Ronald Koeman

[Bobo32]

I wasn't talking about my personal viewpoint with that by the way. The xG stats say what they say but if you don't set store by them much( which I don't) then it's little consolation for the position we find ourselves in now.

Ok
Just wanted to point it out, he didn't show some statistical proof that Barcelona are better this year, the numbers he posted shows nothing so far.
He likes the way they play now and found some numbers going his way, but it's very weak.
Maybe he'll be able to show some significant improvement through numbers, although I doubt it.

I used xG to understand that Barcelona still had a big chance on the liga back in february or whenever, because Atletico had overperformed so much and would probably start to lose points, as they did.
I was also able to predict though that Barcelona would start dropping points again after going clean for a few months, and they did drop points despite "winning" most games in xG.
I predict now that Barcelona will be worse next year if Koeman stays, based on how their games have looked. I hope he doesn't get the chance to prove me right.

 

[SmilerBam]

Just to say Averages per games are terribles.

Since he is here less than a season you can easily go a see every important match, here is a teaser.

Failed EVERY & EACH important moment in the season which is a first low not even Setien has reached.

Now if you have watched our gameplay you could easily see that Koeman is great at "player" management but horrible as tactician, "lets throw 5 players up front the last 5 minutes with hope to score a goal" hilariously bad.

We should find another word for that, because ''abysmal'' doesn't cover it. But hey, we are the kings of xg, whatever the fuck that means.

 

[snowy]

Poor lad, seems like he was quite young at the time too. Disgusting behavior from RK

Sh Vers.: yeah you just don't do that... Good on the lad for not breaking down or losing his cool (ง︡'-'︠)ง

 

[SmilerBam]

But anyway, i don't like the thread title. He may not be a very good coach, but he's still a Barca legend. He bought us our first CL and he was a very important cog in Cruyff's Dream Team. It's embarassing that he has that thread title. It was embarassing when he had Ronald Koemander too.

 

[Birdy]

Tired of this blue balls feeling this season. Before supercopa final vs Bilbao we played great and were even leading only to lose. Before CL second leg we were playing great only to come out with a tie even though we were clearly superior. Before second match with RM we looked deadly on attack only to play like pussies and not attack until the last 15 mins. And then you have the League there for the taking only for us to lose to Granada (in the last 20 mins mind you) and to draw vs a piss poor Levante.

This has been the blue balls season for me.

This is a serious counter-argument to what I write.
It's maybe the first serious response I have read here after pages and pages of trying to discredit xG stats.

What you say pertains to mentality IMO.
Your point can be translated to 'the team has a mentality problem' that appears either when big ties come, or when the moment of high pressure appears (Granada game for instance).
And when this happens, there is a serious mental block, and the fails to perform, or fails to live up to the standards of the football played in general in the season.

That seems to be the case, I agree with that.
What I don't agree with is that it's Koeman's fault.
IMO the mentality problems of this team run very very deep, and go many years in the past.
The only thing I can say is that Koeman has not corrected them yet, not that he is responsible for them

 

[Birdy]

You are tiring out the back and forth, as you repeat points I have already debunked

To give some historical sense and show the trend. If you don't want that, here are the 5 year numbers:
Actual goal difference: 79, 70, 54, 48, 47
Actual # points taken: 90, 93, 87, 82, (82 with 2 more wins)
And if you don't want to see 5 years back, you can see for yourself that the last two seasons are roughly the same in both actual points and actual GD.
Last season had two managers sacked, and was (until this season) the worst season since Rijkaard.

I already showed you that actual goal scored/conceded pertain to effectiveness in front of the goal, effectiveness of the keeper, and have nothing to do with how well the team performs on both ends.
Hence, have also nothing to do with Koeman's work.
Don't make me write it many times.

Ok, I'd think to get a penalty you need to have the ball in the box, it doesn't come from the sky.
It makes sense to exclude penalties from a players goal stats, but I fail to see why you'd do it for a team.
Your (excellent) site fbref doesn't include this stat on the team pages, neither in fixtures or in the tables what I see. I see the simple xG and xGA everywhere.
Could you point me towards where you find these non penalty xG for teams on the site?

You would do it for a team because they taint what you want to see.
The metric here should show how well a team attacks or defends.
Getting awarded 0.75 xG (that's usually enough for 2 goals not even 1, in the open play) for a pen, has nothing to do with laboriously producing chances that amount to 0.75 xG.
AGAIN: simple rationale, used by most data analytic people on the web. Look around and you will see that.

IF you check each season and each competition separately on FBREF, you can see columns for npxG for/against, and at the bottom of the table 'squad total' 'opposition total'

Okay, so we see that Barcelona traditionally overperform their xG.
Less so this year.
It's interesting, maybe this is a significant difference, while I don't think chance creation was.
Do you have an explanation for it, or do you see conversion as pure chance, whereas the xG is pure fact?

Explanation has to do with individual ups and downs of effectiveness.
Many factors behind.
For example, I don't know why Messi was so shit in finishing until Dec. Maybe it had to do with his 'unhappiness'.
But it impacted negatively our points tally, as his crazy finishing under EV impacted it positively.

If there was any massive improvement this year, it wasn't in hindering opponents chances.
Also, if anything it looks as if ter Stegen performed better than in the previous 2 years, or the opponent attackers underperformed.

xG conceded was never good the last 4 seasons.
It still isn't (City who have the best defense in Europe this season have total np xG against 21 or so, to get a sense of where the ceiling is), but not as bad as EV 2 first seasons.

If, if, if.
If these 10 goals you want to count on was given in a single game Barcelona already won, nothing would change but the statistics.
I don't know if you would feel your argument was stronger or weaker then?

They were not.
For example, the Levante game which is fresh in my memory, in the first half we should have been 0-4 up based on clear big chances, which we wasted.
Levante might have never created their chances in the second half if we were clinical in the 1st half.
I can draw many of the games we dropped points this season and show you the xG map that proves we could have won the game if our attackers had converted their big chances.
We are talking about at least 10 points we have wasted like that this season, derbys aside.

Two points that you didn't respond to:

• What's the significance of the change? You can pick either vs last year or vs any other time period, and choose whatever metric you like.
• The expected goal difference based on these numbers would be: 36, 28, 31, 36 roughly. Significant? Satisfied with sample size?

The significance of the improvement of xG for is that the team is not sterile anymore in producing chances, as it was last season, which was the final phase of a steady decline on that department since EV took over.
xG goal difference confirms all my previous remarks.


PS: this discussion is focused on xG which is the 'front-end' stat, but there are other 'back-end stats as well .
I had posted in the past the mid-season review by Statsbomb which listed other significant stats, like pressures, ball regained and others.
There it was evident how Koeman had improved the intensity and the aggression of a pathetic and lethargic team he received after 3 years of EV and his football.
When I see something similar after the end of the season, I will post it again.

xG do not come out of the blue (most of the times at least).
There is some working on the background to produce a good chance. And that is the work of the coach and his tactical plan.

 

[Nazario1985]

@Bidy : Messi contribution this season in xG (mostly solos) is even bigger than last season so i don't see this 'RK tactic or team play' in the stats.

Also i can't find RK style, and i think even himself doesn't know his own style (the guy requested depay fgs so what's style is that mixed with the squad).

i know the season ongoing so %stats can get lower and cumulative stats can only get higher but why not :

And don't tell me he needs 3 seasons to put his style ... it only means he is still looking for one !

 

[Nazario1985]

To this date, every time he took over a squad the first year (which is mostly the work of the previous manager) the team grasp some 'copa del rey' level title, then he sink them next year before getting sacked.

Lost all respect for him after his reaction against Mingueza always a lion against the kids ... hope he try this with Pique !!

He tried to pull a louis van gaal personality with a Cruyff tactic, but without the brain behind it.

Knowing there is a big chance he is going to stay (because no decent name is free), we better get the most we can this summer with players that can break through the game when the "tactic" isn't there.

 

[Bobo32]

You are tiring out the back and forth, as you repeat points I have already debunked

I already showed you that actual goal scored/conceded pertain to effectiveness in front of the goal, effectiveness of the keeper, and have nothing to do with how well the team performs on both ends.
Hence, have also nothing to do with Koeman's work.
Don't make me write it many times.

You haven't debunked anything.
Actual goals scored is a hard stat and what the game is all about. It's high variance and as you say dependent on finisher/keeper isolated. Thus xG was introduced, to estimate what goals are expected on the finishes the team arrived to.
xG is still high variance, and dependent on the effectiveness of the final pass and the final run, and that the finisher releases a shot. It has just about as little to do with Koeman as actual goals, and it obviously doesn't give the whole picture of the game.
But just as actual goals, it obviously says a lot about the performance, and it is lower in variance than actual goals.
If you don't understand this, it'll go in circles and lead to nothing, but I'll still respond to what you've written, and maybe an xG thread can be started where we can discuss this more.

You would do it for a team because they taint what you want to see.

You want to see shots arrived at through open play, and the quality of them?
The simple xG would be a better estimation of expected goals for the team than "non penalty xG + whatever the average penalties given are"

The metric here should show how well a team attacks or defends.

If Koeman told the defenders to just kick the opponents on the feet every time they arrived near the box, they wouldn't defend better, but have a much lower non penalty xGA.
If Barcelona started to shoot twice the amount of what they do, their xG would skyrocket, but the goals probably wouldn't

Getting awarded 0.75 xG (that's usually enough for 2 goals not even 1, in the open play) for a pen, has nothing to do with laboriously producing chances that amount to 0.75 xG.
AGAIN: simple rationale, used by most data analytic people on the web. Look around and you will see that.

Woah woah woah. You say that 0,75 xG usually equals two goals more than 0,75??? Please expand on this.

IF you check each season and each competition separately on FBREF, you can see columns for npxG for/against, and at the bottom of the table 'squad total' 'opposition total'

Okay, so it's just a simple addition of all the values in the table for individual performances.
If it was an important stat for team performance, surely it would be added in a separate column in their team performance tables?
Thanks anyway. If you have some good team analysis at hand where they single out this stat, I'd gladly see it.

Explanation has to do with individual ups and downs of effectiveness.
Many factors behind.
For example, I don't know why Messi was so shit in finishing until Dec. Maybe it had to do with his 'unhappiness'.
But it impacted negatively our points tally, as his crazy finishing under EV impacted it positively.

Yes this is a reasonable explanation.
Of course there are unlimited factors both behind actual goals and xG...

xG conceded was never good the last 4 seasons.
It still isn't (City who have the best defense in Europe this season have total np xG against 21 or so, to get a sense of where the ceiling is), but not as bad as EV 2 first seasons.

No but it is worse than last year.

They were not.
For example, the Levante game which is fresh in my memory, in the first half we should have been 0-4 up based on clear big chances, which we wasted.
Levante might have never created their chances in the second half if we were clinical in the 1st half.
I can draw many of the games we dropped points this season and show you the xG map that proves we could have won the game if our attackers had converted their big chances.
We are talking about at least 10 points we have wasted like that this season, derbys aside.

According to this site it was 2,34xG goals in the first half: https://understat.com/match/15133
One potential good thing with xG is people might start to understand big chances doesn't equal goals...
There are also games which Barcelona did win that they shouldn't have won if both teams converted. You can't pick just one.
Barcelona are on xPTS: 75.33 and actual points: 76, so it is a nice estimation this season for Barcelona. https://understat.com/league/La_liga/2020

The significance of the improvement of xG for is that the team is not sterile anymore in producing chances, as it was last season, which was the final phase of a steady decline on that department since EV took over.
xG goal difference confirms all my previous remarks.

I am talking about statistical significance, ie what is the probability that the change is just variance. Are you familiar with this?

PS: this discussion is focused on xG which is the 'front-end' stat, but there are other 'back-end stats as well .
I had posted in the past the mid-season review by Statsbomb which listed other significant stats, like pressures, ball regained and others.
There it was evident how Koeman had improved the intensity and the aggression of a pathetic and lethargic team he received after 3 years of EV and his football.
When I see something similar after the end of the season, I will post it again.

Please do, it is always interesting. But it is more pleasurable if you think more about what your stats say and what they don't say, and understand that they are all very much subjectively defined.
Try to define where the boundary lies between what a "pressure" is and what is not, for example?
Most stats analysts I read are aware that their stats only help their argument, they don't think the stats are some truth in themselves.

xG do not come out of the blue (most of the times at least).
There is some working on the background to produce a good chance. And that is the work of the coach and his tactical plan.

Here is the main argument.
There is a lot of work to produce a chance, and to produce an opportunity to produce a chance, and to produce a base from which to produce an opportunity to produce a chance etc....
The coach tries to do everything but in reality his contribution is limited.
He isn't some xG focused maniac wanting to just raise these numbers, he wants to win games through actual goals.

 

[Xaviniesta]

When a board is weak unprofessional things happen ...
I'm sure Puig didn't get the same treatment as FCB isn't that stupid.


https://www.youtube.com/watch?v=l0m1f9cZg58

will be interesting to hear what puig says once koeman is gone

 

[snowy]

Irrespective of xG stats, our attack, finishing aside, looks better to me than the sideway snore fest of the previous years...

so yeah now it's clear K needs to fine-tune his human rapport skillz but what IF the next dude sucks even more?

with all 'em managers who got sacked and walked the plank recently, sharks been getting Biiiiiiiiiig!!!

Poch, Mou, Tuc, Wenger, Flick, Lampard, Jardim, Kovac, Lopetegi, Marcelino, Celades, Spaletti, Sarri, Allegri, Vieira, Ancelotti, Pirlo? he got pretty close, uhm who else? Yeah that chess Grandmaster dude and even the Great Guru Grassman himself

why folks congratulate coaches when they get hired? My first question would be damn dawg, when u gettin sacked?

Fire K Hire X and X G forbid turns out to be a dud. Then what? Switch to Y and then ?

And then what, start the whole alphabet again ? switch to hieroglyphs (^^ゞ

 

[Birdy]

Actual goals scored is a hard stat and what the game is all about. It's high variance and as you say dependent on finisher/keeper isolated. Thus xG was introduced, to estimate what goals are expected on the finishes the team arrived to.
xG is still high variance, and dependent on the effectiveness of the final pass and the final run, and that the finisher releases a shot. It has just about as little to do with Koeman as actual goals, and it obviously doesn't give the whole picture of the game.
But just as actual goals, it obviously says a lot about the performance, and it is lower in variance than actual goals.
If you don't understand this, it'll go in circles and lead to nothing, but I'll still respond to what you've written, and maybe an xG thread can be started where we can discuss this more.

Your high-low variance point answers already the need to look first at xG, rather than G.
G is more susceptible to luck/randomness/personal idiosyncrasy than xG

If Koeman told the defenders to just kick the opponents on the feet every time they arrived near the box, they wouldn't defend better, but have a much lower non penalty xGA.
If Barcelona started to shoot twice the amount of what they do, their xG would skyrocket, but the goals probably wouldn't

Nonsensical hypothesis, as no team gains something from deliberately handing pens to the opponents
A pen is awarded. It's simple as that.
Something awarded and not produced through the play of the team does not belong to the way the team plays.

Woah woah woah. You say that 0,75 xG usually equals two goals more than 0,75??? Please expand on this.

I have explained it here in the past as well.
A 0.3-0.5 xG shot is a big chance. When I say big, I mean big. You can reasonably expect your attackers to nail such a chance every time it occurs.
Of course not all players, and especially not all below top- level players, have managed to put out such chances.
Hence, the statistical average will never give you something close to 1, which is statistical certainty.
Even an open goal shot is less than 1 in most xG models around, something like 0.9 or so.
Because statistically some blunders in front of the goal have happened and are included in the way models calculate the relative value.

According to this site it was 2,34xG goals in the first half: https://understat.com/match/15133
One potential good thing with xG is people might start to understand big chances doesn't equal goals...
There are also games which Barcelona did win that they shouldn't have won if both teams converted. You can't pick just one.
Barcelona are on xPTS: 75.33 and actual points: 76, so it is a nice estimation this season for Barcelona. https://understat.com/league/La_liga/2020

Now, the other important factor to pay attention to is the number combined with the value of each chance.
The rough sum one sees in the xG scoreline says something, but not everything. You can get the entire image of xG output only if you look at the xG map as well.
A team can have a total xG of 1.0 in a game, but there are various ways that can be made up.
For instance, if the team has 20 shots of 0.05 xG each, the sum is 1.0 xG. Same if the team has 2 shots of 0.5 each.
But in the first case, the team did not create a big chance the entire game, rather shot too many times from unfavorable positions.
On the contrary, in the second case, the team created 2 huge chances.
Hence, with the same total xG on the first case you can reasonably expect actual 0 goals, whereas in the second you can reasonably expect 2 goals.

Likewise, if you don't just look at the sum of the first half against Levante, but at the shot map as well, you see that Barca should have scored at least another 3, apart from Pedri goal, in the first half (chances of xG 0.38, 0.3, and 0.41 respectively). Hence, a score of 0-4.
What happened instead is that all 3 of these chances were wasted, but Messi scored a minor chance (xG 0.08) in the first half.

I am talking about statistical significance, ie what is the probability that the change is just variance. Are you familiar with this?

No, expand if you want

He isn't some xG focused maniac wanting to just raise these numbers, he wants to win games through actual goals.

You are putting the cart before the horse.
Coach wants the team to score goals, and is doing the best to optimize and maximize chance creation, because that's the way to score goals.
Whether a coach follows a 'scientific' way of doing this and looks at xG maps (like Tuchel did with Aubameyang) depends on the training methods.

When we judge though, how effective a team was in a game, we have to look at the xG output and the xG shotmap to assess that.

PS: More about stats:
1)
FiveThirtyEight is one of the best sites in stats and predictions.
They have some of the most sophisticated models when it comes to football as well.
Read about the SPI index they employ, which shows the relative strength of a team.
Then if you go to LaLiga table and order the teams according to SPI, you will see Barca being top.
It's not an accident that all data analytics sites and people this season rate Barca above RM and Atleti.
They know better than the people here who just happen to hate Koeman.

2) A very promising stat FIVETHIRTYEIGHT also provide there is the 'Non-shot xG', which in the future will be as common as xG is today.
the non-shot xG indicates how dangerous a team was around the penalty box of the opposition. It's a more 'back-end' stat compared to xG.
The definition from the link above reads: "Non-shot expected goals are an estimate of how many goals a team “should” have scored based on non-shooting actions they took around the opposing team’s goal: passes, interceptions, take-ons and tackles"
Non-shot xG does not always translate to xG. For example a team can be very active around the box of the opposition, but fail to find the final pass for example that will get translated to a final shot.
For instance, in the last Barca-Atleti the xG (according to the same site) was close 1.0-0.8, with no big chance for either team. But the non-shot xG was 2.0-1.2, which shows that Barca was much more active around Atleti's box, but failed to translate that to chances.
If you casually scroll down all LaLiga games listed there, you will see that Barca almost always dominates the non-shot xG. For instance, in the Barca - Granada the non-shot xG was 4.3-0.3, whereas the xG was 1.8-0.7/

 

[Nazario1985]

I see 80% RK is staying, but we all know he is an average coach who has been that way for nearly two decades of managerial career.

But the managers market is no good right now, unless Laporta pulls some Perez move (or maybe he can ask him help since he stupidly stood by his side in that super league shit).

 

[Bobo32]

Your high-low variance point answers already the need to look first at xG, rather than G.
G is more susceptible to luck/randomness/personal idiosyncrasy than xG

Variance is one thing, and xG is good in that the variance should be lower, just as I said, yes. The accuracy and precision is another question...
There is no need for anything, but it's good to understand what the flaws are in the models you argue from.

Nonsensical hypothesis, as no team gains something from deliberately handing pens to the opponents
A pen is awarded. It's simple as that.
Something awarded and not produced through the play of the team does not belong to the way the team plays.

No it's not nonsensical, it's a hyperbolic argument showing logically that xG or xGA can't be trusted blindly.
Surely you agree some teams shoot more than others and some teams are more prone for penalties than others.

I have explained it here in the past as well.
A 0.3-0.5 xG shot is a big chance. When I say big, I mean big. You can reasonably expect your attackers to nail such a chance every time it occurs.
Of course not all players, and especially not all below top- level players, have managed to put out such chances.
Hence, the statistical average will never give you something close to 1, which is statistical certainty.
Even an open goal shot is less than 1 in most xG models around, something like 0.9 or so.
Because statistically some blunders in front of the goal have happened and are included in the way models calculate the relative value.

Now, the other important factor to pay attention to is the number combined with the value of each chance.
The rough sum one sees in the xG scoreline says something, but not everything. You can get the entire image of xG output only if you look at the xG map as well.
A team can have a total xG of 1.0 in a game, but there are various ways that can be made up.
For instance, if the team has 20 shots of 0.05 xG each, the sum is 1.0 xG. Same if the team has 2 shots of 0.5 each.
But in the first case, the team did not create a big chance the entire game, rather shot too many times from unfavorable positions.
On the contrary, in the second case, the team created 2 huge chances.
Hence, with the same total xG on the first case you can reasonably expect actual 0 goals, whereas in the second you can reasonably expect 2 goals.

Wasn't it you who said the 0,09 chance of Dembele was scored 9% by absolute fact or am I mixing you up with someone?
The entire model is built around the xG being good enough estimations of the mean of how often the shot is converted. Obviously with some variance.
There should be no difference between 20x0,05 or 2x0,5, but in reality it probably is, as the model is far from perfect.
You need to take into account that Messi (and all Barcelona players) should be expected to score more often than the average, but to say that 0,3 or even 0,5 should be scored all the time is pretty jaw-dropping.

Likewise, if you don't just look at the sum of the first half against Levante, but at the shot map as well, you see that Barca should have scored at least another 3, apart from Pedri goal, in the first half (chances of xG 0.38, 0.3, and 0.41 respectively). Hence, a score of 0-4.
What happened instead is that all 3 of these chances were wasted, but Messi scored a minor chance (xG 0.08) in the first half.

Yep, this is expected and shows the inherent variance of xG.

No, expand if you want

Your null hypothesis in this case would be "Koemans team produce the same xG as Setien/Valverde", then you'd check the xG values and see if the change is big enough to not happen out of pure chance more often than f.e 5% of the time if the null hypothesis was correct. If it wouldn't happen more often than that the change could be said to be statistically significant, and you'd be able to disregard the null hypothesis and have something to report in a paper.
This is teached in any intro stats course, and anyone working with stats or publishing scientific papers will know this, very basic stuff.
I try not to sound like a bitch here, I thought you'd be aware of these things and be able to say something about the significance.
Even if there was significant change though, the control (and more) isn't really that great of course.
Maybe you should take a stats course, as you are clearly interested in it. It would help you argue from statistics...
Though I don't think it should be needed to have an opinion on Barcelona of course
I know you like the way they play now, subjectively. I don't, and to say more about where I'm coming from, I think the squad is actually better this year than previous years. I probably agree with you that Barcelona are playing slightly better than before, but I say it is because of the better squad and despite the worse management.

You are putting the cart before the horse.
Coach wants the team to score goals, and is doing the best to optimize and maximize chance creation, because that's the way to score goals.
Whether a coach follows a 'scientific' way of doing this and looks at xG maps (like Tuchel did with Aubameyang) depends on the training methods.

When we judge though, how effective a team was in a game, we have to look at the xG output and the xG shotmap to assess that.

This is not any more "scientific" than other methods. He just used a tool to help him analyse, could be good or bad, but is not more or less "scientific" based on what you wrote.
No xG is not needed to judge anything. Do you think Cruyff couldn't tell if a team was effective or not? Ferguson? Were they unscientific brutes who would be destroyed by the first coach with access to xG?
xG can be a handy tool if you know how to use it.
Any good coach will understand goals doesn't equal performance, but they'd understand just as well that xG doesn't either.

PS: More about stats:
1)
FiveThirtyEight is one of the best sites in stats and predictions.
They have some of the most sophisticated models when it comes to football as well.
Read about the SPI index they employ, which shows the relative strength of a team.
Then if you go to LaLiga table and order the teams according to SPI, you will see Barca being top.
It's not an accident that all data analytics sites and people this season rate Barca above RM and Atleti.
They know better than the people here who just happen to hate Koeman.

2) A very promising stat FIVETHIRTYEIGHT also provide there is the 'Non-shot xG', which in the future will be as common as xG is today.
the non-shot xG indicates how dangerous a team was around the penalty box of the opposition. It's a more 'back-end' stat compared to xG.
The definition from the link above reads: "Non-shot expected goals are an estimate of how many goals a team ?should? have scored based on non-shooting actions they took around the opposing team?s goal: passes, interceptions, take-ons and tackles"
Non-shot xG does not always translate to xG. For example a team can be very active around the box of the opposition, but fail to find the final pass for example that will get translated to a final shot.
For instance, in the last Barca-Atleti the xG (according to the same site) was close 1.0-0.8, with no big chance for either team. But the non-shot xG was 2.0-1.2, which shows that Barca was much more active around Atleti's box, but failed to translate that to chances.
If you casually scroll down all LaLiga games listed there, you will see that Barca almost always dominates the non-shot xG. For instance, in the Barca - Granada the non-shot xG was 4.3-0.3, whereas the xG was 1.8-0.7/

I rate Barcelona over the Madrid teams as well, based on how I judge their quality and what I saw in games.
Regarding 2) it is a very obvious flaw with the current models that f.e Cascoigne in -96 wouldn't show in xG. It'll be interesting to see how these non-shot stats develop.
As a curiosity; I saw you posting that sites la liga prediction a couple of times and responded on this forum. IIRC I should not be ashemed about my responses then...

 

[Porque]

We back on the Brendan Rogelios hype train.

He just beat Chelsea in the FA cup final.