[Albion] The xG comparisons to last season thread - includes graphs!

[Ⓩ-Ⓐ-Ⓜ-Ⓞ-Ⓡ-Ⓐ]

I know not everybody thinks xG is a useful stat, so for those who don't like it - look away now! I know it's not a perfect science but if you're interested keep reading...

First off let me say I'm no mathematician or stat anorak so if anyone has any suggestions on improvements please let me know!

The table uses "xGoals" to create an "xResult". For example in yesterdays game against Arsenal the xG was Arsenal 0.97 - 1.53 Brighton. If we round the xGoals to the nearest whole number we are able to get a result - in this case it would round to Arsenal 1 - 2 Brighton. In this case xGoals nails the actual result but it's not always like that....

For that reason I've also included ΔGF (The difference between our xGoals and our actual Goals) as well as ΔGA (The difference between our xGoals Against and our actual Goals Against)

Points and xPoints shows how many points we've gained in that match according to actual Goals and xGoals.

ΔPoints shows if we've gained more or less points than our xResult indicated. For example the West Ham game earlier this season we had an xResult of Brighton 1 - 0 West Ham, but the actual result was Brighton 1 - 1 West Ham meaning we lost 2 points from the xResult.

And finally taking a page out of [MENTION=3734]Giraffe[/MENTION]'s wonderful Tracker we have the points tally shown in CUM Points and xCUM Points.

I'm not usually very good at explaining these things, so the best way is to probably just look at the tables below and you'll be able to see what I've been doing!

All xG data is from www.understat.com


Season 2018-19









Season 2019-20




So what does this tell us so far? Well it really shows what I'm sure most of you remember from last season that when we had that run of 3 wins on the bounce last season we got very lucky against West Ham, Newcastle and Wolves. It shows that at this stage last season we were 10 points better off that our play "according to the xG stats" deserved. We'd scored 1 goal more than our play "deserved" and we "should" have conceded 6 more. This season it shows we are more or less doing as our play deserves, we our just 1 point behind our Cum x Points. The surprising thing is that our xG against Watford show we should have drawn the game 1-1, but I remember the game being much more one sided than that. Ah well, like I said before it's not a perfect science but I find it interesting all the same. I can imagine it's the sort of thing Tony looks at and is one of the reasons he decided a change was needed in the summer - we got very lucky with some of our results last season and only stayed up by the skin of our teeth (and Cardiff's ineptitude)

Viva la Potter revolution!

I'll try and update this thread periodically through the season.

 

[Tom Hark Preston Park]

I know not everybody thinks xG is a useful stat, so for those who don't like it - look away now! I know it's not a perfect science but if you're interested keep reading...

First off let me say I'm no mathematician or stat anorak so if anyone has any suggestions on improvements please let me know!

The table uses "xGoals" to create an "xResult". For example in yesterdays game against Arsenal the xG was Arsenal 0.97 - 1.53 Brighton. If we round the xGoals to the nearest whole number we are able to get a result - in this case it would round to Arsenal 1 - 2 Brighton. In this case xGoals nails the actual result but it's not always like that....

For that reason I've also included ΔGF (The difference between our xGoals and our actual Goals) as well as ΔGA (The difference between our xGoals Against and our actual Goals Against)

Points and xPoints shows how many points we've gained in that match according to actual Goals and xGoals.

ΔPoints shows if we've gained more or less points than our xResult indicated. For example the West Ham game earlier this season we had an xResult of Brighton 1 - 0 West Ham, but the actual result was Brighton 1 - 1 West Ham meaning we lost 2 points from the xResult.

And finally taking a page out of [MENTION=3734]Giraffe[/MENTION]'s wonderful Tracker we have the points tally shown in CUM Points and xCUM Points.

I'm not usually very good at explaining these things, so the best way is to probably just look at the tables below and you'll be able to see what I've been doing!

All xG data is from www.understat.com


Season 2018-19

View attachment 117905
View attachment 117907






Season 2019-20

View attachment 117906
View attachment 117908

So what does this tell us so far? Well it really shows what I'm sure most of you remember from last season that when we had that run of 3 wins on the bounce last season we got very lucky against West Ham, Newcastle and Wolves. It shows that at this stage last season we were 10 points better off that our play "according to the xG stats" deserved. We'd scored 1 goal more than our play "deserved" and we "should" have conceded 6 more. This season it shows we are more or less doing as our play deserves, we our just 1 point behind our Cum x Points. The surprising thing is that our xG against Watford show we should have drawn the game 1-1, but I remember the game being much more one sided than that. Ah well, like I said before it's not a perfect science but I find it interesting all the same. I can imagine it's the sort of thing Tony looks at and is one of the reasons he decided a change was needed in the summer - we got very lucky with some of our results last season and only stayed up by the skin of our teeth (and Cardiff's ineptitude)

Viva la Potter revolution!

I'll try and update this thread periodically through the season.

tl:dr

OP is doubtless massively over-thinking this. Maybe just use your eyes?

 

[Badger]

tl:dr

OP is doubtless massively over-thinking this. Maybe just use your eyes?

Well I thought it was quite interesting

 

[Papa Lazarou]

tl:dr

OP is doubtless massively over-thinking this. Maybe just use your eyes?

A bit rude?

Well I thought it was quite interesting

As did I. Thanks @Ⓩ-Ⓐ-Ⓜ-Ⓞ-Ⓡ-Ⓐ

 

[MattBackHome]

Chapeau, @Ⓩ-Ⓐ-Ⓜ-Ⓞ-Ⓡ-Ⓐ.

Keep em coming.

 

[Papa Lazarou]

How come I can't @ Zamora?

 

[Baldseagull]

I know not everybody thinks xG is a useful stat, so for those who don't like it - look away now! I know it's not a perfect science but if you're interested keep reading...

First off let me say I'm no mathematician or stat anorak so if anyone has any suggestions on improvements please let me know!

The table uses "xGoals" to create an "xResult". For example in yesterdays game against Arsenal the xG was Arsenal 0.97 - 1.53 Brighton. If we round the xGoals to the nearest whole number we are able to get a result - in this case it would round to Arsenal 1 - 2 Brighton. In this case xGoals nails the actual result but it's not always like that....

For that reason I've also included ΔGF (The difference between our xGoals and our actual Goals) as well as ΔGA (The difference between our xGoals Against and our actual Goals Against)

Points and xPoints shows how many points we've gained in that match according to actual Goals and xGoals.

ΔPoints shows if we've gained more or less points than our xResult indicated. For example the West Ham game earlier this season we had an xResult of Brighton 1 - 0 West Ham, but the actual result was Brighton 1 - 1 West Ham meaning we lost 2 points from the xResult.

And finally taking a page out of [MENTION=3734]Giraffe[/MENTION]'s wonderful Tracker we have the points tally shown in CUM Points and xCUM Points.

I'm not usually very good at explaining these things, so the best way is to probably just look at the tables below and you'll be able to see what I've been doing!

All xG data is from www.understat.com


Season 2018-19

View attachment 117905
View attachment 117907






Season 2019-20

View attachment 117906
View attachment 117908

So what does this tell us so far? Well it really shows what I'm sure most of you remember from last season that when we had that run of 3 wins on the bounce last season we got very lucky against West Ham, Newcastle and Wolves. It shows that at this stage last season we were 10 points better off that our play "according to the xG stats" deserved. We'd scored 1 goal more than our play "deserved" and we "should" have conceded 6 more. This season it shows we are more or less doing as our play deserves, we our just 1 point behind our Cum x Points. The surprising thing is that our xG against Watford show we should have drawn the game 1-1, but I remember the game being much more one sided than that. Ah well, like I said before it's not a perfect science but I find it interesting all the same. I can imagine it's the sort of thing Tony looks at and is one of the reasons he decided a change was needed in the summer - we got very lucky with some of our results last season and only stayed up by the skin of our teeth (and Cardiff's ineptitude)

Viva la Potter revolution!

I'll try and update this thread periodically through the season.

Alternative conclusion, expected goal formula does not work accurately against teams that focus on defensive play? If my alternative conclusion is correct, then xg against us should not be so high, and we did not get lucky, we got what we deserved.

 

[Brok]

How come I can't @ Zamora?

Your keyboard may have run out of circles and blue ink.

 

[Uter]

Alternative conclusion, expected goal formula does not work accurately against teams that focus on defensive play? If my alternative conclusion is correct, then xg against us should not be so high, and we did not get lucky, we got what we deserved.

If your hypothesis was correct then you'd expect us to have out performed our expected points throughout the season. But we didn't. We came crashing down in the second half of the season. You can only outrun lady luck for so long.

 

[Uter]

So what does this tell us so far? Well it really shows what I'm sure most of you remember from last season that when we had that run of 3 wins on the bounce last season we got very lucky against West Ham, Newcastle and Wolves. It shows that at this stage last season we were 10 points better off that our play "according to the xG stats" deserved. We'd scored 1 goal more than our play "deserved" and we "should" have conceded 6 more. This season it shows we are more or less doing as our play deserves, we our just 1 point behind our Cum x Points. The surprising thing is that our xG against Watford show we should have drawn the game 1-1, but I remember the game being much more one sided than that. Ah well, like I said before it's not a perfect science but I find it interesting all the same. I can imagine it's the sort of thing Tony looks at and is one of the reasons he decided a change was needed in the summer - we got very lucky with some of our results last season and only stayed up by the skin of our teeth (and Cardiff's ineptitude)

Viva la Potter revolution!

I'll try and update this thread periodically through the season.

Great stuff! Keep it up.

Re Watford, you don't goal any xG for OGs, even if it's a situation where a goal was a reasonably likely outcome. So that will go some way to explain it.

xGA seems a lot more variable this season. We seem to be generally good at keeping xGA down to respectable level, but have an aberration every few games. Not ideal, but that's probably preferable to having the same total xGA spread evenly across all games.

 

[Jim in the West]

If your hypothesis was correct then you'd expect us to have out performed our expected points throughout the season. But we didn't. We came crashing down in the second half of the season. You can only outrun lady luck for so long.

Yes - I think that’s why it’s a useful analysis. We had a lot of luck in the first half of last season, and it couldn’t last (we should have seen the warning lights....maybe some did).

I think it suggests that this season we should carry on at roughly the same level of points accumulation...meaning we should be safe.

 

[narly101]

A bit rude?

A bit rude? It was wholly unnecessary.

I like the idea. be interesting to see how it goes for the season. I must admit that win streak of 3 games was incredibly lucky considering the football we were playing, and that shows it.

 

[dingodan]

Very impressive OP, but can you play LB?

 

[Papa Lazarou]

A bit rude? It was wholly unnecessary.

I like the idea. be interesting to see how it goes for the season. I must admit that win streak of 3 games was incredibly lucky considering the football we were playing, and that shows it.

Let's see how the stats stack up for the upcoming 3 game winning streak.

 

[Kalimantan Gull]

I have an xG question that has been bugging me for ages.

Our first goal against Spurs: When Gross slung the cross in straight at Lloris, the xG score for that scenario would have been almost zero. When Lloris dropped it at Maupay's feet, the xG would be almost 1. So what value goes in the book?

 

[Baldseagull]

If your hypothesis was correct then you'd expect us to have out performed our expected points throughout the season. But we didn't. We came crashing down in the second half of the season. You can only outrun lady luck for so long.

We could have outperformed xg goal difference and still lost, you would have to show me the stats that show that we conceded as many or more than xg would predict, to prove your point. For example, xg could predict a 3-0 defeat but we only lost 1-0, no improvement on expected points, but still doing better than xg against.

 

[warmleyseagull]

Good stuff. xG is quite a good measure of the amount of and quality of attacking play, however it is vulnerable to the ‘Vardy’ issue, namely that quality strikers make more of expected chances than others. I mention Vardy because Leicester are the masters this season, having scored twice their expected goals: see https://www.reddit.com/r/FantasyPL/comments/dr73na/leicesters_xg_over_performance_is_incredible/. The article is a month old but still valid.

Albion’s problem has been conversion of chances. Much has been made of having 5/6 players in oppo box at times. That will generate xG but not goals unless the finishing and in box decision making improve. Albion are not scoring enough from open play currently. Doing most other things very well but sharper finishing would make us a real force.

 

[GingerBeerMan]

I have an xG question that has been bugging me for ages.

Our first goal against Spurs: When Gross slung the cross in straight at Lloris, the xG score for that scenario would have been almost zero. When Lloris dropped it at Maupay's feet, the xG would be almost 1. So what value goes in the book?

I imagine it adds up, to give a total at the end of the game. Thus, xG would be 1 because the drop did occur. Add that to the xG of 0 from the initial cross and you get 1.

 

[Kalimantan Gull]

I imagine it adds up, to give a total at the end of the game. Thus, xG would be 1 because the drop did occur. Add that to the xG of 0 from the initial cross and you get 1.

But take a goal mouth scramble, where players are taking potshots from 2 yards and the keeper makes a brilliant double save before they finally score a tap in. The total xG would be greater than 1 for a single phase of play. If that is an error inherent in the system so be it, but it devalues the stat some what

 

[BN41Albion]

Well I thought it was quite interesting

Yeah i thought very interesting, too. hopefully bodes well for the rest of the season, especially as we hopefully continue to improve under potter's revolution.

It really shows how why we dropped like a stone when our luck ran out last season. We got pretty lucky in staying up!