keepers are usually in a good position.

I had little against him going for that mind, it was probably on target and a little more on it and it would have been a goal. It was risking 1 point to get three, assuming we wouldn't have expected to concede a chance in the last minute or so, never mind a fluke goal. He could have squared it or tried to hold onto it, but we would likely have lost the ball anyway.

My larger concern was the first goal we conceded, why Paddy didn't slide for the ball against Pedro, why Mowatt slid in stupidly when Sarr had the ball on his left foot and got turned inside out, and he could have largely blocked a left-foot shot on his feet anyway. Why Mowatt defended the wrong target initially (the lesser threat), why Mowatt was on the line he was when he could see Clarke playing Sarr on-side. How Clarke missed a free header from 6 yards out. All of those are basic individual errors, from players we should expect more from.
xGf - expected gaffs

top of the league in that particular stat also!
 

newyddion

Well-known member
Why don't you like it, don't you believe chances lead to goals? If not, why?

Is the aim of the game not to create more and better chances than the opposition, and hope your players are good enough to score/ not conceded them? What else can we do, not create, and give more chances away, how would that work out? Better or worse?
It’s a meaningless garbage stat, too many variables to use as a genuine tool. Ok to give a post match summary it sounds cool for podcasts and the like but not a genuine indicator for quality.
 

Nano

Well-known member
It's an indicator but it isn't reliable. You can have loads of low xG chances like we have where you're have to be very lucky to score or a single penalty and we'll be winning on xG despite the penalty being scored.

The biggest failing is it only gives a score for actual shots but if you accept that a skit from the penalty spot via a cross results in a goal 1/10 occasions, xG of 0.1 then you have to accept that a cross only results in a shot 1/5 occasions so an xG of 0.1 is really 0.02 from the cross so we're already over performing our xG just by having the shots (because we pass instead of cross) but another team has fewer successful crosses doesn't mean more goals should be scored.

It's not meaningless but it's over valued at the moment. It's an indicator, but an answer
 

Gunslinger

Active member
It's an indicator of how many goals you haven't scored when you maybe should have

Which we already know because we can check the goals for column

Honestly what is the point of this statistic, worse than how much possesion we have had, or throw ins or corners, the reality is the only thing that matters is points
and we are severly lacking in them

We are better than we are because a bit of addign up mixed with a bit of guess work in a subjective mathematical model says so
 

Laughing

Well-known member
The "oh dear" isn't necessary, your opinion is no more valid than mine. Professional clubs can see worth in it (which is largely limited to post analysis), which is good enough for me (and should be for you), and I've largely agreed with it, this year especially.

It's relevant when people are saying we're playing crap or deserve to be where we are, or think a manager change would bring a better difference between chance creation and chances conceded. It would still be the same squad etc.

The only stats that matter are the ones which make up the table, and this is fine if people say this, as long as they're saying the stats which make up the table aren't always a reflection of expected outcomes from post-analysis. I.e if you beat Swansea 2-1 it doesn't mean that result would be more replicable than say the Reading, Sheff Utd, Stoke games, if we repeated the chances (performance) over and over again. It's fine to not get what you deserve some games, and get more in others, that's football, and sometimes you will have a string which is away from the norm, just like having 11 reds in a row on roulette.

My main reasoning on penalties (or anything international-related) is for example, 90% of England's games, we're not playing teams of a comparable level (not like as close as any individual English league is), you would need to break the stats into various sections where they are comparable, i.e maybe compare knockout stages of tournaments, or even better quarter-finals and onwards. There's nothing to gain by including the Faeroe islands taking a penalty against France, and vice versa, not when trying to gauge the chance Harry Kane has of scoring against a top 10 side. Boro wouldn't be taking penalties against a guy who plays for Norton and works a the chip shop full time.

In order to analyse correctly (most accurately), you need to remove anomalies/ non comparable situations from the dataset, and when using for predictions you need to drill down even further. It's like lumping covid survival rates altogether, it's pointless and lacks detail, it needs to be looked at by age group and then level of health if you want to go one further, and want to get the most out of it.

Like I say, my guess is they back analysed it already, or beta tested it, more than any of us are capable of full time, and certainly in our spare time.

I take your comment at face value, as in I don't think you're lying, but it doesn't mean that I think your analysis is correct, and I've no reason to think you're going to be able to come up with something better than Opta. How long do you think Opta have spent on it, and how much money? xG isn't influential on future outcomes, it's not designed to be, it's a post analysis tool, and that's it.

Paying for stats doesn't mean they're being used correctly, just like me paying for a financial times subscription does not make me Warren Buffet, Charlie Munger, Nick Leeson or Jordan Belfort. Just like someone watching a youtube video on my industry doesn't make them an expert, despite them thinking they are.

Like I say, it's the best we have, which is a good indicator when used correctly, and far better than opinions, especially those influenced by peoples feelings, especially on subjects they care about and may have a bias or pessimism about. Some people just can't look passed the results.
My opinion is much more valid than yours Andy, on this subject. I have worked with it, you haven't, as far as I am aware.
 

festa5

Well-known member
Dijksteele wasn't on the pitch when we shipped 3 in against QPR, but was when we scored two and looked far better, second half.

On a point of pedantry he started v QPR (partly responsible for the first goal, think Wilder specifically criticised him for it) was subbed at half time. We looked better when McNair replaced him.

But yeah, as it stands Dijksteel should be playing for now, he's looked much better recently and actually has a bit of pace (unless his form drops etc. of course).
 

festa5

Well-known member
It's an indicator of how many goals you haven't scored when you maybe should have

Which we already know because we can check the goals for column

Honestly what is the point of this statistic, worse than how much possesion we have had, or throw ins or corners, the reality is the only thing that matters is points
and we are severly lacking in them

We are better than we are because a bit of addign up mixed with a bit of guess work in a subjective mathematical model says so

In that's the case what's the point of any statistic that isn't on the league table? Might as well do away with all analysis then and just go on gut instinct Warnock style.

Pretty sure most pro clubs in the top 2 tiers will be using xG (amongst many other metrics) and qualified analysts, statisticians etc will be using that data to help analyse performances. You know, professionals who clearly think it adds value and understand the data.

Or the club should maybe just listen to some moaning blokes on a fans forum who have declared it's utterly useless largely, I suspect, on the grounds they don't really understand it and get rid of it and the rest of the statistical mumbo jumbo.

Tricky one 🤔
 

The Ruthsayer

Active member
and didn't take some great chances, had a number of ref decisions go against us and let in some really poor goals. It highlights what we know, the biggest problem is letting in some goals that should have been preventable.
And equally not scoring some that we should have.
 

Laughing

Well-known member
In that's the case what's the point of any statistic that isn't on the league table? Might as well do away with all analysis then and just go on gut instinct Warnock style.

Pretty sure most pro clubs in the top 2 tiers will be using xG (amongst many other metrics) and qualified analysts, statisticians etc will be using that data to help analyse performances. You know, professionals who clearly think it adds value and understand the data.

Or the club should maybe just listen to some moaning blokes on a fans forum who have declared it's utterly useless largely, I suspect, on the grounds they don't really understand it and get rid of it and the rest of the statistical mumbo jumbo.

Tricky one 🤔
The issue is that some think the stat is worthless and some put more store in it than they ought to. It's the best single indicator of performance that we have. But it still isn't very good, particularly on its own.

The truth lies somewhere in the middle. It's probably a little bit better better than corners but, probably, much worse than corners and possession used together.

I would largely agree that pointing to xG alone and claiming that we were unlucky shouldn't wash with fans, particularly over a sequence of games. Either the results or the xG tend toward the mean. However, in our case, just watching games should be enough to suggest we, probably, deserve a couple more points.
 

festa5

Well-known member
The issue is that some think the stat is worthless and some put more store in it than they ought to. It's the best single indicator of performance that we have. But it still isn't very good, particularly on its own.

The truth lies somewhere in the middle. It's probably a little bit better better than corners but, probably, much worse than corners and possession used together.

I would largely agree that pointing to xG alone and claiming that we were unlucky shouldn't wash with fans, particularly over a sequence of games. Either the results or the xG tend toward the mean. However, in our case, just watching games should be enough to suggest we, probably, deserve a couple more points.

Absolutely. The club won't be using it in isolation either.

I think the criticism of Wilder referring to it is unfair. He's not (as far as I'm aware) using to say the league table is lying or that everything is fine. He's pretty much saying "look, we are creating chances, we are causing problems for the opposition" which is a fair interpretation of the stat I think?

It's become a stick to beat him with for some, but not sure what they want to him to say "yeah we are **** and I'm ignoring all the data that suggests anything otherwise because the league table is all that matters"?
 
Top