I mentioned computer analysed grading in another thread…but I cant really see it happening. Also the amount of jobs that would be lost, and it wouldnt be difficult to find dents / minor scratches? the tech would have to be very advanced and I could see issues. But it is an interesting thought and is it possible? maybe. Computer scanned and analysed then a grader looks over them with gloves ;p or the scans.
EDIT:computers could measure the centering etc but they would need extra resources, increased time un less they expand and have more people grading
Modern machine vision and machine learning techiques are advanced to the point where it shouldn’t be difficult to write a card-grading algorithm given a large enough dataset of scans of cards graded by humans. Especially for things that remain consistent between cards such as backs or border centering.
It wouldn’t be a 100% perfect system since there may be weird cases where something in the artwork might appear to the algorithm as damage.
I think the ideal case would be having a machine identify and highlight what it believes is damage and have a human verify and assign the final grade.
I’m 100% for this. I have a feeling it might become a thing sooner than we think. There’s already the technology in place for it, you just need the right software and a company willing to invest in such a project. A machine will spot altering or tiny dents/creases/scratches 100% of the time.
I’m not talking about actual robots handling the cards, but what the last 2 posters have in mind would be perfect. I actually brought this up in another thread last week.
I would also like the opportunity to receive some actual data from items I send for grading. I don’t care for having the subgrades on the actual case like BGS does, but it would be cool if you look up the cert number you get the centering %, surface wear/damage,etc…
I agree with this. A machine could be used to identify the centering ratio, areas of whitening, etc. The human could then use quantitative info from the machine along with his/her own human experience grading cards to assign the final grade. I don’t think strictly using a machine would be the best idea, but having quantitative info for the human grader would go far in adding justification to a grade.
Indeed. It would be a win/win for everyone. The only negative is they’d lose business from the crackers/regraders, but that’s actually a positive for any hobby.
I know grading is not an exact science, but it doesn’t have to be so subjective and there are tools the big companies can use to pretty much eliminate misgrades.
@reinasierpe I’m pretty sure there’s equipment out there that does 360°/3D scanning in high resolution of any object in just a couple seconds. We’re not talking your scanner in your office or a HD camera, but actual professional equipment these companies could use to help with the grading process. It would be very expensive, but it would pay off on the long run.
I mean, one of the motivations for adding an automated grading software into the pipeline is that the 20-second human evaluation does not appear to be sufficient. I’ve seen plenty of examples of very clearly misgraded cards. People who continuously crack and resubmit also depend on graders to miss damage. Certainly adding an additional step will take more time there is the benefit of higher consistency and accuracy. Such a system could even detect if a card was already graded and assign it the same grade, totally destroying any motivation to crack and regrade. If turnaround time was all we cared about, we could just take 1 second to assign a random grade and just fire all professional graders completely.
Also on a side note, PSA could hire minimum wage employees to scan the cards into the machine prior to the graders receiving them. That way, the amount of time a professional grader has to look at the card is the same amount of time (or less).
In many other vision-related tasks, machine learning algorithms have already exceed what humans are capable of, so on an objective level I’d imagine an computer could grade a card better than a human. If you combined an image scan of the card with a second scan that provides depth information along the card, that’s all the data you’d need. Whether or not someone submitting their card will be pleased with the job of a computer over a human is another story. One of the hurdles that self-driving cars still have to overcome.
There are always 2 main holes in this idea, the cost & reality. The reality is a huge nuanced dilemma.
Realistically a card grading machine is a terrible idea for a grading company. The risk of damage alone would be through the roof. People complain that psa put a ruler next to a card, but putting it in a machine is somehow safer?
Also, grading different eras requires different eyes. I wouldn’t want a machine to grade any historic rookie card. More specifically, anything with a programmed understanding increases the potential for forgeries.
Then you get into the reality of cards that sit between grades. I am sure we could do a test by handing 50 cards to seasoned collectors and their opinions would vary on grades. Some cards don’t have “one finite answer” and that is the point.
Machines are already used at PSA to assist signatures & authentication, which is great. They do wonders for autos, but the final opinion should always be done by a human.
Oh and even if it were a reality, then you have the cost. People want a business to invest in the the most advanced technology, but still expect to pay $7 a card…
Ultimately the customer is paying for an opinion, not a programmable algorithm. If the latter were the standard, there would be a decrease in submitters, myself being one of them.
This is a overly simplistic view of how an actual digital grading assembly line would likely work. Modern scanning technologies allow visualization of surface topology and other characteristics that the human eye can easily miss. A complex scanner with multiple cameras would be not limited to holding a card at a certain angle in order to detect imperfections. There is no doubt that on corners and edgewear alone, a basic scanner + data crunching algorithm would far outstrip any grading company in terms of efficiency and consistency.
Also, this idea that a mech-based process would be slower or less efficient than human beings even if a human grader is thrown into the mix is simply untrue. The loader would just load a stack of cards into the machine and sit back while the computer handles the rest. At the end, multiple graders would go through to verify, but for the most part the grading would be done, so they’d spend far less time grading and just giving the cards a ‘once over’ for QA.
This process sounds like a nightmare for any collector. Imagine a stack of 1st ed base holos, vintage mtg cards, rookie cards, etc. No collector would want that process for any of those items.
In general I don’t think people understand the risk aspect as a business. We are dealing with collectibles, and historical items. Not only is there significant money involved, there is a high level of emotional value. PSA already receives numerous damage claims, some legitimate, some ridiculous. Regardless they all stem from emotion. Telling the average submitter who submits $1000+ cards on the low end that their items will be slotted in a machine doesn’t exactly scream consumer confidence.
I think the technological aspect would be great for an individual pre-grading. However I don’t think this is realistic for this type of business. They would open themselves up to so much litigation.
I’m all for using tech to pre-screen for surface damage, but it should be a step in the process, not the process itself. The main barrier is the fact that technology evolves rapidly. There’s no way to develop the definitive card grading tech on the first try, and any improvements would invalidate the grades issued by the previous version.
I think people complain, not so much about the ruler, but the fact that someone was taking their collectible that was in a sleeve/top loader for X number of years and handling it like a carpentry project. It looks unpleasant.
In card grading, people are not paying for opinions, they’re paying for a standard. People pay for opinions because that’s the only means to achieve standardization of collectibles today. When someone buys a PSA 10, they’re buying the card because it’s perfect/close to perfect, not because some guy said he thinks it’s perfect. There’s a difference. The idea of perfection is the standard that PSA promises and “guarantees” in its 10 grade, ironically, because they’re trying to work like a machine would in the first place. Ideally PSA would like 100% consistency across the board, where every 10 looks and feels the same in quality. PSA can’t do this but they come closer than anyone on the market, that’s part of why they earn their reputation as the industry standard. If someone is able to replace PSA with an actual machine that can grade cards based on data, there would be a far more stringent standard that people would gravitate towards because of its consistency.
Imagine a complex scanning apparatus that renders case-cracking useless because statistically, 9999 out of 10000, you’re going to get the same grade. The machine doesn’t have a different opinion of an 8 than it did yesterday. An 8 is an 8. A 9 is a 9. You’re not going to get an 8 upgraded to a 9 because another person is grading it more leniently.
The thing about forgeries is that, there’s going to come a time when forgeries will become virtually indistinguishable from regular cards. MTG is already having issues with people producing forged cards that are barely distinguishable from real ones: Alpha Investments has a few videos on this. Pokemon is far more popular as a collectibles market than Magic so that should be an indication of where we’re headed.
Once forgeries become amazing, digital grading may be our only resource as collectors for weeding out the impostors from genuine articles. A machine can detect microscopic differences in texture, layering, and even transparency of a card that the human eye could never detect. At the very least, I’d say a digital any-forgery machine is something that needs to be added to the card grading process in the near-future. And I agree about the part where a machine can’t detect everything, but who says it has to be either/or? Let a machine do the grading and a human do the authenticating.
In the long run, mechanization always greatly reduces the cost of any process. Human labor is very expensive. In the short run, a grading machine would cost more, in the long run as the process is perfected, it would be more efficient and cheaper.
I think you’re imagining a printer feeding cards through a roller. I agree, a paper jam sounds pretty horrifying when there’s a 5 figure item involved.
I don’t have the blueprints for a perfect machine, but ideally, it would work like I described: you’d feed a stack into the machine and it would grade. How the machine handles the cards would probably need to be invented with a dozen or more patented processes.
I don’t think they’d open themselves up to any more litigation than they already have at the moment. Also, there’s nothing that says more expensive cards can’t be handled differently or more carefully. A grader could personally handle 4-5 figure items. The rest of the cards could be mass fed through the process, etc.
There are robots that assist in surgery and can maneuver with more precision and accuracy than any human hand. The robots provide visual insight that can’t be seen with the human eye. These machines are being used with a human life at risk. The most valuable collection PSA has ever graded is worth a fraction of a human life. The problem isn’t that a machine can’t be build to grade cards more objectively, because it can be. The problem would most likely be cost and/or collector acceptance.
@jj1 I agree with implementing technology for authentication, that is what PSA does for autographs. Forgeries do ultimately come down to an opinion. Having a tool to help assist that opinion is beneficial.
For modern hobbies forgeries are negligible. For historical items, there are reprints with the correct era paper stock, ink, etc, but are still illegitimate. This is where a machine would always fail.
The implementation could help authentication and the general process. I just can’t see this being a reality for a business. There is so much risk involved.
In an ideal world, it would probably be 50/50 if I want to answer the OP.
I agree with most of JJ’s points, except for the feeding of stack of cards. You don’t want to leave a stack of raw valuable cards unattended and just let some machine shuffle them and possibly damage them. This can easily be done manually.
At the very least, there’s no reason a company like PSA can’t have equipment that measures a card for possible trimming and also determine the exact centering. Just as an example, I’ve sometimes seen identical BGS cards and one has a 8.5 centering and the other has a 9. Why is that a thing? That small error prevents a card from being Gem Mint. In other cases, it also inflates the overall grade.
There’s many little things like that they can implement to reduce human error or damaging a card with their metallic ruler.
Some of you think most ideas in this thread are unrealistic, but a few years from now most everything will be automated whether we like it or not.
I actually like this idea, any card with a declared value of $50 or less gets handled by a mass feeding process. A machine that mass grades cards could be very beneficial. Cards can get graded much faster, more consistent, make grading cheaper and it’s only an option.
I think machines should grade all cards but handling is different based on value. With items that are worth more than $50, they can get handled by graders with a personal grading machine.
This is a win win for everybody, mass grading could cost like $3 and it’s only risky to those who choose it while lessening the turnarounds for standard grading services.
If the human eye can’t see it or spot it easily then I don’t want it figured into the grade.
Plus, I don’t care what you do. This new crop of Pokémon graders are going to snivel about their results and/or the process.
Can anybody tell us what approximately would be the total cost of R&D, production of etc to put some of these machines online at PSA?
I know just to purchase a robotic surgery machine for minor surgeries is about 2 Million US dollars. Plus expensive parts need to be replaced daily and upkeep is intensive. Now to develop that machine the cost was enormous.
It’s possible grading costs would have to be doubled or tripled which would be hard for customers to accept. Heck, weve had entire treads devoted to people complaining about not having 2.00 off specials each month.