Would this kind of grading company interest you?

[Peter_Spaeth]

People love to talk about technology. Has anyone here seen a machine, computer, device, whatever, reliably authenticate and grade a card? Until that happens, with due respect, it just feels like feel-good talk to me.

[chalupacollects]

Originally Posted by Peter_Spaeth

People love to talk about technology. Has anyone here seen a machine, computer, device, whatever, reliably authenticate and grade a card? Until that happens, with due respect, it just feels like feel-good talk to me.

Where are those Microsoft engineers that patented a new automated grading process???

Maybe planning their emergence?

[Mark17]

Originally Posted by Peter_Spaeth

People love to talk about technology. Has anyone here seen a machine, computer, device, whatever, reliably authenticate and grade a card? Until that happens, with due respect, it just feels like feel-good talk to me.

Tools that measure size, thickness, etc. are of course in existence. Other methods to determine foreign substances also exist. Think of what's been done in the areas of crime scene forensics.

And more advancements will be developed. You may have missed the first part of my quote:

Originally Posted by Mark17

It might be some years from now, but I believe eventually there will be a tool one will be able to pass over a surface (like a baseball card) and the tool will be able to perform carbon dating without damaging the card at all. So if there is anything modern on a vintage card besides a human fingerprint, someday it should be detectable.

So, can I go to Walmart today and buy a card grading/alteration detection machine? No. Does technology exist to catch a lot of trimming, pressing, and other stuff? Yes. Will detection methods continually improve over time? Again, yes.

Peter, what alternative do you propose? Sticking with table after table of people looking at cards, ignoring technological tools that could be of great assistance?

[Peter_Spaeth]

If and when the technology exists and it's proven to be reliable then fine, but it feels too futuristic to me to be a realistic solution to the current problem. I am glad to be proven wrong as we obviously are not in a good place.

[steve B]

The problem with using most technologies to authenticate cards and detect alterations, is that you also need the baseline information to compare to.

A card that's pressed and trimmed will be thinner, or thinner in the pressed areas. And measuring thickness is pretty easy, even without modern technology.
But the question becomes "what thickness should that card be, and how much could that naturally vary based on the papermaking technology at the time. in the 1880's (where I've seen studies of printed items specifically looking at the paper thickness of hundreds of copies) +/- .003 isn't unreasonable.
That spec may be a bit tighter by 1910, and should be much tighter for modern cards. I just measured a handful of modern cards - list on the postwar side - and didn't find much variation at all.

Other groups have studied the chemical makeup of the inks and seizing(sp?) of the paper. The equipment used is expensive, and requires a bit of interpretation of the results, but some interesting things have been learned.

As far as I know, there isn't anyone who has studied this stuff for cards. If it's in a database, I don't think it's been publicly available.
There also isn't one for any of the other things that would be of interest, like UV reaction. The only bit I know is out there is that the 1991 Topps backs have two different inks, one reactive the other not. Other Topps sets in the junk wax era also have the same issue, which I mentioned once, but haven't pursued yet, not even as far as dividing my sets. I collect a lot more that isn't really cataloged or known anywhere. (And yes, I really need to do at least a few posts on those)

If it isn't know exactly what a normal card should be, no entirely reliable comparison can be made.
Currently, someone with experience can tell. After years you get a feel for when something just isn't "right". But to have a machine do that you have to give it the knowledge.

[marzoumanian]

For decades I was a trade press editor covering the paperboard packaging industry (cardboard boxes and folding cartons) so I'm very aware that paper "breathes (expands and contracts)." Perhaps it will take someone offering a "reward" of, say, $5000 to a computer software wiz to develop that "base" and then test card alternations to see if the software can do its job. If PSA were smart it would develop a way to initially test a card by computer and THEN follow up with the human eye. Just some off-the-cuff ideas. This hobby is still so young but it has to grow and evolve or it will (slowly) die.

[conor912]

What I want to see is a grading company that grades and slabs graded and slabbed cards. I think a card in a slab in a slab would be really spiffy. And slap some f***ing stickers on that thing for good measure. Oh, and LAOs...lots of them....and guranteees.... can't forget the guarantees.

[wondo]

Originally Posted by conor912

What I want to see is a grading company that grades and slabs graded and slabbed cards. I think a card in a slab in a slab would be really spiffy. And slap some f***ing stickers on that thing for good measure. Oh, and LAOs...lots of them....and guranteees.... can't forget the guarantees.

This seems like a ridiculous idea. How would you see the card in between the additional layers?

[Peter_Spaeth]

Originally Posted by conor912

What I want to see is a grading company that grades and slabs graded and slabbed cards. I think a card in a slab in a slab would be really spiffy. And slap some f***ing stickers on that thing for good measure. Oh, and LAOs...lots of them....and guranteees.... can't forget the guarantees.

Lol

[Bigdaddy]

Originally Posted by conor912

What I want to see is a grading company that grades and slabs graded and slabbed cards. I think a card in a slab in a slab would be really spiffy. And slap some f***ing stickers on that thing for good measure. Oh, and LAOs...lots of them....and guranteees.... can't forget the guarantees.

And I want a 3D printer that can print 3D printers.

[steve B]

Originally Posted by marzoumanian

For decades I was a trade press editor covering the paperboard packaging industry (cardboard boxes and folding cartons) so I'm very aware that paper "breathes (expands and contracts)." Perhaps it will take someone offering a "reward" of, say, $5000 to a computer software wiz to develop that "base" and then test card alternations to see if the software can do its job. If PSA were smart it would develop a way to initially test a card by computer and THEN follow up with the human eye. Just some off-the-cuff ideas. This hobby is still so young but it has to grow and evolve or it will (slowly) die.

I think that some functions could be automated. Centering would be fairly easy. there's already software for stamps that measures the perforations, and that's more difficult. Plus, the margins on each pair of sides would change the same with temperature/breathing etc, (And it may not be necessary to measure that closely. Those changes will be pretty small.

I think a first step would be scanning every card in for grading, at a decent resolution, and maybe adding the centering info that would be measured by the software. Maybe dimensional info too, a decently air conditioned office isn't a climate controlled metrology lab, but should be close enough for now.

The data about overall size, what sort of cut is factory etc, could be built, I could do a decent job on thickness just from my own collection. I have other more complicated database things I want to do, but I'm not a computer person, so any of them would be pretty difficult for me.
I did a spread sheet with images of as many of the 48/9 leaf variations as I could find, and that took a lot of time. It came out pretty nice though.

[T205 GB]

Originally Posted by steve B

I think a first step would be scanning every card in for grading, at a decent resolution, and maybe adding the centering info that would be measured by the software. Maybe dimensional info too, a decently air conditioned office isn't a climate controlled metrology lab, but should be close enough for now.

Pretty darn close. Lots of transparency is needed

Pete you are also correct. It takes a lot of $$ to make a machine.

[lowpopper]

Sorry guys but this is nonsense.

A peripheral device authenticating cards? Aside from the storing of the image for
future comparisons, the only usable data that can be reliably extracted from
a scan are measurements for total size and centering. Even that can easily be
thrown off by a print defect or a slightly irregular cut.

As far as alteration detection, the only way to use imaging technology to
identify altered cards would be to have a massive database of card
"before images" to compare to. Even if compiling a database with millions
of entries was feasible, it could never be 100% accurate due to the amount
of cards that come off the same print runs and have virtually identical physical
qualities. There would be too many "perfect matches" to identify the true
original.

I know not everyone has a tech background so I understand the optimism
when a discussion like this starts. At best, implementing technology can ASSIST
us with card grading. It could never be done with perfect accuracy. There will
always be errors: human and mechanical. The best we can do is use before and
after pictures on currently graded or serial numbered cards. Anything beyond
that would not improve upon the accuracy of a human physically measuring and
inspecting a card.

Has anybody calculated the current accuracy rate of the current grading process?
I hear a myriad of complaints regarding the grades that were assessed wrong, but
how many are done correctly? I would imagine it is something very north of 99%,
no? Millions upon millions of graded cards out there and a few thousand are turning
up bad? As far as ratios go, that sounds irrefutably reliable for a paid service. Yes,
one could make the argument that the sample size is smaller because the focus
seems to be on the higher value pieces, but that should not render the lesser value
slabbed cards to be irrelevant here. There may never be a way produce a process that
is void of error, collusion, or dubious behaviors. However, the current grading process
as a whole IS accurate based on the numbers we are aware of.

[Bigdaddy]

I would wager that a AI/machine learning tool set could do a much better job of authenticating and grading cards than a human ever could. At a minimum, it would be more consistent than the current 'which grader got this submission?' system that we seem to have.
There are free apps for your iphone that can measure items, are you telling me that a system specifically designed to measure cards, in all three dimensions, would not be better than a human?
Look what computers are doing in solving crimes, tracing genealogy, facial recognition, image processing, etc. Cameras in the Dallas Cowboys stadium scan the crowd and identify areas where trouble may be brewing in the crowd by analyzing the dynamics of the people. Just today I read about Stanford researchers who used Google street view images (50 million of them) to predict local demographics and election voting patterns. The machines were able to process the images in approximately two weeks, where a human expert would have taken over 15 years. And that was just to classify the images, not to use that data to predict anything.
Now these machines are only as good as the software and sensors on them, but their ability to quickly process images and extract useful data from them is far beyond what a human can do. And they do it much quicker and more consistently.
Are they perfect? No. Is this an easy task? No. But it is one where machine vision along with AI/machine learning would make a tremendous leap in the card grading/authentication clap-trap that we have now.

[topcat61]

Originally Posted by Bigdaddy

I would wager that a AI/machine learning tool set could do a much better job of authenticating and grading cards than a human ever could. At a minimum, it would be more consistent than the current 'which grader got this submission?' system that we seem to have.
There are free apps for your iphone that can measure items, are you telling me that a system specifically designed to measure cards, in all three dimensions, would not be better than a human?
Look what computers are doing in solving crimes, tracing genealogy, facial recognition, image processing, etc. Cameras in the Dallas Cowboys stadium scan the crowd and identify areas where trouble may be brewing in the crowd by analyzing the dynamics of the people. Just today I read about Stanford researchers who used Google street view images (50 million of them) to predict local demographics and election voting patterns. The machines were able to process the images in approximately two weeks, where a human expert would have taken over 15 years. And that was just to classify the images, not to use that data to predict anything.
Now these machines are only as good as the software and sensors on them, but their ability to quickly process images and extract useful data from them is far beyond what a human can do. And they do it much quicker and more consistently.
Are they perfect? No. Is this an easy task? No. But it is one where machine vision along with AI/machine learning would make a tremendous leap in the card grading/authentication clap-trap that we have now.

I use machines already when inspecting my cards -a black light, gram scale, digital microscope and most importantly, my brain. All these things help, but the one thing a machine cannot do is use critical thinking, logic and experience to know what to look for.

[Bigdaddy]

Originally Posted by lowpopper

Sorry guys but this is nonsense.

A peripheral device authenticating cards? Aside from the storing of the image for
future comparisons, the only usable data that can be reliably extracted from
a scan are measurements for total size and centering.

Ah, much more than that. Surface indentations and irregularities (indications of erasures), thickness, weight, florescence of the inks (like when folks use a black light), very small wrinkles, physical characteristics of the edges, foreign substances, etc. By using a sample of known good cards, an machine learning algorithm could be trained to detect abnormalities that are otherwise invisible to the human eye.

Think of it as looking at a card with a different set of eyes - highly magnified and sensitive to things that we don't normally see. Just like a bat sees objects at night when we cannot, or a dog detects smells that are beyond what we can smell. We don't use people to sniff for drugs, we use dogs.

[swarmee]

Originally Posted by lowpopper

I hear a myriad of complaints regarding the grades that were assessed wrong, but how many are done correctly? I would imagine it is something very north of 99%, no?

When you mean "wrong," are you only classifying it as cards that are altered being given a numerical grade?

I would also declare these as wrong:
1) Mechanical errors: this includes wrong set information, wrong card variant, wrong card number, wrong grade attached to card by accident or process failure, multiple cards put in wrong slabs at same time, spelling errors, etc. I believe just this category would easily exceed 1% of cards/coins submitted. It's around 3-5% on the hundreds of cards I submitted.
2) Cards that are marked or miscut that are not labeled with the appropriate "we never remove these MC or MK" qualifiers, even if you ask us to?
3) Cards that are MINSIZ but slabbed with a number grade anyways: See 1975 Topps Mini set collectors thread on CU/PSA board.
4) Cards that PSA could easily identify with an internet search but are unwilling to and return as N9: NO SPEC INFO.
5) Cards that are NOT MINSIZ or EOT but are returned ungraded or slabbed AUTHENTIC ALTERED anyways.
6) Hand Cut cards given number grades despite not following PSA's own rules that the borders must be present?

You still want to tell me their failure rate is less than 1%? Watch some of the PSA reveals from Vintage Breaks and see how many times PSA slabs O-PEE-CHEE cards as Topps and Topps as O-Pee-Chees, even on easily distinguishable sets like 1971 OPC Baseball with the yellow-orange backs and different layout. They are awful at identifying modern card variants, even screwing up superfractors and labeling many hundred dollar variants as the base cards.

If their error rate isn't closer to 10%, I'll eat my hat.

[lowpopper]

Originally Posted by swarmee

When you mean "wrong," are you only classifying it as cards that are altered being given a numerical grade?

I would also declare these as wrong:
1) Mechanical errors: this includes wrong set information, wrong card variant, wrong card number, wrong grade attached to card by accident or process failure, multiple cards put in wrong slabs at same time, spelling errors, etc. I believe just this category would easily exceed 1% of cards/coins submitted. It's around 3-5% on the hundreds of cards I submitted.
2) Cards that are marked or miscut that are not labeled with the appropriate "we never remove these MC or MK" qualifiers, even if you ask us to?
3) Cards that are MINSIZ but slabbed with a number grade anyways: See 1975 Topps Mini set collectors thread on CU/PSA board.
4) Cards that PSA could easily identify with an internet search but are unwilling to and return as N9: NO SPEC INFO.
5) Cards that are NOT MINSIZ or EOT but are returned ungraded or slabbed AUTHENTIC ALTERED anyways.
6) Hand Cut cards given number grades despite not following PSA's own rules that the borders must be present?

You still want to tell me their failure rate is less than 1%? Watch some of the PSA reveals from Vintage Breaks and see how many times PSA slabs O-PEE-CHEE cards as Topps and Topps as O-Pee-Chees, even on easily distinguishable sets like 1971 OPC Baseball with the yellow-orange backs and different layout. They are awful at identifying modern card variants, even screwing up superfractors and labeling many hundred dollar variants as the base cards.

If their error rate isn't closer to 10%, I'll eat my hat.

John, you just read into my post wrong.

to clarify...

Mechanical meaning anything to do with a machine, device, computer, etc.

not referring to "mech error" when PSA mislabels a card.

Bad was referring to the cards on the new suspect list. Not in reference to
under/over grades, mislabels or anything like that. My accuracy rate was
only referring to those cards that are on or will be on the list compared
to anything that is currently sitting soundly in a holder free of suspicion.

Purple Label would not exist if I believed every card was accurately graded
just in regard to the numerical grade.

Hope that clears things up

[CobbSpikedMe]

Originally Posted by lowpopper

Millions upon millions of graded cards out there and a few thousand are turning
up bad? As far as ratios go, that sounds irrefutably reliable for a paid service.

But the millions upon millions of graded cards out there are not altered to begin with. So the actual ratio of accuracy should be calculated on the number of Altered cards that were submitted related to the number that were caught. The millions upon millions of unaltered cards shouldn't be included in the calculation. This would bring the percentage way down from 99% accuracy to a much lower number, maybe 25%. Just a guess. That's pathetic reliability for a paid service.

[T205 GB]

Originally Posted by Peter_Spaeth

People love to talk about technology. Has anyone here seen a machine, computer, device, whatever, reliably authenticate and grade a card? Until that happens, with due respect, it just feels like feel-good talk to me.

Nope. Not even possible.

[ullmandds]

I just don't think the financial investment in such a "device" that would/could grade cards is feasible to create to be cost effective?