xG

[StanSP]

23 minutes ago, Col city fan said:

What is it Stan?

Expected Goals. 

 

 

23 minutes ago, RumbleFox said:

Probably just ignore it then. 

It's hard to when it's mentioned a lot in several threads on here! 

[daddylonglegs]

xG is a real nuisance if you’re a Brendan fanboy. 
 

Fairly useful if you’re looking to see how many chances a team create/concede versus how many they score/concede. 
 

By no means the be all and end all though.

[ALC Fox]

If you're not a fan of xG, just wait until you find about xA and xT 

[Strokes]

2 hours ago, Col city fan said:

What is it Stan?

A noncey stat.

[HighPeakFox]

This forum would do well on xE for overuse of the word EMBARRASSING  

[Chelmofox]

1 minute ago, HighPeakFox said:

This forum would do well on xE for overuse of the word EMBARRASSING  

Or 'bottle job'.

[RumbleFox]

3 hours ago, StanSP said:

Expected Goals. 

 

 

It's hard to when it's mentioned a lot in several threads on here! 

I’d say that other than match of the day having it as one of many stats they show at the bottom of the screen and a few posts on here about it, it hardly feels all over the place. Doesn’t bother me at all, I’m fact I rarely notice any mention of it.  X 

Edited 22 April 2022 by RumbleFox

[RumbleFox]

I think it’s probably confirmation bias, on both sides. xG doesn’t bother me so when I see someone talking about it I just think “oh, that’s mildly interesting, we outperformed our xG, I wonder if we can keep that up” where as folk that don’t like them probably see “omg xG is 5 but we only scored 2 so in reality we should have won”. 

[Nalis]

3 hours ago, ALC Fox said:

If you're not a fan of xG, just wait until you find about xA and xT 

I can see the logic in xA and xT as tools to form part of the wider range of stats available.

 

I would tell you about xX but it needs no Intro

[majaco]

4 hours ago, daddylonglegs said:

xG is a real nuisance if you’re a Brendan fanboy. 
 

Fairly useful if you’re looking to see how many chances a team create/concede versus how many they score/concede. 
 

By no means the be all and end all though.

You make it sound problematic if people back the manager.

[Dan]

7 hours ago, jamesp26 said:

In reality you highlight why the statistic isn't the be all and end all. Because of 5 non defensive players with quality, we are doing better than a statistic suggests we might do. Compare that to teams who don't have the same quality, and they do not have as many points as us. Losing those quality players suggests we might be worse off - nobody saw that coming!  If you take Brentford for example, chances get generated by long ball play, but they lack quality, excluding Eriksen of course, but he's not played for all of the season. Interesting to see how some quality  for Brentford had expecting impact.

 

Therefore, there is no place we 'deserve' to be beyond where we are.  The job of the performance metric is to highlight the importance of the quality, area's we could improve on and what might happen if we could improve those areas. However, just doing what Brentford do just to improve xG isn't the way to go about it. 

 

In reality, i wish they called the metric something else, like 'Expect Impact' or something vague. Expected goals sets the wrong message.

You are correct and the quality of some of our attacking players is likely going to lead to us overperforming against the actual flow of matches. I just don't think it's particularly sustainable to keep scoring wondergoals. It's not solely attacking quality either, it's the chances missed against us. Richarlison the other night a perfect example.

 

It's a guide as much as anything. I see xG as adding context to shots and shots on target.

 

To put it simply what I'm getting at is I think the manager is fortunate we're as high as we are.

[Fox92]

On 21/04/2022 at 12:22, Finnegan said:

 

No more impossible to ignore than some thick fvck screaming that Daka is a pub footballer and needs to be sent back on every fvcking page of the match thread. 

Well I've actually never said any of those exact words, so that can't be me you're aiming that at  

[NewEnglandFox]

 

[coolhandfox]

58 minutes ago, NewEnglandFox said:

 

Like two drunk tramps fighting over a chip.

[Number 6]

It's a really useful metric even if it is still a bit inchoate at the moment. Think the models could be improved fairly easily so I imagine they'll get better, though as has been highlighted it can't be viewed in isolation.

 

James Tippett who runs the Twitter posted above worked for Benham and knows his stuff. But like Benham and Franck has drank the kool aid on it a bit and seems to think it's a silver bullet. Also given how quickly he posts his xG results after a game I presume his model is based entirely on shot location and nothing else, so very rudimentary.

[Dave Fishwick]

bloody dinosaurs

[Dahnsouff]

5 hours ago, Dave Fishwick said:

bloody dinosaurs

Agree, they definitely should have paid more attention to the xE (expected extinction) figures, and not dismissed them out of hand

[leicsmac]

3 minutes ago, Dahnsouff said:

Agree, they definitely should have paid more attention to the xE (expected extinction) figures, and not dismissed them out of hand

Asteroid (10^9) 10^10 - 0 (0.1) Dinosaurs

[Dahnsouff]

4 minutes ago, leicsmac said:

Asteroid (10^9) 10^10 - 0 (0.1) Dinosaurs

[Zear0]

Liverpool xG 2.01. Let's all bow down to our xG overlord. 

[Ric Flair]

Have we ever had an xG above 2? Probably about 1.74 even when we won 9-0 at Southampton

[Aus Fox]

21 minutes ago, Ric Flair said:

Have we ever had an xG above 2? Probably about 1.74 even when we won 9-0 at Southampton

 

[Aus Fox]

I’d say with Vardy in our team Xg, becomes even more meaningless, I would imagine he’s goal scoring rate throughout his career would be almost double his Xg - I can’t back this up with stats, be great if someone has the information and can either confirm or deny.

Iheancho I would imagine to be right on his Xg and Daka slightly behind.

Therefore a game with us having an Xg around 2, with Vardy could be a 3 or 4 goal win, with Iheanacho 2 goals and with Daka maybe one goal. 
Maddison and Tielemans are both quality from distance, I dont think most of their goals would register as an expected goal,however, because of their quality it’s a genuine chance for us if either of these get the ball in space 30 yards or closer to goal. 

[blaaklint]

On 24/04/2022 at 23:39, Aus Fox said:

I’d say with Vardy in our team Xg, becomes even more meaningless, I would imagine he’s goal scoring rate throughout his career would be almost double his Xg - I can’t back this up with stats, be great if someone has the information and can either confirm or deny.

Iheancho I would imagine to be right on his Xg and Daka slightly behind.

Therefore a game with us having an Xg around 2, with Vardy could be a 3 or 4 goal win, with Iheanacho 2 goals and with Daka maybe one goal. 
Maddison and Tielemans are both quality from distance, I dont think most of their goals would register as an expected goal,however, because of their quality it’s a genuine chance for us if either of these get the ball in space 30 yards or closer to goal. 

tl;dr: Maddison and Iheanacho are our best finishers, Barnes is worst, nobody in the world is good enough for 2x xG.

 

Fbref has Vardy's finishing over the last five seasons as averaging 0.02 non-penalty goals more than non-penalty xG per 90 (np:G-xG per 90). This equates to Vardy scoring an extra from the same chances as the average player every 50 games. Iheanacho over the same time frame is 0.06 np:G-xG per 90, or an extra every 17 games.

 

According to the Statsbomb data fbref uses at least, Vardy's finishing is probably above average, and may be a slight underestimation if we assume that goalkeepers in the PL are better on average than goalkeepers over the whole dataset the xG model was trained on.

 

If you exclude last year, which appears to be a complete outlier compared to other seasons, Vardy has a much more respectable np:G-xG rate of 0.07, or an extra goal every 14 games. This probably greater reflects my own estimation of his finishing ability from watching him play.

 

Either way, what is more notable is that Vardy only shoots when the chance is of high quality (0.19 xG/shot over the last five years) . As a result, he scores a lot from not many chances, making his finishing appear exceptional at first glance (this is the same issue with using chance quality-blind measures such as G/shot).

 

Vardy is still an excellent player, it's just that his goalscoring comes more from his ability to be in the right place at the right time (i.e., movement, anticpation, pace, etc.) than from his finishing.

 

Maddison's np:G-xG per 90 is 0.06 and Tielemans' is 0.03, indicating that they are both also above average finishers. The same measure finds that, of the rest of our more attacking players, Perez is above average (0.04), Lookman and Albrighton are average (0.01) and Barnes is by far the worst (-0.05).

 

A starting lineup of Vardy, Iheanacho, Barnes, Lookman, Maddison and Tielemans scores just under five more goals than can be expected of average finishers over a 38 game season.

 

For comparison, the highest average np:G-xG per 90 I could find is Haaland (0.22; 8.36 goals per 38 game season). The highest single season value is Messi (0.42; 15.96).

 

All of this should be taken with a pinch of salt: the volume of data for most players is fairly small, so inferences regarding finishing quality may be just natural variance. There may also be other factors which effect the quality of chance which aren't included in Statsbomb's xG model.