The Philatelic Foundation Grading Upgrade

Hi ikey,
I can see how the data could show that sentences before noon differ than those after noon; but unless there is more data it is a huge jump to say that this is related to breakfast. Scraping data often produces strange and unexpected results but root cause correlations are much more elusive. For all we know the delta could be related to being frustrated by the morning commute.

AI is heuristic and this is similar to human subjectivity since as time/experiences pass the software modifies itself. But given the lack of drive our hobby has for technical advancement and the fact that there are few related commercial opportunities that we could piggy back upon, I doubt that AI grading is within reach in the next 25-30 years. Frankly, our hobby lacks the technical vision and stakeholders that are even willing to drive an existing hobby-wide grading specification as an standard. This makes it difficult for me to see how it would ever come together enough to agree upon a grading AI module.
Don

Rodgcam

I think AI could easily analyze questions like heavy cancel, cancel over face, shorter perf tips, brightness of paper, sharpness of impression.

But in my opinion PSE really focuses mainly on centering,

Reality there are a lot of graded 95 classics but they surely are not all equal
Buy the stamp not the grade

I prefer jumbos over high grade. Meaning I'll take a jumbo perforated 90 over a regular 95, 98 or even 100.
I personally cannot discern difference between a 98 and 100 with my eyes anyway. So what is the point?
Point is that a 100 can sell for 5x a 98 and that is the subjective corruption that is available to humans to make money.

Hell if You worked at PSE , You could buy a bunch of 98 from PF and regrade them at 100 at PSE.
Boom profit
This scares me but not saying that someone does this.

Morning all,

Yep, to me this is a hobby or entertainment but not an investment or money maker except for the P.T. Barnum types who successfully hype their wares. At times philately is a refuge for ice salesmen who can no longer make a living in Alaska during this record hot summer.

That being said what is the functional value of subjective grading other than a marketing tool? A used car is a used car even if the salesperson assured you it is a cream puff.

Rgstamp - Could not agree with you more. I buy the stamp. It so happens that the stamps I like are also graded relatively high and I love jumbos. Personally after 95 I cannot see the difference by eye and at that point it would become a purchase for the sake of bragging rights or perhaps playing the market or both.

AI and analytics is my business. AI can absolutely handle eye appeal, plating, authentication, and all sorts of other things that we do manually now. You don't need AI, but technology can handle color and centering MUCH more accurately than a human. Doing "Eye Appeal" with AI would be a textbook example of supervised machine learning. I agree with Don that there's not much aptitude or interest in using technology in our hobby, but I don't think it will take 25-30 years to get there. I think the switch to doing more with technology in our hobby will happen in 10 years when more the now 30- and 40-year olds start replacing the 70- and 80-years olds. The only limiting factor now is the access to the data. The low-hanging-fruit would be to just load historical catalog and auction data (without pictures) into a database and do some simple analytics on it to figure out pricing trends. But, all that data (for US stamps, at least) is locked up with the luddites at Amos and only available on paper. The technical problems have all been solved already - using AI in our hobby will just take someone with interest to put the effort into getting the data.

I'm with rgstamp on grading. It's almost entirely centering-focused. I can't tell you the number of graded items I've seen on SAN with 90+ grades, yet with short or blunted perfs, subpar color or impression, or paper flaws.

It's a centering game and nothing more IMO, and the premiums paid for it are staggering.

I've gotten to the point where I will only consider buying an item with a graded cert if there is no premium associated with the grade. Luckily for me, U.S. revenues are not included in SMQ, and I hope they never are.

As rgstamp also states, a nice jumbo is much more pleasing to me aesthetically than mathematically perfect centering.

Here is an example of two 315's that have different grades but are both of the same qualities.

This stamp is from a Siegel sale graded at 98:



I just purchased this stamp graded at 95J:

Which grading companies and how old are each of the certs? The top example looks more J to me than the bottom.

Siegel stamp is 2009 PSE and my stamp is 2003 PSE.

Bizarre. Why wouldn't PSE have given the 2009 stamp a J when they did on the 2003 stamp? That makes no sense to me. Another negative mark in my book for PSE, but then again my opinion of them has never been very high.

Grading standards have changed! Early certs with grades not equivalent to more recent grading.
Nice stamps rogdcam!

It's only a 6-year difference. It's not like we're talking about a grade from the '80s or early '90s vs. 2009.

But more importantly, what would disqualify the top stamp from receiving a J designation... PSE graded it numerically higher than the bottom one... but didn't consider it a jumbo.

"To J; or, not to J?" This is a question from a novice grade person.

If you destroy a block of nine of this stamp making the center stamp a jumbo center with as big as possible margins on all sides will it wind up taller and wider than the perf version of the same stamp printed on the same press? Any consideration positive or negative if it winds up oversized?

Here is what I think, not that it means anything.

Centering should be its own quantitative measurement, not subject for debate. It should have 2 numbers a horizontal and vertical.


On perf stamps, It should contain a standardized measurement. This is the measurement from the stamp graphic to the inside lip of the perfs, or another agreed upon edge. The left should be subtracted from right and the result is the skew delta. The ideal number should be 0. The same should be done vertically. Then add those 2 numbers. The closer the skew is to 0, the better the centering. Computers should be used and the rubric for measurement using some defined dpi. Websites should be able to analyze centering trivially within some acceptable standard for error. It just needs consensus. I'd imagine imperfs would work same way with a predefined minimum margins.

Finally qualitative analysis is required and a grade can be assessed. The F/XF, etc should be sufficient without a focus on centering at all. Pulled perfs, cancel quality, etc.
Then that state of the stamp should be MNHOG.

My 2 cents. The issue is trying to roll up all of these things when they are orthogonal. Maybe you get the gem when you hit perfect across the board as a magic term, but everything short of that, is just not a gem.

I also think that standards need to be different for early perforated classics as opposed to say Wash-Franks.

Here is an example:



This Scott 27 had a clean 2000 cert from the PF. The 2019 PSE cert notes a short perf on the top right and assigns a grade of 75 with a value of $1100. Without the short perf it would have recieved at least an 80. Grade 80 is $1500. Grade 85 is $2500. All for a short perf. It does start to get ridiculous IMHO.