To see less ads Register or Login ----- Daily Fantasy Sports games 18+

One stat to rule them all (xG/xA)

A Fantasy Football forum for news on fantasy football games run by the Premierleague (FPL).
User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

One stat to rule them all (xG/xA)

Post by Stemania »

To blatantly steal a phrase that's been used in the FPL context before, it seems the era of one stat to rule them all may nearly be upon us. Now being displayed regularly on TV via MOTD (leading to many a geeky orgasm), the stat of expected goals (xG) has finally made it to the mainstream. The bbc even produced an introductory guide at the start of the season. All the stat fanciers among us have been jealously chattering about its potential significance/usefulness in FPL for a few years - as in the linky, it's basically an aggregate measure of how many goals a player 'should' have got from the precise position/locations they've taken their shots (edit: BEWARE, that is: how many the average striker would have scored from those chances) - but we've long had the same problem:

One stat to rule them all, get all the stats and bind them,
Well, two stats (xG and xA), but where the hell to find them?


Now OPTA have started putting out an xG stat it seems the availability issue may be loosening, meaning xG could (hopefully) become the new language of fantasy football talk. More and more fantasy football articles/blogs had been mentioning xG over the last few years (along with expected assists (xA) based on chances created). The seminal article has always seemed to be this marvelous one by 11tegen11, where evidence was presented to suggest xG is a better predictor of future goals than the traditional stat pervert's staples of shots in the box (SiB)/shots on target (SoT)/total shots (TS) etc. Edit: Oddly, it seems to have been linked far more than this one by xG maker Michael Caley on the same thing, which I didn't know about until just now. :)

But anyway, drumroll over. There seems to be very good free resources that have hit the internet and are worth sharing. Mav3rick found the following gem (an independent non-OPTA source of xG/xA) a few weeks back - so credit to him - and it's been slowly permeating various other sites/blogs too:

https://understat.com/

As is common with all sources of xG/xA, the process/formula used to calculate their particular numbers is not provided (so it's up to the reader what level of credence to give them). But this particular site has some quite ridiculous features - it has xG/xA for every player, you can filter players by position in formation (so for example, you can see the difference in Antonio's different xG at RB vs RWB vs RW etc - press 'Position') for the last three years. It also can be sorted by set pieces, shot position and body part used!! It also has xG for teams (so for example, Swansea have actually been the worst defence this season in their metric, and City the best), and for every shot taken or chance created it has the precise xG/xA. Dino has already been making some excellent posts using these numbers - and twas he who suggested we start this thread. :D

Also worth pointing out is the outstanding blog of Chris Glover (http://www.plfantasy.com/), which contains a whole host of OPTA's (?) xG numbers and some quite exceptional analysis and ways of displaying them (just look under 'Team Snapshot' for example!!). :shock:

The one/two potential downsides of xG (and xA) is that every model is obviously a tad subjective (though the probabilities assigned to shooting positions would be based on historical shooting records from similar positions in idealistic models). Also somehow the stat might not adequately take into account how much better at finishing one player is to another, but this was always the case with SiB, SoT and TS anyway, same for the popular Big Chances (BC). There's also a suggestion that there's a slight issue in accuracy with not being able to easily take into account GK/defenders' positions, and debate on how to best incorporate to the binary SiB/BC stats into something fundamentally non-binary. For the real purist, there's a ridiculously good ongoing study/blog about comparing and maintaining xG models http://mackayanalytics.nl/football-analytics/ (thanks to Blue Fire for pointing this out a while back). Every model is slightly different, for example, the OPTA numbers and the understat ones don't match exactly - so, there's a warning there (i.e., none of the numbers should be taken as gospel, obviously, only as a rough guide). A second caveat is that, as far as I understand, professional statisticians don't actually use xG as a predictive tool, only as a very good way of 'judging what's happened' (they use much more advanced stuff for prediction) - but we're not professional statisticians, so extrapolate away. :lol:

Edit: Finally, I now realise it's probably a must to link an explanatory article (also discussing critisisms) by the guy widely attributed with coming up with xG as a concept, (Michael Caley), here, along with an explanation of his method here. Edit again: Turns out that may be a miscreditation. OPTA (and Caley himself), at least, credit Sam Green & Devin Pleuler for xG's inception. :)

Edit 2 (much later): And a nice other source of OPTA numbers is Paul Riley's here - thanks Blue Fire for pointing this out below (I now actually recall Mav may have mentioned this in the past too, so great work both). Edit again: it seems these are not OPTA's numbers, rather xG derived from some OPTA stats by Riley himself.

Enjoy. Please feel free to post any other resources for xG stats and or other articles and/or critisise away the whole concept and/or analyse some of the data. :mrgreen:

User avatar
Ruth_NZ
Grumpy Old Gorilla
Posts: 9156
Joined: 25 May 2015, 22:46

Re: One stat to rule them all (xG/xA)

Post by Ruth_NZ »

Thanks for putting this up. I have bookmarked understat. :)

I won't comment on xG/xA until I have used them more, which this post may cause me to do. But I would raise a general question on the use of stats, and that's "over what period?". I often see stats being used in a micro way, meaning over short time-frames like 4 weeks, when the macro picture can tell you as much if not more. I suspect that's an issue that won't go away with xG/xA either.

When circumstances have changed - for example a player in a new position or at a new club - then micro (short-term) stats are the best guide you can have apart from what you see watching games. But when nothing fundamental has changed then micro-stats can mislead badly. Take Kane as a case in point. I see lots of stats already being adduced on FFS about how things have changed for him this season, he's having more headers, his chances aren't as good or aren't "big" enough chances, his xG/xA is way behind Aguero and Lukaku and so on. But frankly that is an abuse of stats in my view. He's a proven player who most certainly isn't past his best (he should actually still be improving at his age) and is playing in the same position as ever in the same team. The only change of circumstance that might be relevant is Wembley.

If nothing material has changed then I'll take the macro view over the micro view in a case like that all day long. If Kane's form slumps alarmingly (meaning he is playing badly rather than his short-term stats are bad) then I'll react to that. But I don't need recent stats over a very short period to tell me how good a FPL asset he is. I think that would risk failing to see the wood for the trees.

User avatar
baganboy
Comfortably Dumb(ledore)
Posts: 5874
Joined: 05 Aug 2008, 06:59
FS Record: 2011/12 - 212. 2019/20 - 222.
Altogether 6 top 10Ks. 8 top 20Ks. 9 top 50Ks.

Re: One stat to rule them all (xG/xA)

Post by baganboy »

Oooh!

Cannot wait to get through the info.
Also, I know that Chris Glover had taken a hiatus, glad that he is back!

Thanks a lot for sharing, Stemania.

User avatar
blahblah
FISO Viscount
Posts: 108494
Joined: 13 Oct 2005, 18:46
Location: .. he thinks that he knows something which he doesn't, whereas I am quite concious of my ignorance.

Re: One stat to rule them all (xG/xA)

Post by blahblah »

re Kane: I think it is in his head, as per the "discussion" a season or so ago about him snatching at chances. Also to rebuff the stats: he doesn't need clear cut chances, which is one reason I erred against him in the first season as he gets a fair few from what I then considered pot-shots, but they kept going in... Once he calms his head down then he will fire for FPL purposes.

That said: Spurs at Wembley seems to support the Doom-mongers :wink:

Football Hero
Grumpy Old Man
Posts: 1256
Joined: 19 Aug 2014, 14:05

Re: One stat to rule them all (xG/xA)

Post by Football Hero »

If xG and xA trend towards actual goals and assists over a reasonable period of time, then why not. If they don't then these stats are fairly useless as they can't help you predict anything.

User avatar
Blue Fire
Grumpy Old Man
Posts: 1633
Joined: 05 Oct 2014, 14:19

Re: One stat to rule them all (xG/xA)

Post by Blue Fire »


User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

Blue Fire wrote: 22 Sep 2017, 06:52 https://public.tableau.com/profile/paul ... ard2017-18

PL shot map and player xG
Fantastic! So OPTA's numbers are covered then. :)

User avatar
Aldershot Rejects
Dumbledore
Posts: 9585
Joined: 03 Aug 2011, 17:15
Location: Kent
FS Record: 5th Metro (2010-11); 146 - Sky (2015-16); 218 - Sky (2014-15); 386 - Sky (2020-21); 636 - FPL (2017-18); last 16 Sky Cup (2018-19)

Re: One stat to rule them all (xG/xA)

Post by Aldershot Rejects »

Thanks so much for posting this Stem; I'm currently putting the finishing touches to my wildcard and a couple of the links have proved really useful.

User avatar
Joshhua
Red & Blue Braces
Posts: 481
Joined: 18 Sep 2017, 20:57

Re: One stat to rule them all (xG/xA)

Post by Joshhua »

Hi,
there is also another site with xG and xA, https://understat.com/league/EPL . Also easy for scrapping data if you are interested in that kind of things. :D

User avatar
Archy
Grumpy Old Man
Posts: 3210
Joined: 22 Oct 2011, 10:09
FS Record: Maybe all is not lost after all

Re: One stat to rule them all (xG/xA)

Post by Archy »

Interesting sites, thanks for posting guys :)

How do we make sense of it all though? On the site Blue Fire posts, it suggests Lukaku has been significantly overperforming, with xG of 3.0 compared to actual goals scored of 5. On the site Joshhua posts, it suggests Lukaku has been underperforming, with xG of 5.49 compared to his actual total of 5.

Similarly, Salah is overperforming on one (xG of 2.3, compared to actual 3) and underperforming on the other( xG of 3.38 compared to actual 3).

It's great having one stat to rule them all, but different versions of the same stat are ruling my confusion at the moment!

User avatar
Blue Fire
Grumpy Old Man
Posts: 1633
Joined: 05 Oct 2014, 14:19

Re: One stat to rule them all (xG/xA)

Post by Blue Fire »

Archy wrote: 22 Sep 2017, 13:55 Interesting sites, thanks for posting guys :)

How do we make sense of it all though? On the site Blue Fire posts, it suggests Lukaku has been significantly overperforming, with xG of 3.0 compared to actual goals scored of 5. On the site Joshhua posts, it suggests Lukaku has been underperforming, with xG of 5.49 compared to his actual total of 5.

Similarly, Salah is overperforming on one (xG of 2.3, compared to actual 3) and underperforming on the other( xG of 3.38 compared to actual 3).

It's great having one stat to rule them all, but different versions of the same stat are ruling my confusion at the moment!

The link I posted doesn't take into account xG from free kicks and penalties. That's probably the reason for the variation in Lukaku's xG. Not sure about Salah though

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

Joshhua wrote: 22 Sep 2017, 13:01 Hi,
there is also another site with xG and xA, https://understat.com/league/EPL . Also easy for scrapping data if you are interested in that kind of things. :D
Aye, this is the one Mav found. :D

As mentioned, their features are amazing, though their numbers (non-OPTA) seem significantly different to OPTA's in one or two places. Always good to have two different consistent sources though, with Paul Riley's tableau and Chris Glover's PLFantasy covering the latter. So...
Archy wrote:It's great having one stat to rule them all, but different versions of the same stat are ruling my confusion at the moment!
...yeah. One stat to rule them all, but loads of people have different ways of producing that one stat. The mackayanalytics link actually mentions 10 or so different xG creating methods/creators. It's quite a similar situation to the less advanced class of derived stats (PPM/VORP etc), where basically they're all the same idea calculated in slightly different ways.

User avatar
Tall Paul
Dumbledore
Posts: 7517
Joined: 27 Aug 2008, 12:57

Re: One stat to rule them all (xG/xA)

Post by Tall Paul »

Archy wrote: 22 Sep 2017, 13:55 Interesting sites, thanks for posting guys :)

How do we make sense of it all though? On the site Blue Fire posts, it suggests Lukaku has been significantly overperforming, with xG of 3.0 compared to actual goals scored of 5. On the site Joshhua posts, it suggests Lukaku has been underperforming, with xG of 5.49 compared to his actual total of 5.

Similarly, Salah is overperforming on one (xG of 2.3, compared to actual 3) and underperforming on the other( xG of 3.38 compared to actual 3).

It's great having one stat to rule them all, but different versions of the same stat are ruling my confusion at the moment!
As long as you use the same method to compare your players/teams, does it really matter that there are different versions?

User avatar
Ruth_NZ
Grumpy Old Gorilla
Posts: 9156
Joined: 25 May 2015, 22:46

Re: One stat to rule them all (xG/xA)

Post by Ruth_NZ »

Well, it's kind of disconcerting when the xG factors for key players are widely different. :?

User avatar
MoSe
Dumbledore
Posts: 9550
Joined: 10 Sep 2014, 12:25
Location: next door S.Siro stadium
FS Record: FISODAS CUP Winner Season 25
FISO H2H Winner: 15/16 Div2 - 16/17 Div1
FISO Mirror: 16/17 PL Winner

Re: One stat to rule them all (xG/xA)

Post by MoSe »

so now we're torn between

G = what happened
xG = what didn't happen (but should have)

User avatar
MoSe
Dumbledore
Posts: 9550
Joined: 10 Sep 2014, 12:25
Location: next door S.Siro stadium
FS Record: FISODAS CUP Winner Season 25
FISO H2H Winner: 15/16 Div2 - 16/17 Div1
FISO Mirror: 16/17 PL Winner

Re: One stat to rule them all (xG/xA)

Post by MoSe »

Tall Paul wrote: 22 Sep 2017, 15:39 As long as you use the same method to compare your players/teams, does it really matter that there are different versions?
without having investigated the details

it depends HOW the sources differ

source X: A=1 B=2 C=3 D=4
source Y: A=2 B=4 C=6 D=8
OK

source X: A=1 B=2 C=3 D=4
source Y: A=1 B=4 C=5 D=8
hhmmm....

source X: A=1 B=2 C=3 D=4
source Y: A=1 C=2 B=3 D=4
Ykes! :o

obviously, if you rely on indicators to even partially help you take your decisions
you understand that in 3rd case using source Y rather than source X would lead you to different decisions, even with the same method...

in simpler words, using Archy's example

if Lukaku and Salah both similarly overperform in "BF source" (3.0, 2.3) and both similarly underperform in "J source" (5.49, 3.38), then as you say it's just a matter of coefficient and scale, but the relative strength and/or expectations are the same.
should the values be (3.0 & 2.9) and (6.0 & 3.1), i.e. in the same order but very close on one and very far on the other, some skew is introduced
Should finally the two sites invert the reuslts, i.e. on one Lukaku is over and Salah is under, while on the other source Luk is under and Sal is over, then houston we have a problem

User avatar
Turd Ferguson
FISOhead
Posts: 943
Joined: 16 Aug 2015, 03:32

Re: One stat to rule them all (xG/xA)

Post by Turd Ferguson »

Blue Fire wrote: 22 Sep 2017, 14:46
Archy wrote: 22 Sep 2017, 13:55 Interesting sites, thanks for posting guys :)

How do we make sense of it all though? On the site Blue Fire posts, it suggests Lukaku has been significantly overperforming, with xG of 3.0 compared to actual goals scored of 5. On the site Joshhua posts, it suggests Lukaku has been underperforming, with xG of 5.49 compared to his actual total of 5.

Similarly, Salah is overperforming on one (xG of 2.3, compared to actual 3) and underperforming on the other( xG of 3.38 compared to actual 3).

It's great having one stat to rule them all, but different versions of the same stat are ruling my confusion at the moment!

The link I posted doesn't take into account xG from free kicks and penalties. That's probably the reason for the variation in Lukaku's xG. Not sure about Salah though
If it doesn't take into account set pieces then it's tough to use it for FPL purposes.

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

You can actually filter all the data to exclude penalties by selecting NPxG on understat (and in fact filter by every type of set piece on a given individual's profile).

Both top 10-15 or so are for the most part the same players with a wiggle here and there so it's really not a huge thing. With no pens (underS vs OPTA):

1 Lukaku vs Kane
2 Aguero vs Lukaku
3 Jesus vs Jesus
4 Kane vs Aguero

5 Salah vs Alli
6 Welbeck vs Salah
7 Joselu vs Joselu
8 Morata vs Welbeck

9 Pogba vs ECM
10 Alli vs Pogba
11 Mane vs Mane
12 ECM vs Lacazette

Morata and especially Lukaku seem to really miss out in OPTA relative to underS, Kane and Alli lose out on underS relative to OPTA, all for unknown reasons. (Though the comparison isn't perfect due to direct freekicks being included by underS). The difference with Lukaku is quite odd really as most of the other differences look pretty minor in size.

User avatar
Turd Ferguson
FISOhead
Posts: 943
Joined: 16 Aug 2015, 03:32

Re: One stat to rule them all (xG/xA)

Post by Turd Ferguson »

Stemania wrote: 22 Sep 2017, 18:15
7 Joselu vs Joselu
Wow, am I reading this wrong or doesn't this imply we should all be looking very hard at 5.5 Joselu for our teams? I haven't watched Newcastle live so I have no idea how threatening he's been.

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

MoSe wrote: 22 Sep 2017, 17:20 so now we're torn between

G = what happened
xG = what didn't happen (but should have)
Ha ha, indeed. Clearly the second is just a refinement of the first and I suppose the important thing to point out is that none of the numbers should be directly extrapolated to produce points projections for the future. All the analytics are still at their heart a description of what has happened so far (even if only theoretically) and over the short term depend just as much on nice previous fixture runs/purple patches of form as any other. So in that vein...
Turd Ferguson wrote:
Stemania wrote: 22 Sep 2017, 18:15
7 Joselu vs Joselu
Wow, am I reading this wrong or doesn't this imply we should all be looking very hard at 5.5 Joselu for our teams? I haven't watched Newcastle live so I have no idea how threatening he's been.
...yeah, his numbers are mighty impressive so far! But it's all such a small sample that I think all it's really telling us after 5 weeks is that he's had many good chances so far (he missed at least three of them last game alone) and is one to watch. The question of whether he will keep it up is the important one - if he's still got those kind of numbers in 5 or so more weeks then certainly the evidence will be building - that's if he keeps his place over Gayle. :wink:

Edit: That's in a nutshell why xG is better than G I suppose, because xG isn't biased by things like whether in reality a shot hits the post and goes in or hits the post and stays out (which is 'basically' luck), though you could (and should) argue that a sustained disparity in xG vs G is evidence of whether a player is an excellent or terrible finisher.

casualnutpi
Cheetah
Posts: 3
Joined: 22 Sep 2017, 21:06

Re: One stat to rule them all (xG/xA)

Post by casualnutpi »

Its probably more the method rather than OPTA vs non-OPTA. Rileys method is quite basic. My own OPTA based non penalty expG for Lukaky is 4.4 for example

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

casualnutpi wrote: 22 Sep 2017, 21:11 Its probably more the method rather than OPTA vs non-OPTA. Rileys method is quite basic. My own OPTA based non penalty expG for Lukaky is 4.4 for example
Welcome to the forum!!

Ah, are Riley's numbers not the official OPTA xG numbers (OPTA actually produce an xG stat themselves now)? My impression from his tableau was that they were official rather than calculated himself, but happy to be corrected. It would indeed explain a lot if not, though being OPTA official ones wouldn't necessarily be a sign of quality anyway. :D Edit: see post below.

Your 4.4 would be roughly in like with understat have (4.7). Are you in a position to share your numbers generally, or alternatively (as someone who clearly is experienced in this) do you have an informed opinion on which, if any, open source is best?


(BTW I realise now I may have miscredited the initial idea - OPTA themselves credit an article by Sam Green from 2012: http://www.optasportspro.com/about/opta ... alscorers/.)

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

Actually, from what I've just read (having looked around the web), casualnutpi seems indeed spot on that Riley's xG are his own derived from OPTA's stats, rather than OPTA's numbers themselves. Further, they don't seem to be considered so strong in the xG community. In fact, they are one of the sets in the mackayanalytics blog that performs the worst and is actually described there as simplistic (it's this one https://differentgame.wordpress.com/201 ... nutes-ish/ of Paul Riley).

Happily, whilst most models have used logistic regression one of the best performing ones in the opinion of mackayanalytics was that of Trainor & Chappas over at statsbomb.com. I say, happily because that one is unique in the list as it's the only one to use what's is classed as the 'neural network approach' rather than logistic regression. And that's also the method understat use (as per the understat homepage). Given that the statsbomb's model/numbers are not public that's potentially a big plus in understat's column.

In lieu of being able to see any of the methodologies in the above, I read it's the same method (or category of method) as those at teouchanalytics use for their models (which I don't think are available to see) - I didn't read through their general methodology thoroughly as it's...quite heavy and not xG specific.

And btw, I haven't delved too far here either, but there might be some hidden gems in deepxg.com, which I came across - e.g. a cheeky go at expected saves a couple of years back. :D

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

Looks like FFS are also now adding (OPTA-based) xG and xA to their data:

http://www.fantasyfootballscout.co.uk/2 ... ted-goals/

Will be interesting to see how their numbers compare to understat and Riley etc.

User avatar
gallus
Grumpy Old Man
Posts: 3921
Joined: 06 Sep 2014, 11:55

Re: One stat to rule them all (xG/xA)

Post by gallus »

Turd Ferguson wrote: 22 Sep 2017, 18:24
Stemania wrote: 22 Sep 2017, 18:15
7 Joselu vs Joselu
Wow, am I reading this wrong or doesn't this imply we should all be looking very hard at 5.5 Joselu for our teams? I haven't watched Newcastle live so I have no idea how threatening he's been.
No, but we all should have Choupo Moting. He's rated 8th among all players and he's a 5.5 mid! :shock: There are 5 better strikers than Joselu, but only 2 mids are better than Moting.

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

Both of them have actually dropped off a bit in recent weeks though, and neither of them look like very good finishers; Josulu in particular looks like he might need multiple times the xG of anyone else just to break even, and after last weekend his best bet may well be not to shoot at all. :lol:

Still very early in the season, and too early to say anything about the cheapies with much certainty imo. :D

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

So I had a quick play around with FFS' numbers vs understat's and (on xG at least) they reassuringly seem to generally agree
- maybe surprising considering that they use as far as we understand very different methods of calculation? Good news though.

Here's a (excel-sorry!) graph of the top 35 xG performers so far on FFS vs the numbers for the same players on understat (with the number of actual goals scored in grey) - both are remarkably similar, and generally regard the same players to have 'over-' or 'under-' performed:

Image


And also an interesting expected assists version, where there seems to be far more noise/disagreement between the two methodologies.

Image

Notably, it seems on initial glance that expected assist numbers correlate nowhere near as well to actual assists - but I guess we should remember that the definition of an 'assist' in FPL is generally slightly different to that used by data gathering people (e.g. OPTA). It could also be down to the possibility that the best passers of the ball are a significantly better passer (than the average passer) compared to how much better the best finisher is (than the average finisher) - i.e., perhaps being world class at creating chances is a far more impressive/difficult skill than just being a world class finisher of chances. Or simply that chances created are far harder to quantify than shots.

(I didn't include Riley's numbers as his are no pen/no FKs and are widely thought to be relatively simplistic it seems - no offense intended to him - plus he doesn't have xA numbers).

Edit: For clarity, I ordered the players in the graphs by actual goals scored/actual assists made, not by their xG/xA, then by the xG/xA values on FFS (which is why the xA line for US looks 'spikier' than the line for FFS for players with the same number of assists, for example.).

Any thoughts? How have people found using the understat (or FFS) numbers so far?

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

Just in case anyone wants to look up who's who:

xG:(1Kane 2Aguero 3Lukaku 4Salah 5Jesus 6Morata 7Sterling 8Vardy 9Lacazette 10Sane 11Richarlison 12Abraham 13Okazaki 14Murray 15Chicharito 16Niasse 17Tadic 18Martial 19Ramsey 20Alli 21Firmino 22Welbeck 23Rashford 24Fellaini 25Joselu 26ECM 27Rodriguez 28Mahrez 29Sturridge 30Mane 31Silva 32Rooney 33DLC 34Ayew 35Alexis)

xA:(1Silva 2De Bruyne 3Sane 4Mkhitaryan 5Groß 6Azpilicueta 7Mahrez 8Shaqiri 9Ritchie 10Walker 11Aguero 12ECM 13Holebas 14Kolasinac 15Rashford 16Eriksen 17Sanchez 18Ozil 19Ibe 20Redmond 21Sterling 22Firmino 23Davies 24Knockaert 25Fabregas 26Brady 27Alli 28Baines 29Rooney 30Tadic 31Townsend 32Cabaye 33Mata 34Sigurdsson 35Sissoko)

User avatar
Pirlo's Beard
FISO Jedi Knight
Posts: 20554
Joined: 21 Aug 2013, 17:48

Re: One stat to rule them all (xG/xA)

Post by Pirlo's Beard »

Stemania wrote: 11 Nov 2017, 13:13 33DLC

Dominic Loftus-Cheek? :D

User avatar
Stemania
FISO Jedi Knight
Posts: 20448
Joined: 27 Aug 2006, 11:54
Location: On the Iron Throne of xG, the seat of The Crown Prince of the Stat Perverts. Or if not, in the STC!
FS Record: Best: TFF 321st. FPL 129th. FFS Career HoF Peak 2nd (Live 1st). Ability since lost.

Re: One stat to rule them all (xG/xA)

Post by Stemania »

Oops, DCL. :mrgreen: :lol:

View Latest: 1 Day View Your posts
Post Reply

Return to “Fantasy PremierLeague.com (FPL)”