Maybe not but I wonder if the future could be automated / robotic process that allows for faster, cheaper and more objective grading?
I’d think you could lay a card down on a piece of glass and take high resolution pictures from multiple angles, including below the glass. Some sort of algorithms could analyze the condition of the surface, edges, corners, etc. The centering would be a breeze to analyze. Grade determined.
The algorithm could be tweaked. The pictures themselves could be reviewed for overall appeal and to ensure nothing is out of whack.
Hell, why couldn’t robots be designed to remove cards from card savers and even place them in cases?
Not that any of this is immediate but am I crazy to think this would be the future?
Not only could you do that, but I think it’s inevitable. I think there’s already room for a new grading company to fill this space with current machine learning algorithms, and the huge reduction in labor costs that would be achieved by doing this would be a massive competitive advantage.
While a large hurdle is obviously gaining the trust of the market due to the inevitable high volume of initial grading inaccuracies that would occur vs a trained human eye, the lower prices and faster turnaround time that such a company could offer due to the lack of labor costs compared to PSA, Beckett and CGC would immediately make the company an attractive alternative (I think). Trust could also be established by offering free first time re-grades to dissatisfied customers, which would be handled by trained human graders similar to PSA and others. The size of this crew would still be substantially smaller than competitors, and would give an opportunity for the company to collect additional data to continue to improve its grading algorithm. Once the accuracy eventually improves to the P95 level of human graders due to additional data collected the company could then increase prices and begin charging for re-grades.
I think that there is enough sample data (high definition photos of graded and ungraded cards) freely available on the internet that if properly catalogued and used as training data would result in an algorithm with an accuracy of 80% compared to competitors when first brought to market. Keep in mind that auto insurance companies already use similar algorithms to process claims (those photos you submit to show damage in your State Farm or Progressive app? That’s additional training data for them to improve their model further). While correctly identifying a holofoil print line or a small white dot on a Pokémon card requires a higher level of precision than, say, determining if an animal is a cat vs a dog, or determining whether a car bumper is crushed or not crushed, it can be done.
Cards are made in different ways with different processes and different card stock. This would be hard for a computer to analyse as its all just 1’s and 0’s with the human element it is prone to mistakes but “hopefully” they are knowledgeable enough to pick up on those little intricacies.
One thing ive been thinking lately is they could potentially do is implement a process to take high resolution photos for a grader offsite to examine and grade it remotely then they just have a final check through a grader to ensure no small dings or creases and to make sure cards authentic. Im not sure the concept would work but its something that might be worth looking into
It’s entirely possible, but you have the same problem CGC is facing: PSA is already super dominant in the space and you have to get customers to trust your grades will actually increase value of the card. Otherwise you’re just an automated GMA / MNT / etc.
Don’t know if they even do TCG, just watched a review of them on YT. Thought it was quite neat that you could choose custom colors for your slabs to match sports teams etc.
I think one major issue would be scratches since, even under 50 different light positions, only 1 may actually show a scratch. Additionally, I can’t imagine any photos would be able to capture surface defects, regardless of lighting conditions, due to how small many are.
If they were to program a robot to do that they’d need to program it to actually see, and identify, the card in question, comprehend what is the card and what is just visible though the Cardsaver, adjust for different brands of plastic to be able to position/open the Cardsaver correctly, and detect the location/positioning of the card within the cardsaver and adjust accordingly. And that’s just before the card is removed from the submission holder. That’s not a simple task, that’s a serious challenge to overcome that will require years of development.
Even if they did create such a technology, I’d much rather a well-trained human handle my cards than hope that the machine is programmed correctly.
