Calling on all opinionated coin graders, your help is requested! Following up on my previous post, it has been a blast building out my AI coin grading and identification app these past two weeks. Folks here have given a lot of great feedback and I've made numerous updates to Numi since. Here is a live video demo showing the latest update with PCGS API integration. A key part of Numi's logic is that it references detailed grading criteria I've compiled based on the top TPG's public grading standards. Whenever Numi is asked to grade a coin, it references these grading standards to give a final estimated grade. I've been running around to various coin shops to talk with dealers to get their feedback on the wording of Numi's grading criteria, but I'm getting a lot of mixed messages. One coin shop dealer said, However the next day another dealer told me, So now I'm turning to the wisdom of the crowds. I'm looking for opinionated coin graders to share their thoughts on Numi's grading criteria. Below is what Numi currently uses: I would love to start a discussion and see what you believe should be changed. I have also created a public Google doc with these criteria for easier sharing and commenting. I'll update Numi's logic based on the discussion here, continue to test the app with graded slabs, and post the results here. This continues to be a wonderful community. Thank you all for the feedback you've shared so far. Let's see if we can make coin collecting more approachable. If you want to play around with Numi directly, you can check it out here. - Dansco Dude
I'll ask you one question. Are you really sure you want to design this system to grade coins based on the way the TPGs grade coins in today's world ?
I understand your intent and admire your energy concerning this project, but I do not understand the foregoing statement. How would such a tool make collecting more approachable? The programming behind this tool will be no less fluid than the opinions which its development is intended to address. Don't get me wrong, but no automated tool is going to supplant the sensitivities of different individual buyers / sellers to varying problems and attributes on many coins. There will always be subtleties and nuances to address, and tradeoffs and contradictions to reconcile. How would such a program make collecting more approachable if, after a transaction, the customer still feels remorse because the coin falls short of expectations for the grade, or the seller still feels shafted by a lower grade than seems justified?
I like what looks good to me regardless of what the tpgs or AI say about it. I am the best grader I know for what appeals to my eye.
It's a good question. I would say initially yes, I do want Numi to consistently give grading estimates that are in line with TPGs today. I struggle to see the numismatic world moving away from the Sheldon Scale anytime soon. NGC is trying to mimic the card-collecting scoring system via NGCX to make coin collecting easier to approach for new collectors. But my sense is that NGCX hasn't picked up adoption since the Sheldon Scale is so entrenched. And in reality, has just added more confusion to the marketplace. I also don't see AI grading as replacing TPGs with its own grading criteria. At least not anytime in the near future. Rather AI grading apps like Numi would be an initial tool used to see if it may make sense to ultimately send a coin in to the TPG. So for now, I'd argue the quickest way for AI apps like Numi to assist coin collectors is to remain in line with the TPG standards. A key expectation I try to lay out with Numi is that it should be considered a preliminary opinion or grading sanity check for collectors. Basically a second grading opinion [the first opinion being the collectors] before they send it to the TPGs. In order for Numi to be valuable to collectors, I'll first have to adjust it to be as accurate as much as the model allows. It's entirely possible the GPT4 model is currently too weak, but I feel I have not hit that barrier yet given the progress I've made and the increased accuracy I've seen. I'd argue that those feelings of remorse or feeling shafted already happen today with the results folks have gotten from the TPGs The difference is that with AI grading [assuming AI becomes powerful and accurate enough], collectors would get grading results immediately. Before they had to sign up for TPG membership, pay submission fees, and wait months for the results as they have to today. As long as AI app like Numi can lower the barrier to knowledge for collectors, be it with grading, identification, or general access to information, then it will make coin collecting more approachable and is worth the effort to investigate AI as a solution. In case you are curious, here are the grading standards from the TPGs I based Numi's grading criteria off of PCGS NGC
OK, but given your purpose, which I've highlighted above, have you decided to use the grading standards of one specific TPG or are you trying to create some kind of aggregate and or generic set of grading standards ? I ask this because each of the 4 major TPGs - ANACS, ICG, NGC, PCGS - all have their own specific and individual set of the grading standards and all four of them are quite different from each of the others. And, it's worth noting that the grading standards used by each TPG are different for any specific series - i.e. the grading standards are different for a half dollar than they are for a dime, even when it's the same TPG grading both. They are even different for a half dollar from the 1800s than they are for a half dollar from the 1900s. And given all of that, after using Numi, which of the 4 TPGs would Numi recommend a collector send his coins to ? And how would Numi decide which of the 4 should be used ? And lastly, what would Numi do should the TPGs decide to change their grading standards again as they have already done a few time in the past ?
Gee, you left out CAC and CACG. Oh well, maybe you don't know about them, or you do, but they don't meet your exhaled Grading expertise. Not to mention ICG. Good Lord. I think your definition of "major" might be a bit skewed.....
I'm aligning it to match PCGS' standards since PCGS has the strongest reputation in the hobby. Not to say that NGC is bad or a distant second, but I'd rather Numi align with PCGS. I'm tailoring Numi so that it gives grading results that align with PCGS for non-ancient coins. And with NGC for ancient coins. It's not so much defining grading standards towards one TPG's standards, but rather having so that Numi gives results consistent with a certain TPG [i.e. PCGS]. i.e. When it analyzes an MS65 coin in a PCGS holder, it always gives back an MS65. Numi will not give a recommendation to the user on which TPG to use for submission. Thought I might include context to the user that estimated grades best align with PCGS. If/when that happens, I may end up updating Numi to align with the new standards to match PCGS' new results. Or I may not update and make a disclaimer that results align more closely with PCGS before their update. It'll depend on what collectors ask for. I am always polite to our new robot overlords
Good luck to you on that PCGS thing, and your endeavors, but (there is always a "but"), you may want to research a bit, as to what is occurring in the world of pricing and evaluation FOR and on BEHALF OF the collector.
A computer program, any computer program, is only as good as the person that writes the code for that program.
No, because I subscribed to the school of thought that if a person is introducing a better mousetrap/methodology to an industry/entity, and particularly the evaluation of same, the person should have researched and be well versed in the subject and know all of the market aspects and the methods being presently used, and more appropriate to your particular endeavor...knowing what the public really wants, not what you may perceive it wants. I completely understand you are pursuing making a market and any entrepreneur doing so is his own best sales person, as it should be. You appear to have a gift for answering those with negative or doubtful thoughts, with erudite phrasing and positive self-endorsement. What is missing is your in-depth understanding of the Market and the collecting community: the hobbyists, the dealers, the Graders, the basis of pricing AND the participants that set the present level of the hobby, and indeed the worldwide scope of the hobby, and the relatively quiet sub-groups that have a significant impact. It isn't the entity you think it is. Yes, yes, I expect you will dismiss any observation that you are attempting to make a market. That is always the Play Book. I have observed many individuals and entities that have tried over many years and the road signs of making a market never vary. I do wish you well, and you may get there. However, as of now, you are not well versed. The simple aspect that you request me to explain what I mean is support of my observations. Truth in editing: changed a "t" to a "y". It happens.
There will always be naysayers in anything anyone does, I for one think this is a super cool concept. If for no other reason than the average joe to play around with his probably not that valuable collection and just see what grades it comes up with. Not everything about coin collecting has to be well is it worth this or that. Maybe you just want to compare 2 of the same coin you have to see which is better and you suck at grading (I kinda suck at grading even after years of collecting). Just my .02 on it. If I had a GPT4+ subscription I'd give this a whirl cause I love playing around with stuff and anything that ignites some new interest in collecting is a win in my mind.