I'm obsessed with XG

[chave]

Sorry to be a pain but do you have the equivalent for Barnsley. From reading their forums I’d say their perception of their team’s performance is the opposite of ours with most thinking they’ve been getting progressively poorer over the last month or so.

As luck would have it, Experimental 361 has put out some more graphs, showing long term trends of xG against actual goal difference over the last two years. Also shows examples of other clubs who exceeded xG for good portions of last year. Worth a look if you're interested by this sort of thing

League 1 trends, 12 Nov 2018

Also shows Barnsley's drop in both xG and actual goal difference, which are almost identical.

 

[phunboy]

Isn't xg meant to be a long lead predictor, rather than a result predictor?

Using xg to predict specific results is like using the half life of uranium to measure when your egg is soft boiled i.e. f***ing useless.

Its not even meant to be that but some stattos use it for that. Its an analytical tool for managers and sports analysts to be used in conjunction with other statistics to highlight areas to improve on.

 

[Howard the Duck]

As luck would have it, Experimental 361 has put out some more graphs, showing long term trends of xG against actual goal difference over the last two years. Also shows examples of other clubs who exceeded xG for good portions of last year. Worth a look if you're interested by this sort of thing

League 1 trends, 12 Nov 2018

Also shows Barnsley's drop in both xG and actual goal difference, which are almost identical.

Class, thanks. That’s exactly what I was after.
Shows what you said earlier in the thread.

 

[pfceggchaser]

xG is an interesting one. As Chave has pointed to, I am a big fan of the work Ben Mayhew does, who always maintained we would win League 2 in 2016/17, when we looked for a long spell a clear 3rd or 4th place side in the league table. xG is looked at in analysis by clubs in general as an area where you can improve on, and where the opposition are weak. It's not everything though, as you often get sides that buck the trend. Plymouth for a number of years now have overachieved against their xG in matches.

For us anyway, the recent postings suggest we will have the best defence by the end of the season. But if current trends continue, we will be out-scored by Barnsley, yourselves and Peterborough, although we seem to pose more of a threat than yourselves and Posh on 'balance of play'. I'm sure our manager knows this and will be in the market for a centre forward in January and we will need to resist any big bids for Clarke.

Ignoring the xG for a second, your performance stats suggest to me you have good technicians in your team, and players capable of 'moments'. McGeady, Gooch and Maguire are all good strikers of a football, and the goals you scored in the game at the weekend suggest that. If teams do not close you down 25-30 yards from goal, you have players more likely to score. Maja too is taking his chances. Posh are similar because they have Godden who is an utter machine if you give him a chance.

 

[RascaI]

Who's that and when has he been out since?

Kenneth Dougall, hes missed the last 4 and half games, which coincides with the drop in this bollocks.

 

[Charmless Man]

Exactly, it can't factor in injuries/red cards etc and a change in personnel e.g. the future involvement of Wyke, Watmore for starters.

It does factor in all those things, as they feed into it. Just seems it hasn’t been explained/interpreted very well here.

Last I looked we led the league for red cards. Less players on the pitch for more time will naturally lead to lower xGF and higher xGA. It doesn’t account for it all, but all those little factors probably do add up.

Fwiw - I think Maja might fall away and he’s the real big outlier in our team. I suspect we will have enough without him though. And if we’re using Wyke who we assume can hold a ball up in bring people into play, then we may well see an uptick in xG. Maja doesn’t really do that. He just confounds xG predictions by stroking the ball into the net.

Also, the lad has alluded to it but the real big issue is match situation. There’s been a few games where we’ve raced into a comfortable lead taking basically every chance that fell and then coasted thereafter. That’s the benefit of context you don’t have if you aren’t watching the games regularly.

I don't believe it, Barnsley and Peterborough win, whilst safc don't

that's that then

It’s not but you’re being a little silly about it. Over long periods of time sides tend toward what xG suggests for them. Either by not nicking wins/throwing them away when they’re doing better or by improving/declining. This is what people seem to be getting wrong - it’s a measure of how likely sustaining a run of results might be given how a team has been playing. Not a foolproof method for predicting the outcome of a league.

It’s flawed, yes, but there does seem to be some value in it.

 

[SAFCOldie]

It does factor in all those things, as they feed into it. Just seems it hasn’t been explained/interpreted very well here.

Last I looked we led the league for red cards. Less players on the pitch for more time will naturally lead to lower xGF and higher xGA. It doesn’t account for it all, but all those little factors probably do add up.

Fwiw - I think Maja might fall away and he’s the real big outlier in our team. I suspect we will have enough without him though. And if we’re using Wyke who we assume can hold a ball up in bring people into play, then we may well see an uptick in xG. Maja doesn’t really do that. He just confounds xG predictions by stroking the ball into the net.

Also, the lad has alluded to it but the real big issue is match situation. There’s been a few games where we’ve raced into a comfortable lead taking basically every chance that fell and then coasted thereafter. That’s the benefit of context you don’t have if you aren’t watching the games regularly.



It’s not but you’re being a little silly about it. Over long periods of time sides tend toward what xG suggests for them. Either by not nicking wins/throwing them away when they’re doing better or by improving/declining. This is what people seem to be getting wrong - it’s a measure of how likely sustaining a run of results might be given how a team has been playing. Not a foolproof method for predicting the outcome of a league.

It’s flawed, yes, but there does seem to be some value in it.

Cheers for info.

 

[stephen cartwright]

It does factor in all those things, as they feed into it. Just seems it hasn’t been explained/interpreted very well here.

Last I looked we led the league for red cards. Less players on the pitch for more time will naturally lead to lower xGF and higher xGA. It doesn’t account for it all, but all those little factors probably do add up.

Fwiw - I think Maja might fall away and he’s the real big outlier in our team. I suspect we will have enough without him though. And if we’re using Wyke who we assume can hold a ball up in bring people into play, then we may well see an uptick in xG. Maja doesn’t really do that. He just confounds xG predictions by stroking the ball into the net.

Also, the lad has alluded to it but the real big issue is match situation. There’s been a few games where we’ve raced into a comfortable lead taking basically every chance that fell and then coasted thereafter. That’s the benefit of context you don’t have if you aren’t watching the games regularly.



It’s not but you’re being a little silly about it. Over long periods of time sides tend toward what xG suggests for them. Either by not nicking wins/throwing them away when they’re doing better or by improving/declining. This is what people seem to be getting wrong - it’s a measure of how likely sustaining a run of results might be given how a team has been playing. Not a foolproof method for predicting the outcome of a league.

It’s flawed, yes, but there does seem to be some value in it.

Does the equation account for sending offs, yellow cards, weather conditions, pitch condition, players playing when not 100%, deflected goals, miskicked goals, referee, location of the game, standard of the pitch, game on tv, mistakes by officials and if there is a partridge in a pear tree?

 

[Charmless Man]

Does the equation account for sending offs, yellow cards, weather conditions, pitch condition, players playing when not 100%, deflected goals, miskicked goals, referee, location of the game, standard of the pitch, game on tv, mistakes by officials and if there is a partridge in a pear tree?

I’ve just said man. Yes things like red cards yellow cards feed in because it’s a measure of the chances a team creates and allows the opposition to have.

Have less players on the pitch? It’s likely you’ll have less chances and opposition will have more.

It does also represent game situation, but can give a false impression of a team’s play because of that. E.g if you blast a team away in the first half and then coast it gives you a false impression of chances created.

Think it could be worth actually looking into what it’s about before poo pooing it. It’s not a be all and end all but it’s not as useless as you seem to think. Again, the evidence suggests it’s usually relatively accurate.

 

[andymcnish]

I

Over long periods of time sides tend toward what xG suggests for them.

Except when they don't. Like Stoke last year, Chelsea when they won the league with Costa and, from the looks of it, us this year.

Every year in every league there are 'outliers' who significantly out or under perform xG over the whole season.

Some of this may no doubt be statistical variance (or luck as we normally call it) but some of it is undoubtedly because of flaws in the model, which everyone can see are there.

Primarily that Diego Costa bearing down on Lee Camp is given the same xG score as Jozy Altidore v Pickford. The relative quaility of players in critical areas of the pitch really matters. An assumption that all players in a match are of the same quality is obviously flawed and can and does significantly affect the accuracy of the model.

So whilst the number and quality of chances created and given away in a match is obviously useful data for the coach/manager, it, on its own, is not quite the moneyball for soccer statto nirvana it's portrayed as being.

 

[stephen cartwright]

I’ve just said man. Yes things like red cards yellow cards feed in because it’s a measure of the chances a team creates and allows the opposition to have.

Have less players on the pitch? It’s likely you’ll have less chances and opposition will have more.

It does also represent game situation, but can give a false impression of a team’s play because of that. E.g if you blast a team away in the first half and then coast it gives you a false impression of chances created.

Think it could be worth actually looking into what it’s about before poo pooing it. It’s not a be all and end all but it’s not as useless as you seem to think. Again, the evidence suggests it’s usually relatively accurate.

I’m all for accurate equations helping punters, and I have read about it, but imo, it does not reflect reality as humans play the game not computer controlled statistical beings.

If it was perfect, why play the games as the outcomes are known and week by week someone would win millions using the equation.

For all the calculations out there the simple fact of safc being second top scorers and the least conceded is the actual reality

If xg was accurate we would be mid table.

 

[Charmless Man]

I


Except when they don't. Like Stoke last year, Chelsea when they won the league with Costa and, from the looks of it, us this year.

Every year in every league there are 'outliers' who significantly out or under perform xG over the whole season.

Some of this may no doubt be statistical variance (or luck as we normally call it) but some of it is undoubtedly because of flaws in the model, which everyone can see are there.

Primarily that Diego Costa bearing down on Lee Camp is given the same xG score as Jozy Altidore v Pickford. The relative quaility of players in critical areas of the pitch really matters. An assumption that all players in a match are of the same quality is obviously flawed and can and does significantly affect the accuracy of the model.

So whilst the number and quality of chances created and given away in a match is obviously useful data for the coach/manager, it, on its own, is not quite the moneyball for soccer statto nirvana it's portrayed as being.

Aye they do it for a while, but eventually they do revert. See Leicester.

One of the most interesting aspects of it is how and why teams do that. A lot of people seem to ascribe it to ‘luck’ but clearly that can’t be the case. I actually suspect most of it relies in player mentality but that’s just a hunch.

 

[flusheding]

Aye they do it for a while, but eventually they do revert. See Leicester.

One of the most interesting aspects of it is how and why teams do that. A lot of people seem to ascribe it to ‘luck’ but clearly that can’t be the case. I actually suspect most of it relies in player mentality but that’s just a hunch.

what is 'form' if not sustained luck. A player scoring 10 games in a row is fairly rare but a player scoring 10 goals in a season of 38 matches is normal.

 

[eddtheduck]

I just check the scores

 

[andymcnish]

Aye they do it for a while, but eventually they do revert. See Leicester.

One of the most interesting aspects of it is how and why teams do that. A lot of people seem to ascribe it to ‘luck’ but clearly that can’t be the case. I actually suspect most of it relies in player mentality but that’s just a hunch.

Hang on - general squad mentality should be factored in already according to the stats geeks - as motivated players should create more chances at one end and prevent more at the other and 'that is all that matters'.

The gap in the model remains player quality - in particular clinicality of finishing for players who get chances and GK perfomance in saving shots. I suppose a player who is confident will perform better in those areas too, so mentality helps, but underlying technical skill is more important than the odd purple patch for a player over a season.

This reversion to the mean is also a concept much misuderstood by fans. Coins have no memory - you are never 'owed' bad luck to make up for previous good luck. If you 'run hot' in terms of good luck for half a season, the expectancy is that you will run with average luck over the remainder - meaning you will still have a luckier than usual season overall.

A s for your examples of reversion Leicester lost Kante (who did the job of 2 men as a defensive midfielder); Chelsea lost Costa who had bulldozed them to the title. They weren't the same teams the following years. The examples almost prove the opposite - that key player technical ability (or lack of it) is what breaks the model and leads to outliers.

Overall if 20% of the teams buck the model for a full season and have to be termed 'outliers', then the model is of limited value as a predictor of anything.

You can't tell which of the teams currently outperforming or underperforming 'the norm' will revert towards it by season's end.

Nor can you tell how many teams currently close to the norm may shift away from it.

It might suggest how a league is likely to shape up in general - but you could just go off player salaries and probably get as accurate a prediction.

It's a speciifc tool for coaches when analysing a game that the stat geeks have jumped on and tried to make the be all and end all of 'soccer' and that's bollox.

 

[Windypants]

Aye they do it for a while, but eventually they do revert. See Leicester.

One of the most interesting aspects of it is how and why teams do that. A lot of people seem to ascribe it to ‘luck’ but clearly that can’t be the case. I actually suspect most of it relies in player mentality but that’s just a hunch.

Leicester weren't much of an xg outlier when they won the league, were they? Just checked... bang on for xg. Xga was better than predicted (9 for the season).


P.S. Kante leaving might account for Leicester not being as good the following season... who'd have thought, sell your best player and you get worse.

 

[chave]

Chelsea lost Costa who had bulldozed them to the title. They weren't the same teams the following years. The examples almost prove the opposite - that key player technical ability (or lack of it) is what breaks the model and leads to outliers.

A quick note about Costa - when he won the title with Chelsea, his xG was 0.13 per 90 minutes less than he actually scored, which equates to about 5 goals over the course of the season. The season before with Chelsea, and the two seasons since, he's scored marginally less than his expected heals, which again suggests he had one lucky/hot season where he outperformed expectations.

For comparison, Josh Maja is currently about 0.4 per 90 minutes, so 3 times as much of an outlier than Costa.

 

[Windypants]

A quick note about Costa - when he won the title with Chelsea, his xG was 0.13 per 90 minutes less than he actually scored, which equates to about 5 goals over the course of the season. The season before with Chelsea, and the two seasons since, he's scored marginally less than his expected heals, which again suggests he had one lucky/hot season where he outperformed expectations.

For comparison, Josh Maja is currently about 0.4 per 90 minutes, so 3 times as much of an outlier than Costa.

In his three seasons at Chelsea, in two of them he exceeded xg by about 5 goals each season. In the one he didn't, Mourinho was having his melt down, Costa was suspended and mired in angry man controversy, so missed a lot of games, Hazard missed quite a lot of games and, in general, Chelsea were in chaos. I remember watching them... they were shit and didn't look like they wanted to play for Mourinho. In his two seasons since leaving Chelsea, he's played less than 20 games in total. According to you, the 70-80 games where he's consistently exceeded xg are lucky, whereas the seasons where he was disrupted by injury and suspensions were 'normal'.

P.S. Re. Maja... this is a crap division with crap defenders and goalkeepers. The disparity in ability (and, by extension xg) should be amplified, shouldn't it?

All that aside, you're ducking the basic premise of 'the sum of the parts'. Also, you're conveniently ignoring the fundamental fact that teams at the top of leagues exceed xg, whereas, teams at the bottom tend to underachieve. Example: in the year that Chelsea exceeded xg by 23 goals, I think that five of the top six significantly exceeded xg.

You're trying to do things with xg that it simply won't do. You can't use xg to argue that players who are exceeding expectation, on account of being better players in poorer divisions are lucky, because, statistically, it doesn't hold water.

 

[chave]

In his three seasons at Chelsea, in two of them he exceeded xg by about 5 goals each season. In the one he didn't, Mourinho was having his melt down, Costa was suspended and mired in angry man controversy, so missed a lot of games, Hazard missed quite a lot of games and, in general, Chelsea were in chaos. I remember watching them... they were shit and didn't look like they wanted to play for Mourinho. In his two seasons since leaving Chelsea, he's played less than 20 games in total. According to you, the 70-80 games where he's consistently exceeded xg are lucky, whereas the seasons where he was disrupted by injury and suspensions were 'normal'.

P.S. Re. Maja... this is a crap division with crap defenders and goalkeepers. The disparity in ability (and, by extension xg) should be amplified, shouldn't it?

All that aside, you're ducking the basic premise of 'the sum of the parts'. Also, you're conveniently ignoring the fundamental fact that teams at the top of leagues exceed xg, whereas, teams at the bottom tend to underachieve. Example: in the year that Chelsea exceeded xg by 23 goals, I think that five of the top six significantly exceeded xg.

You're trying to do things with xg that it simply won't do. You can't use xg to argue that players who are exceeding expectation, on account of being better players in poorer divisions are lucky, because, statistically, it doesn't hold water.

I don't disagree with any of that, and I don't think anyone is disagreeing that player ability plays a large part in this. However, it seems as though the difference in skill still doesn't account for how big the gap is. Teams like Man City may have a similar xG difference to Sunderland, however they're expected to win comfortably on xG. If Sunderland's players are better than the rest of the league, surely xG should also be high. There's no record of any team doing what Sunderland are currently doing on an extended basis, and that's why it's an anomaly that most people think isn't sustainable.

For Maja specifically, the nearest players to his xG/goals ratio in the Prem are Hazard and Richarlison. Again, they're both great players but unlikely to carry on their current form for the entirety of the season.

 

[Windypants]

I don't disagree with any of that, and I don't think anyone is disagreeing that player ability plays a large part in this. However, it seems as though the difference in skill still doesn't account for how big the gap is. Teams like Man City may have a similar xG difference to Sunderland, however they're expected to win comfortably on xG. If Sunderland's players are better than the rest of the league, surely xG should also be high. There's no record of any team doing what Sunderland are currently doing on an extended basis, and that's why it's an anomaly that most people think isn't sustainable.

For Maja specifically, the nearest players to his xG/goals ratio in the Prem are Hazard and Richarlison. Again, they're both great players but unlikely to carry on their current form for the entirety of the season.

Maja scored again, last night... did fuck all for 90 minutes, then got the ball in the box. Like so many of his shots, this one was on target and just inside the post giving the 'keeper little chance - it's almost like he's aiming, or something.