Computerized Technical Grading

[Beijim]

In the Regarding the Coin World grading test thread, Todd wrote:

 

In a perfect world grading would be an exact science that something other than an unperfect human who is prone to error would grade coins and there would be no need for such tests.

 

I had a thought about this and felt it might merit a new thread.

 

It is, in fact, possible with today's technology to create an automated system to evaluate a coin and assign it a technical grade. Such a system would comprise a laser scanner providing input to a computer. The computer would have at its disposal a database of all such gradable coins (in all their varieties). In the database would be the physical properties of the ideal specimen of each variety as it would have come off the dies. In other words, the database would in essence represent the 70 grade level, along with any coin-specific deviations from the general algorithm by which the computer determines grades falling short of 70. The computer would compare the coin's scanned physical properties with those of the ideal and compute a grade.

 

Obviously, I'm simplifying the description, but the technology is without question feasible. If you trust a computer laser system to perform surgery on your eyes, I think you can trust it to evaluate a coin's surface for technical merit.

 

The intangible aspects of grading are another matter. I suppose you could design algorithms based on known physical properties that contribute to such things as eye appeal, but I suspect most folks would still want a person doing this. I believe I would.

 

Here's my straw proposal: assign two grades to each coin. The first is the technical grade as determined by the computer. The second is the human evaluation of the coin's overall effect on the viewer.

 

The beauty of this approach is this: I hear people griping up and down about how too many 70's are given out these days. I see much hand wringing and gnashing of teeth over grade deviation and the lax standards of the second and third tier services. In general, I perceive dissatisfaction with the existing level of subjectivity in the grading process. I've personally viewed numerous MS70 or PF70 coins whose surfaces, in my opinion, didn't merit the grade. Okay, fine. The computer would catch every single one of these. A surface defect is a surface defect. The wavelength of light is really small compared to these imperfections. It's totally objective evaluation.

 

Once the coin's technical characteristics are determined (and assuming an accurate and trustworthy database, the result would not be open to debate), the humans can address themselves to the important matter: the coin's net appeal, taking the technical grade into account, along with all intangibles. Buyers and Sellers can assign their own weights to these grades as they deem appropriate, but the technical grade would form an objective anchor in the light of which the Buyer can consider the coin's subjective grade and asking price.

 

Opinions?

 

Beijim

 

[Julian]

Compugrade had a system in the early 90's. I think that PCGS has a system, or may have even purchased the Compugrade system.

 

I totally agree that the technology probably exists to be able to assign a consistent technical grade, with the caveat that there really are no two coins that are exactly the same, with the possible exception of modern proofs . If one disagrees with the computer grade, then a written or verbal case would have to be made to change it.

 

Subjectivity would still be the main force behind the value, just the way it is today, whether the coin is raw or slabbed.

 

There would be no crackouts. The services would suffer financially.

[Oldtrader3]

The technology is now available, IMHO and experience, to computer grade coins with accuracy from an equipment performance standpoint. The optics, algorithym and the computing know-how has come a long way since the days of the first PCGS Grading Computer.

 

Now the software is a different issue. Where a system such as this would get in trouble, would be programmming a system with algorithyms that simulate human judgement criteria on grading per se. Once that has been accomplished, there would probably still be a large component of collectors/dealers who would not accept machine grading.

 

A second (human determined) grade would complicate pricing greatly. If they are different, which grade do you go by? The present system, imperfect as it is, has improved since inception but will always have an element of human fallability.

 

Maybe, the industry takes grading too literally. We all have experiences with Third Party graded coins, even from the most reputable companies, where we don't agree with the assigned grade. Third Party Grading should always be regarded as an educated human opinion, not gospel. Most knowledgeable collectors buy-the-coin, not-the-grade anyway.

[cladking]

Computer technology likely isn't even close to being able to assign the proper grade to a coin consistently. It won't be able to spot altered and counterfeit coins and will overlook large problems. They probably can be programmed to give a fairly consistent grade to virtually all coins but for many coins this will simply mean consistently wrong. Perhaps a better solution would be to have the computer grade it and then a human to make corrections. There would be many benefits in using computers. And in the future they can take over increasingly large portions of the "responsibilities".

[Oldtrader3]

Modern vision systems can scan a coin's devices and match them against every known die set combination in an instant. My money is on the computer being able to spot alterations and counterfeits better than a human can by 100X. I watched (and Project Managed as Plant Manager) a computer vision system project on a high speed medical products assembly machine that could resolve, compare against a standard and accept/reject complex product geometry and defects that the human eye would sometimes miss at 12X. The machine also did it 60 times per minutes with a .1% error rate. Humans did it at 2X per minute with a 15% error rate. This is real life and reduced customer complaints for defects by more than 90%.

 

Seeing through toning may be more of a problem (to determine and measure luster).

[TomB]

I might be a minority of one in this debate, but, I don't think it easily done at all to have computers grade coins consistently and accurately and have those grades be in line with human graders. There are likely several issues to stunt this endeavor. First, as Julian has already written, there would likely be fewer crackouts if everyone truly believed that the system worked. This does not help the third party certification services as it would greatly reduce resubmissions. Also, the generation of computer code sufficient to grade each coin series in each coin grade would seem to require so much effort as to make it financially prohibitive to undertake. Lastly, the computer only knows what it is programmed to know, therefore, a human grader has to program in every definition and permutation of every grade and that would require a human grader to have the precision and consistency of the computer that you are hoping could be made functional. I just don't see it happening.

 

On another note, don't you think that two grades assigned to each coin would exponentially increase the chaos in the hobby/industry? How do you price the coin?

[Oldtrader3]

I agree with Tom. Two grades assigned to a coin would cause mass confusion in the sight unseen market. How do you price such a coin?

[cladking]

There is already mass confusion in the sight unseen market. People value coins differently based on percieved value anyway. If there's anything that can be done to improve perceptions of GRADE on the sight unseen market than it should be done.

[GDJMSP]

There is already mass confusion in the sight unseen market. People value coins differently based on percieved value anyway. If there's anything that can be done to improve perceptions of GRADE on the sight unseen market than it should be done.

 

Well that's easy - do away with market grading and have all the grading companies use the same set of standards.

[michael]

NUTS!!

 

 

 

sincerely michael

[jom]

There would be no crackouts. The services would suffer financially.

 

I've said this a 1000 times: The grading services have absolutely NO incentive to have consistant grading. If they were they wouldn't be in business long. This is the BIG contradiction today's coin market.....

 

jom

[cladking]

I've said this a 1000 times: The grading services have absolutely NO incentive to have consistant grading. If they were they wouldn't be in business long. This is the BIG contradiction today's coin market.....

 

jom

 

 

I don't believe this is true. The grading company's very existence hinges on how collectors view their consistency. They could intentionally misgrade a few coins probably just to increase revenues but it seems unlikely they'd need to so long as there are plenty of coins that are liners. Since coins are graded by price, there will be plenty of repeat business as coins fluctuate in value in various grades because of their various attributes.

 

 

[jom]

Since coins are graded by price, there will be plenty of repeat business as coins fluctuate in value in various grades because of their various attributes.

 

Hence, the inconsistancy. BTW, I never said they intensionally misgraded but there is a perception the services change their standards. Even dealers talk about how "..PCGS is being liberal now" or "PCGS is pretty tough at the moment". I've been to auctions where the auctioneers were happy PCGS under-graded material because that would be there would be stronger bidding.

 

What I meant by "no incentive" is that the services won't make huge efforts for computer grading because if grading ever becomes completely consistant they would be out of business.

 

jom