I'm obsessed with XG

XG is another way for nerds to think they actually know something about football. Suits the new fans brought along by Sky. So many variables, like when McGeough had that shot but the ball bobbled up. You can’t account for it.
 


I've taken another look at this (as it won't go away) and despite my massive reservations on so many aspects of it, I'm prepared to suggest that maybe there is something (very small) in it. The guy who posted on here really didn't do the cause any good - with his suggestions that it shows we were lucky or were outperformimg and therefore we will 'probably' drop away. I'm looking at it a completely different way. Those teams that conform to the 'lots of good shots and stopping lots of good efforts wins you games' theory (it's not a theory its generally how the game works) tell us very little, its all about the outliers. Let me suggest the fact that we are an outlier does not mean we are overperforming , it suggests we have the necessary quality to win through, it actually suggest we WILL walk this league as things are not going wonderfully on the stats yet we win - so either this continues (why should it change) or we get better, either way we win. Equally Luton will NOT rise as predicted they are UNDER performing because they do not have the quality, again either it will continue or they will stop being so good on these stats. I believe xG does not suggest that Luton will rise and we will fall, it actually highlights these teams actual quality as being out of line, their productive efficiency if you like.

My xG prediction: Sunderland 1st, Luton outside top six.
 
Last edited:
chickenbaltipie said:
I'd argue the complete opposite from what I have seen and read.

Sunderland's xG against Charlton will bare no relationship to the performance of the team against Wycombe. Any similarities will be purely coincidental. Different personnel, different context, different form, different tactics, etc.

You could argue that the game against Plymouth, is a better predictor of Wycombe - as the form, personnel & tactics are more likely to be similar. Intangibles like team chemistry and individual player confidence are transient. In that sense, you could apply a moving average, weighting more recent performances as being of higher value - again, something that day traders do when looking at the stock market.

Put simply, performance metrics drastically depreciate in value as they age. It's the reason form guides tend to only look at the last 5-6 games. So I'd push this forward in time and say the same applies in reverse. You could look at weighted xG and try and use it to predict the outcomes over the next month. But beyond that... flip a coin.
Click to expand...
I don't think you can use it to predict any game with any more certainty than 'that Man City are a good team, I think they'll win'.

The data does show a correlation over large data sets, but you'd expect that with a distribution curve. Over the course of a season, goals scored and conceded does, in most instances, correspond, reasonably well, with actual goals scored and conceded. That doesn't mean it is reliable as a predictor, though... example: when Chelsea won the league by skewing their gx by 23 goals, Man City won the league by 9 points on predicted xg. There will always be outliers, due to external factors. All that said, I'm not entirely sure what xg is actually good for.
 
You don't need to be a mathematician to recognise that we've been far more clinical in front of goal this season than most of our opponents, anyone who has watched our games will have noticed it. Plenty of teams in this league (and Championship) are decent tee to green but just cannot finish or mess up the final pass when it matters - the attacking quality just isn't there. Plus we have probably the best keeper in the league. On the other hand, when we go forward we're pretty ruthless most of the time, simply because the likes of Maja and Maguire in particular so far this season (also Honeyman, Gooch, Power and McGeady) have the quality to strike a ball into one of the corners of the goal from inside and outside the box (you get a touch more time and space in this league too which helps). It really is as simple as that.
 
phunboy said:
The score is the main thing but after a little bit of digging I actually like it. What it's telling us is offensively we're very clinical around and further back from the 18 yard line. defensivly while we concede plenty of chances we somehow don't conceded much
Click to expand...

We've got a good keeper.

Easy this.
 
stephen cartwright said:
Good points, but I’m not basing safc on this xg model. I’ll stick with the reality of the table and my own eyes

There is no consideration taken into the sending offs we’ve had, the lack of the best team out and general tactics

If these stats were spot on, every single game would be known prior to kick off and you couldn’t get a bet on

Is the weather included in the equation? Are the referees? Any crowd numbers involved?
Click to expand...

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.
 
SAFCOldie said:
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.
Click to expand...
Miskicked shots go in. Deflections, ogs, dodgy bounce of the ball, weather conditions, the list is endless.

This is based on computer says yes, so it must be correct
 
chave said:
Don't have these to hand (although can put out together from experimental 361s site), however from talking to a few of the people who run the data, I know that Barnsley's figures have been decreasing over the last month, while Portsmouth's have been holding steady.
Click to expand...

We’re missing our best player, he dictated the pace of the game by breaking up most opposition attacks, and getting us on the front foot very quickly with his fantastic passing abilities. It was inevitable we’d struggle to keep up the early season form without him, but he’s due back in January so hopefully we’ll be back on tack then.
 
RascaI said:
We’re missing our best player, he dictated the pace of the game by breaking up most opposition attacks, and getting us on the front foot very quickly with his fantastic passing abilities. It was inevitable we’d struggle to keep up the early season form without him, but he’s due back in January so hopefully we’ll be back on tack then.
Click to expand...
Who's that and when has he been out since?
 
stephen cartwright said:
could someone predict three results this week according to the XG statistics?

scores as well
Click to expand...
The xG (and xGD in brackets) for all league one games this weekend:
Fleetwood 1.13 (-0.19) Walsall 1.24 (-0.32)
Oxford 1.04 (-0.13)Gillingham 1.15 (-0.15)
Doncaster 1.37 (0.12) Wimbledon 1.17 (0.24)
Sunderland 1.26 (-0.1) Wycombe 1.24 (-0.02)
Southend 1.26 (0.2) Blackpool 1.44 (0.38)
Burton 1.36 (0.04) Coventry 1.38 (0.42)
Peterborough1.31 (-0.05) Bradford 1.04 (-0.79)
Luton 1.42 (0.41) Plymouth 1.01 (-0.86)
Stanley 1.42 (-0.02) Barnsley 1.44 (0.56)
Shrewsbury 1.11 (0.16) Rochdale 1.44 (0.01)
Bristol 1.01 (-0.16) Scunthorpe 1.22 (-0.42)

From those, I'd choose
Oxford to draw with Gillingham
Luton to beat Plymouth
Peterborough to beat Bradford

Wouldn't go for scores, as xG doesn't predict that as well. If those are too likely, I would also pick
Sunderland to draw with Wycombe
Barnsley to beat Stanley
Burton to draw with Coventry
Fleetwood to draw with Walsall

In reality, I'd be happy with four of those seven being right...
 
stephen cartwright said:
could someone predict three results this week according to the XG statistics?

scores as well
Click to expand...
No, because that's not how it works.

Just did a comparison of a XP after 14 games for the PL last season against actual points for the end of season. There is reasonable correlation, although there are significant outliers e.g. Burnley get relegated according to XP, but finish 7th in real life. Bournemouth also get relegated on XP, but, in reality finish 12th, whereas Stoke finish an imaginary 12th, but in reality get relegated. Southampton are set for an imaginary 7th place, but, in reality, finish 17th. That's 20% of the teams are huge outliers.

When you look at XP in terms of 20% of the time, xg / xp is massively wrong, it doesn't look an infallible system.
 
Last edited:
I don't believe it, Barnsley and Peterborough win, whilst safc don't

that's that then

Windypants said:
No, because that's not how it works.

Just did a comparison of a XP after 14 games for the PL last season against actual points for the end of season. There is reasonable correlation, although there are significant outliers e.g. Burnley get relegated according to XP, but finish 7th in real life. Bournemouth also get relegated on XP, but, in reality finish 12th, whereas Stoke finish an imaginary 12th, but in reality get relegated. Southampton are set for an imaginary 7th place, but, in reality, finish 17th. That's 20% of the teams are huge outliers.

When you look at XP in terms of 20% of the time, xg / xp is massively wrong, it doesn't look an infallible system.
Click to expand...
But is not exactly that predicting expected goals in a game?

So what are the expected scores for three games this coming weekend?
 
Last edited by a moderator:
stephen cartwright said:
I don't believe it, Barnsley and Peterborough win, whilst safc don't

that's that then


But is not exactly that predicting expected goals in a game?

So what are the expected scores for three games this coming weekend?
Click to expand...
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.
 

Back
Top