Rovers…Are We Any Good ?

[RoversTilliDie]

On 23/10/2022 at 09:12, M_B said:

It's a good question, I keep saying I'm waiting for us to click and play well, even over a full 90 minutes. 

Eventually either the results will suffer, you can't keep hanging on in games and winning,or the results will bring with them more quality to the performance. 

At the moment, what we are good at is getting the 3 points when we go in front. Long may it continue, it's a lot easier for Tomasson to do whatever it is he wants to do nearer the top than the bottom.

Being booed off at half time against Millwall, yet finishing the day in the top 6 is evidence enough. 

The way Rovers are playing at the moment, I will be disappointed if we don't make the top 6 minimum.

[JacknOry]

On 23/10/2022 at 22:56, Tyrone Shoelaces said:

Maybe JDT has given Gally some one to one coaching. After all he’ll know a lot more about playing up front than Mowbray or Venus.

Thinks its more simple than that, he is playing more centrally - or at least having licence to go there. 

[JacknOry]

11 hours ago, Displaced Rover said:

I know a lot of people don't love xG but I was taking a look at Championship stats through the season so far.

Much like last year, our actual for and against goals are performing far beyond expected. Like last year are the wheels going to come off spectacularly?

23rd for xG (0.96)

8th (highest) for xGA (1.5)

If we're anywhere near the mix for top 6 come April/May I think we will have had a fantastic season.

Yeah but this xG malarky swings in all directions. i read somewhere that when we finished 15th, we had the second highest xG in the league. 🤷‍♀️

[J*B]

15 hours ago, Herbie6590 said:

xG can only be calculated after the game so not sure how that entices betting TBH 🤔

Surely xG can be calculated mid-game? They tell you what xG each specific chance has against it real time. Eg that was a 0.38 xG opportunity.

So whilst the final xG can only be calculated at the end of the game, the current xG can be averaged out at any point up until that moment? 

[smiller14]

xG is so misunderstood - I don't swear by it, but to suggest it's nonsense is just daft. How many times have we said 'we had 20 shots today and didn't score' or 'their keeper made 10 saves'. This is just an extension of that. It's obviously flawed but so are any other statistics in football. I appreciate it doesn't factor in the 'well if we didn't score then we would have been attacking more and creating more chances' but you can apply that to corners, shots, possession too.

One thing that rang true last year is xG was highly predictive of performance and a lot of teams who were performing well/creating chances earlier in the season but perhaps underperforming on points often ended up doing alright. Celtic were 5 or 6 behind but my xG loving (too much in his case) mate swore they'd end up as champions because their performances were statistically much better than Rangers. 

Obviously it doesn't factor in individual players. Neither does shots, or shots on target. Clearly Diaz is more likely to score a similar chance than Chris Brown. But better players tend to outperform their xG. We saw Kaminski make saves that Walton wouldn't have made - otherwise we'd have lost or drawn on the weekend. If anything, how much a player outperforms their xG is a great sign of how talented they are or how clinical they are.

I'm by no means an xG fanboy, and choose instead to see it as a small bit of context alongside all the other statistics and events that occur in a game. But I must disagree that it's pointless - it's usually pretty accurate (if you have a big enough sample size of data, game by game it will fluctuate). 

Seen a lot of posts decrying statistics recently as pointless - 'only one thing matters' - clearly the score is the most important, but I really really hope and expect JDT and our staff are much more receptive to statistics than some supporters. And I agree there's no substitute for watching the game with your own eyes and drawing conclusions, but anyone watching on the weekend would have said Birmingham were unlucky - a conclusion supported by statistics.

Back on point, we are good. We have a great keeper, solid defence, a midfield who are developing and the best striker in the league. Are we top 2 worthy? Who knows, but I'm going to really enjoy finding out. 

[J*B]

I think I’m right in saying - although would need @JoeH to confirm - that xG doesn’t/can’t take into account the player, just the environment. So Brereton-Diaz being 18 yards out one on one with the keeper has the same xG as Hirst 18 yards out one on one with the keeper. 

Is that correct?

[Guest]

7 minutes ago, J*B said:

I think I’m right in saying - although would need @JoeH to confirm - that xG doesn’t/can’t take into account the player, just the environment. So Brereton-Diaz being 18 yards out one on one with the keeper has the same xG as Hirst 18 yards out one on one with the keeper. 

Is that correct?

Yeah, xG doesn't account for the quality of the striker. xG trends can indicate the quality of a striker, but the actual xG model doesn't fluctuate and account for the perceived quality of the striker.

If you are consistently outscoring xG, it's down to having above average forwards or potentially having luck. If you are conceding less than xGA puts you at, your keeper is likely above league average.

I think BBD and Tommy K are both well above average for their positions and it reflects in our poor xG/xGA but strong actual goals/goals against.

[Guest]

22 hours ago, roversfan99 said:

You can't objectify the quality of a chance

That's your opinion, but the reality is that industry experts, all of elite football and the mainstream media agree that you can objectify the quality of a chance. That's what xG models are, and they are constantly improving. 9 times out of 10, xG trends are proved correct over a season, it's more accurate than any other mainstream data, like Shots on Target, Shots & Possession that you'll find on Sky & BBC's sites.

[Mashed Potatoes]

30 minutes ago, J*B said:

I think I’m right in saying - although would need @JoeH to confirm - that xG doesn’t/can’t take into account the player, just the environment. So Brereton-Diaz being 18 yards out one on one with the keeper has the same xG as Hirst 18 yards out one on one with the keeper. 

Is that correct?

Yes.xG is the same for the environment for every player. Thus a penalty has an xG of 0.74 goals, regardless of whether it is Ivan Toney or Diana Ross taking it.

[Mattyblue]

 Certainly keeps a load of folk in work…

[Nuttall is lost]

22 minutes ago, Mashed Potatoes said:

Yes.xG is the same for the environment for every player. Thus a penalty has an xG of 0.74 goals, regardless of whether it is Ivan Toney or Diana Ross taking it.

She hit the ball so hard the crossbar broke. 

[roversfan99]

47 minutes ago, JoeH said:

Yeah, xG doesn't account for the quality of the striker. xG trends can indicate the quality of a striker, but the actual xG model doesn't fluctuate and account for the perceived quality of the striker.

If you are consistently outscoring xG, it's down to having above average forwards or potentially having luck. If you are conceding less than xGA puts you at, your keeper is likely above league average.

I think BBD and Tommy K are both well above average for their positions and it reflects in our poor xG/xGA but strong actual goals/goals against.

 

38 minutes ago, JoeH said:

That's your opinion, but the reality is that industry experts, all of elite football and the mainstream media agree that you can objectify the quality of a chance. That's what xG models are, and they are constantly improving. 9 times out of 10, xG trends are proved correct over a season, it's more accurate than any other mainstream data, like Shots on Target, Shots & Possession that you'll find on Sky & BBC's sites.

I have loads of reservations about it for many reasons, but one thing I am sure of is it is that its general usage seems totally not fit for purpose. In the media especially on podcasts, in the absence of anyone having the time to watch every team in the Championship, xG figures are used as a direct barometer of performance, which is a load of rubbish. This idea that teams will revert roughly to their current xG levels is a load of nonsense by people trying to sound well educated about every team. Over the last 18 months, we have scored well in advance of our xG mainly because of Brereton, he will score more goals with the same chances compared to basically any other forward in the league. It also doesn't factor in teams taking a lead and how a game totally changes from there on in, we have taken 10 leads, so naturally 10/17 times, our opponent has then had an onus to get back into the game,

One good use that surely could come from it is judging the performance of individual strikers, its not even about luck IMO. 

[Herbie6590]

2 hours ago, J*B said:

Surely xG can be calculated mid-game? They tell you what xG each specific chance has against it real time. Eg that was a 0.38 xG opportunity.

So whilst the final xG can only be calculated at the end of the game, the current xG can be averaged out at any point up until that moment? 

Well…yes…which was kind of my point, albeit seemingly ambiguously made…

[Herbie6590]

31 minutes ago, roversfan99 said:

I have loads of reservations about it for many reasons, but one thing I am sure of is it is that its general usage seems totally not fit for purpose. In the media especially on podcasts, in the absence of anyone having the time to watch every team in the Championship, xG figures are used as a direct barometer of performance

It clearly is fit for purpose as it has been adopted by the professionals in the industry who see value in it.

As for using it on podcasts - it does provide a proxy for “balance of play”, which previously was the domain of stats such as %possession, % territory, corners, shots on target & so on. It is but one metric that indicates performance.

In business, the headline stat is Sales…but there are loads of others that the sales director, the finance director and the CEO will be looking at to inform future decision-making. Football uses xG similarly.

[roversfan99]

27 minutes ago, Herbie6590 said:

It clearly is fit for purpose as it has been adopted by the professionals in the industry who see value in it.

As for using it on podcasts - it does provide a proxy for “balance of play”, which previously was the domain of stats such as %possession, % territory, corners, shots on target & so on. It is but one metric that indicates performance.

In business, the headline stat is Sales…but there are loads of others that the sales director, the finance director and the CEO will be looking at to inform future decision-making. Football uses xG similarly.

To be fair, Steve Kean was made a manager, Owen Coyle has had loads of jobs, we at one stage had Chris Brown and Luke Varney in the same squad, just because professionals in the industry see value in it doesnt necessarily make it fit for purpose!

Regarding the point about podcasts or the media in general, I still think that it is often used poorly. You would never really have heard previously that a team could not sustainably perform beyond its possession stats or beyond the level of shots on targets. I feel that xG is regularly used as a massively flawed and inferior substitute to watching games to try and seem really informed in what they say about the whole league.

I dont doubt that data has its uses but the key to all data is the person using it and how it is used. You could easily look at our xG being low without watching our games and presume that we are lucky, and whilst I dont think that our performance level has been particularly great, I having watched our games would think of 3 things straight away that contribute to that. We have a striker as ruthless as any in the league, that isnt a temporary thing, he will continue to outperform his xG if fit. We have one of the better goalkeepers in the league too, same logic. And we have gone ahead in 10/17 games, so nearly 60%, I think in 9 we scored in the first half and quite a few particularly early, so we have been content to sit in.

[superniko]

2 hours ago, JoeH said:

Yeah, xG doesn't account for the quality of the striker. xG trends can indicate the quality of a striker, but the actual xG model doesn't fluctuate and account for the perceived quality of the striker.

If you are consistently outscoring xG, it's down to having above average forwards or potentially having luck. If you are conceding less than xGA puts you at, your keeper is likely above league average.

I think BBD and Tommy K are both well above average for their positions and it reflects in our poor xG/xGA but strong actual goals/goals against.

Whilst this is true, luck is a larger factor, and we will eventually regress to the mean….The season only being 46 games means it’s a relative small sample statistically speaking and that regression may happen too late as we march to promotion. 

[smiller14]

I think there are flaws, particularly where you score first and 'sit in'. Although clubs utilising it would take this into account and wouldn't just lazily parrot the 'expected score' without any context like you see some pundits do. I don't think anyone should rely on it fully, it's just another bit of context to add to all the other analysis that every top team uses nowadays. 

I don't see our over-performance as luck, I see it in the same way you do - we have a great striker and a great keeper. The best teams overperform their xG because they have players who will do better than expected.

 

[superniko]

13 minutes ago, roversfan99 said:

To be fair, Steve Kean was made a manager, Owen Coyle has had loads of jobs, we at one stage had Chris Brown and Luke Varney in the same squad, just because professionals in the industry see value in it doesnt necessarily make it fit for purpose!

 

It’s used widely in the bookmaking industry and I’m yet to see a poor one. 

Look at Bloom and Bentham - some of the first to use it for their personal gain. Now owning Brighton and Brentford respectively and they’ve both seen some uptick in their fortunes.

It has flaws yes, but it’s the best stat for pricing odds on matches and at simulating seasons to predict finishing positions. 

[roversfan99]

38 minutes ago, superniko said:

Whilst this is true, luck is a larger factor, and we will eventually regress to the mean….The season only being 46 games means it’s a relative small sample statistically speaking and that regression may happen too late as we march to promotion. 

Why would we regress to the mean in relation to the quote you highlighted? Kaminski and in particular Brereton will surely as quality operators in their positions will over-perform expected goals/goals against.

[Guest]

2 hours ago, roversfan99 said:

one thing I am sure of is it is that its general usage seems totally not fit for purpose

Then why does every professional club in the country take it so seriously? Why do the industry leading data analytics companies bury so much cash into the R&D of xG models? Why do clubs even down in League Two and below use xG for both performance and recruitment analysis?

48 minutes ago, smiller14 said:

Although clubs utilising it would take this into account and wouldn't just lazily parrot the 'expected score' without any context like you see some pundits do. I don't think anyone should rely on it fully, it's just another bit of context to add to all the other analysis that every top team uses nowadays.

Yep very true, xG is never looked at alone and game state is massively important. That's why those xG "race" charts are so relevant. Like this one for an example:

1 hour ago, roversfan99 said:

just because professionals in the industry see value in it doesnt necessarily make it fit for purpose!

That would be a very fair comment if it was a bit like Big Sam and his headset back when he was one of the only people doing it. This isn't just a few oddball managers, it's 99% of clubs across the professional game, down to the National League.

 

 

No metrics has ever more accurately predicted outcome, predicted trend and predicted success/failure than the modern xG models we have today. That's not an opinion it's a proven statistical fact.

 

[Guest]

2 hours ago, Mattyblue said:

 Certainly keeps a load of folk in work…

And if this is in any way a reference to me, which it may not be (and my bad if it's not) I'll just explain that I'm leading the Live Scouting at Barrow AFC, with quite a limited influence in our usage of data.

[superniko]

7 minutes ago, roversfan99 said:

Why would we regress to the mean in relation to the quote you highlighted? Kaminski and in particular Brereton will surely as quality operators in their positions will over-perform expected goals/goals against.

Because whilst Kaminski and Brereton are better than average and a factor in our over-performance in actual vs expected I stated that I believe luck or statistic size (as a broad term but perhaps could be worded better) is a bigger factor. Those two players alone aren't enough to keep outperforming xG as much as we have this season.

Personally, the flaw I don't like with it is how it doesn't represent game scenarios i.e. Sunderland only started to lead in xG once we were 2-0 up and we sat back and they dominated - would this have happened if Breretons screamer went over the bar? Unlikely to the extent it did. We take the lead, we sit back and we lose on xG - if it keeps getting us 3 points then happy days.

To be blunt though, there is a reason we're 8th favourites for promotion despite being joint top of the league.

[roversfan99]

4 minutes ago, JoeH said:

Then why does every professional club in the country take it so seriously? Why do the industry leading data analytics companies bury so much cash into the R&D of xG models? Why do clubs even down in League Two and below use xG for both performance and recruitment analysis?

Yep very true, xG is never looked at alone and game state is massively important. That's why those xG "race" charts are so relevant. Like this one for an example:

That would be a very fair comment if it was a bit like Big Sam and his headset back when he was one of the only people doing it. This isn't just a few oddball managers, it's 99% of clubs across the professional game, down to the National League.

 

 

No metrics has ever more accurately predicted outcome, predicted trend and predicted success/failure than the modern xG models we have today. That's not an opinion it's a proven statistical fact.

 

That Varney/Brown comment was clearly not totally serious. But much of my point has been about the key to data being who uses it, how it is us and accepting its considerable limitations and how I feel that the media often uses it (underpinned by this notion that everything must revert to the mean, see my 15th v 2nd example below) poorly in the absence of a more rounded view.

Another thing about xG that is a limitation unless I am wrong. A striker is put through on goal, he takes the chance normally, that has say 0.6xG. Equally, say he takes a bad touch at the end and the keeper smothers it, the xG is 0. If he rounds the keeper and rolls it in from a yard, it is probably 0.99 xG. All 3 are the same chance created.

2 minutes ago, superniko said:

Because whilst Kaminski and Brereton are better than average and a factor in our over-performance in actual vs expected I stated that I believe luck or statistic size (as a broad term but perhaps could be worded better) is a bigger factor. Those two players alone aren't enough to keep outperforming xG as much as we have this season.

Personally, the flaw I don't like with it is how it doesn't represent game scenarios i.e. Sunderland only started to lead in xG once we were 2-0 up and we sat back and they dominated - would this have happened if Breretons screamer went over the bar? Unlikely to the extent it did. We take the lead, we sit back and we lose on xG - if it keeps getting us 3 points then happy days.

To be blunt though, there is a reason we're 8th favourites for promotion despite being joint top of the league.

I totally agree on game state, and don't think we are favourites for the league but not due to xG. We had the 2nd best xG when we came 15th.

[Guest]

8 minutes ago, roversfan99 said:

Another thing about xG that is a limitation unless I am wrong. A striker is put through on goal, he takes the chance normally, that has say 0.6xG. Equally, say he takes a bad touch at the end and the keeper smothers it, the xG is 0. If he rounds the keeper and rolls it in from a yard, it is probably 0.99 xG. All 3 are the same chance created.

But then you'd pair xG with something like OBV or xGChain. There are loads of xG related metrics which address these issues. Nobody would look at xG alone, it's always in groupings. 

This one might be an interesting read on non-shot models and xG Chain/xG Buildup and the gaps those can plug.

[roversfan99]

15 minutes ago, JoeH said:

But then you'd pair xG with something like OBV or xGChain. There are loads of xG related metrics which address these issues. Nobody would look at xG alone, it's always in groupings. 

This one might be an interesting read on non-shot models and xG Chain/xG Buildup and the gaps those can plug.

Much of my point though was about how it is commonly used (not within football clubs) and how the idea that teams will fall in line with their xG is massively flawed.