The elusive 9.9 - 10.0 1 1

[theCapraAegagrus]

3 minutes ago, kav said:

ok he's doing limits and matrices-thats a bit harder-

[kav]

Just now, theCapraAegagrus said:

Now we're talkin.

[romanheart]

45 minutes ago, valiantman said:

What grade are you placing at the center (median) of the curve? 

The median is the chosen middle item of the submitted samples. In every case you should see the semblance of a bell curve. Most samples should hover towards the 50th percentile (top of the bell).

For moderns you will likely see the higher end of the curve only, since most people will submit only the high end items for grading. For older items, you are more likely to see the full shape.

I don't think a 9.9 should be a unicorn. There should be a pattern to their occurrence.

[valiantman]

2 minutes ago, romanheart said:

53 minutes ago, valiantman said:

What grade are you placing at the center (median) of the curve? 

The median is the chosen middle item of the submitted samples. In every case you should see the semblance of a bell curve. Most samples should hover towards the 50th percentile (top of the bell).

For moderns you will likely see the higher end of the curve only, since most people will submit only the high end items for grading. For older items, you are more likely to see the full shape.

I don't think a 9.9 should be a unicorn. There should be a pattern to their occurrence.

You stated that "From a pure stati[sti]cal point of view, grades should roughly reflect a Gaussian curve and those stellar grades should pop up with a known percentage over time."

Here are Gaussian curves:

 

In each case, the median of the curve shows the same volumes on the left and right of the median, relative to the standard deviation from the median.  Therefore, if there is a Gaussian curve for the stellar grades, as you stated, we'll expect some Y% for 9.9 and the same Y% for some other grade Q, and there will be some Z% for CGC 10 (which is 0.1 higher than 9.9) and the same Z% for some other grade which should be Q-0.1 (0.1 lower than Q), and there should be no copies graded Q-0.2 or lower.  Since that's ridiculous no matter where you put the median, and since there is no reason that there won't be tons of CGC 0.5, 1.0, 1.5, etc., graded books regardless of how many CGC 9.9 and CGC 10 there are, there is no reason to expect a Gaussian curve.

[Foley]

Now do a Poisson distribution 

[kav]

2 minutes ago, valiantman said:

You stated that "From a pure stati[sti]cal point of view, grades should roughly reflect a Gaussian curve and those stellar grades should pop up with a known percentage over time."

Here are Gaussian curves:

 

In each case, the median of the curve shows the same volumes on the left and right of the median, relative to the standard deviation from the median.  Therefore, if there is a Gaussian curve for the stellar grades, as you stated, we'll expect some Y% for 9.9 and the same Y% for some other grade Q, and there will be some Z% for CGC 10 (which is 0.1 higher than 9.9) and the same Z% for some other grade which should be Q-0.1 (0.1 lower than Q), and there should be no copies graded Q-0.2 or lower.  Since that's ridiculous no matter where you put the median, and since there is no reason that there won't be tons of CGC 0.5, 1.0, 1.5, etc., graded books regardless of how many CGC 9.9 and CGC 10 there are, there is no reason to expect a Gaussian curve.

Finally someone who understands math.  I suspect a non-normal distribution for comic grades, based on many factors.

[valiantman]

2 minutes ago, Foley said:

Now do a Poisson distribution 

You're on to something with that one. 

Poisson (cumulative):

Actual CGC Census numbers (cumulative):

[kav]

The biggest problem I have with the idea that CGC is artificially keeping 9.9s and 10.0s low is how would they do that?  Many 9.9s and 10.0s are coming in and CGC is giving them all 9.8s to keep the numbers low.  So what triggers when they do give a 9.9 or 10.0?  A computer program?  THIS IS THE FIVE THOUSANDTH 9.9 WORTHY COMIC-GIVE IT THE 9.9.

And since 2 or more graders are involved, how does that work?  Hey Joe this looks like another 9.9, I was gunna give it the 9.8 like usual but it's the five thousandth book so we give it the 9.9.

"Roger that"

 

[romanheart]

6 minutes ago, valiantman said:

You stated that "From a pure stati[sti]cal point of view, grades should roughly reflect a Gaussian curve and those stellar grades should pop up with a known percentage over time."

Here are Gaussian curves:

 

In each case, the median of the curve shows the same volumes on the left and right of the median, relative to the standard deviation from the median.  Therefore, if there is a Gaussian curve for the stellar grades, as you stated, we'll expect some Y% for 9.9 and the same Y% for some other grade Q, and there will be some Z% for CGC 10 (which is 0.1 higher than 9.9) and the same Z% for some other grade which should be Q-0.1 (0.1 lower than Q), and there should be no copies graded Q-0.2 or lower.  Since that's ridiculous no matter where you put the median, and since there is no reason that there won't be tons of CGC 0.5, 1.0, 1.5, etc., graded books regardless of how many CGC 9.9 and CGC 10 there are, there is no reason to expect a Gaussian curve.

I can't believe I have to debate stats after all these years. You're proving my point here. In all cases it's a bell.

3 of the 4 curves indicate proportionality at dead center with varying standard deviation. Let's say the green curve were further to the right, than the left. That could be skewed as in the case where everyone tanked their final, and the prof had to slide the curve over to prevent failures.

[Foley]

6 minutes ago, valiantman said:

You're on to something with that one. 

Poisson (cumulative):

Actual CGC Census numbers (cumulative):

You never cease to amaze 

[valiantman]

2 minutes ago, romanheart said:

I can't believe I have to debate stats after all these years. You're proving my point here. In all cases it's a bell.

3 of the 4 curves indicate proportionality at dead center with varying standard deviation. Let's say the green curve were further to the right, than the left. That could be skewed as in the case where everyone tanked their final, and the prof had to slide the curve over to prevent failures.

You're describing a symmetrical bell distribution.

It will not be symmetrical bell for comic books... ever.

[kav]

4 minutes ago, romanheart said:

I can't believe I have to debate stats after all these years. You're proving my point here. In all cases it's a bell.

3 of the 4 curves indicate proportionality at dead center with varying standard deviation. Let's say the green curve were further to the right, than the left. That could be skewed as in the case where everyone tanked their final, and the prof had to slide the curve over to prevent failures.

A bell curve of the standard deviation does not necessarily mean a bell curve of the items being analyzed, if I remember correctly.

[Get Marwood & I]

But where does the bell end?

[valiantman]

Just now, Get Marwood & I said:

But where does the bell end?

Don't Google it. 

[romanheart]

17 minutes ago, valiantman said:

You're describing a symmetrical bell distribution.

It will not be symmetrical bell for comic books... ever.

What's causing the asymmetry?

In the case of moderns comics, and fair equally balanced grading scale, I would think it should happen with 0.1-1% prob of slabbed books.

Edited October 17, 2019 by romanheart

[valiantman]

33 minutes ago, romanheart said:

What's causing the asymmetry?

In the case of moderns comics, and fair equally balanced grading scale, I would think it should happen with 0.1-1% prob of slabbed books.

The asymmetry is that 9.9 and 10 will be some percentage, whatever it is, and 0.5 (and 0.4?) will not have any symmetrical relationship to 9.9 and 10.  Even if you allow for 9.9 and 10 to be "the two possible grades on the extreme high end" and 0.5 and 1.0 to be "the two possible grades on the extreme low end", there's no symmetry between 9.9 and 1.0, and no symmetry between 10 and 0.5.  The shape will basically always be skewed to the "slab-worthy" grades for a particular book or if all copies are "slab-worthy" (such as Amazing Fantasy #15), you'll see (in the case of AF #15) a grand total of 5% for all of the grades 10, 9.9, 9.8, 9.6, 9.4, 9.2, 9.0, 8.5, 8.0, and 7.5 combined, and then on the other end, 0.5 will represent 5% all by itself.  No symmetry.

It doesn't really change for modern comics, even if only 0.1% are 10 or 9.9, there's no other grade combination that should be 0.1% the same standard deviation(s) away. It's always asymmetrical. 

Edited October 17, 2019 by valiantman

[Hollywood1892]

On 10/15/2019 at 4:52 PM, kav said:

If you seek the elusive 9.9s and 10.0s you will be a very frustrated individual.  If it were this easy, people would already be doing it up the yin yang.

It does seem this easy for Batman Damned

But not for much else

Maybe the book was graded in inches

[romanheart]

10 minutes ago, valiantman said:

The asymmetry is that 9.9 and 10 will be some percentage, whatever it is, and 0.5 (and 0.4?) will not have any symmetrical relationship to 9.9 and 10.  Even if you allow for 9.9 and 10 to be "the two possible grades on the extreme high end" and 0.5 and 1.0 to be "the two possible grades on the extreme low end", there's no symmetry between 9.9 and 1.0, and no symmetry between 10 and 0.5.  The shape will basically always be skewed to the "slab-worthy" grades for a particular book or if all copies as "slab-worthy" (such as Amazing Fantasy #15), you'll see a grand total of 5% for all of the grades 10, 9.9, 9.8, 9.6, 9.4, 9.2, 9.0, 8.5, 8.0, and 7.5, and then on the other end, 0.5 will represent 5% all by itself.  No symmetry.

Quoting CGC stats and saying it is asymmetrical because "it is" doesn't really answer the question. Why is the deviation so small, or the skew for the curve so far to the left? If there is a factor that I haven't thought of I'd be happy to say I was wrong or that I missed something.

In the case of modern comics though, I think the factors should warrant a more regular occurrence of those grades.

[kav]

4 minutes ago, romanheart said:

Quoting CGC stats and saying it is asymmetrical because "it is" doesn't really answer the question. Why is the deviation so small, or the skew for the curve so far to the left? If there is a factor that I haven't thought of I'd be happy to say I was wrong or that I missed something.

In the case of modern comics though, I think the factors should warrant a more regular occurrence of those grades.

Many flawless things are rare beyond all proportion.  Flawless diamonds are very rare—so rare, in fact, that it's possible to spend a lifetime in the jewelry industry without ever seeing one.

[valiantman]

38 minutes ago, romanheart said:

54 minutes ago, valiantman said:

The asymmetry is that 9.9 and 10 will be some percentage, whatever it is, and 0.5 (and 0.4?) will not have any symmetrical relationship to 9.9 and 10.  Even if you allow for 9.9 and 10 to be "the two possible grades on the extreme high end" and 0.5 and 1.0 to be "the two possible grades on the extreme low end", there's no symmetry between 9.9 and 1.0, and no symmetry between 10 and 0.5.  The shape will basically always be skewed to the "slab-worthy" grades for a particular book or if all copies as "slab-worthy" (such as Amazing Fantasy #15), you'll see a grand total of 5% for all of the grades 10, 9.9, 9.8, 9.6, 9.4, 9.2, 9.0, 8.5, 8.0, and 7.5, and then on the other end, 0.5 will represent 5% all by itself.  No symmetry.

Quoting CGC stats and saying it is asymmetrical because "it is" doesn't really answer the question. Why is the deviation so small, or the skew for the curve so far to the left? If there is a factor that I haven't thought of I'd be happy to say I was wrong or that I missed something.

In the case of modern comics though, I think the factors should warrant a more regular occurrence of those grades.

I'm not sure what factors you're hoping I'll be able to point out but here are a few things that keep symmetrical bell curves from happening (symmetry may happen temporarily by chance (see #6 below), but not because it is the expected result):

1) Comics which are 9.9 and 10 are essentially flawless, but the process of creating comic books doesn't protect comics from damage, they're (unfeeling) machine-pulled, pushed, rolled, smashed, stacked, and then (uncaring) human-handled, boxed, rehandled, shelved.

2) Different materials used for the cover impact the rates at which 9.9 and 10 are observed, that is, chromium, metallic, cardstock, lenticular, and squarebound are far more likely to be CGC 9.9 or CGC 10 according to the CGC census.

3) The average comic book is in "the best condition it will ever be" the first day it is on sale and increases damages from that point forward.

4) As damages increase over time, the "bell curve" for any comic book would always be shifting higher quantities toward lower and lower grades.

5) If every copy was CGC-worthy, such as Amazing Fantasy #15, we'd see there are 10 high grades (7.5 to 10) which (combined) account for 5% of the CGC Census and just one low grade (0.5) which accounts for 5% of the CGC Census: asymmetry by definition.

6) Because damages increase over time, any "bell curve" identified which happens to be symmetrical at one point in time will not be symmetrical at the next point in time, because of #4.

7) Comics which are 9.9 and 10 are (almost literally) shiny objects, which draw attention to themselves, and are submitted to CGC more quickly than lower grade copies, meaning that the rate of 9.9 and 10 submissions to CGC for any comic should always decrease over time.

8) Weird stuff.  Example: Someone, somewhere, did once (or twice? looks like 2004 and 2011) submit an unopened case of Spider-Man #1 (1990) to CGC and receive many copies of CGC 9.9 and CGC 10 at a much higher rate than previously seen.  It's weird to have an unopened case of a single issue decades after its release, but whatever the curve looked like before, it most definitely shifted as a result of that one action.  So... #8 = weird stuff. 

Edited October 17, 2019 by valiantman