MORE ON QUANTITATIVE COLLECTORS GROUP

Bill Eckberg writes:

First let me thank you for your thoughtful and engaging questions. Hopefully, we can shed some light on your issues.

In hopes of trying to add some continuity to my answers, I will address your second question first.

It was also very odd that none of the people involved in creating the system are coin collectors. All of their claimed expertise is in manufacturing, software development and optical technology.

This is very true, as myself and my two partners are not what people would call your traditional coin collector. However, the company was started based on the frustration I experienced when trying to find the value of a 1929 quarter Indian head that was given to me back in 1976. And I started to ask the questions, who can I trust to evaluate my coin? How can I maximize the value of my coin if I want to sell it? How are Sheldon grades determined? What is TPG? I must send my coin where for independent TPG? How long does TPG take? What will TPG cost me? What kind of transparency do I get when my coin is graded by TPG? As you can imagine, it was the answers or non-answers to these questions, combined with my 35 plus years' experience in the world of inspection, while reading ANA grading standards and Jim Halperin's N.C.I grading guide, that led us to the development of the OSC200 CoinGradr. What I can tell you from our collective industrial experience of using computer vision to solve complex inspection problems for some of the largest companies in the world. It's always better to have people from outside of your passion or industry with technical expertise that can dissect the problem without bias as they apply realistic and scientific solutions to some of those problems.

I found it especially curious that neither the press release nor the company's website, nor the video interview with Richard Colonna of QCG ever mentioned anything about the device's accuracy or consistency in grading.

In the world of scientific instrumentation repeatability is the goal we strive for, just ask Google Gemini AI, "repeatability is considered more important than accuracy because consistent, repeatable measurement can be calibrated to correct for systematic errors. If a measurement is not repeatable, it cannot be trusted, regardless of how close a single attempt is to the true value." In other words, anybody can be accurate once, but can they repeat the process? As for the OCS200 CoinGradr, in our light tight enclosure, with illumination feedback, using four different light sources, precision coin placement and calibration we can achieve over 95% repeatability of measurement results from one machine to another.

As for accuracy, we are currently conducting a validation study comparing quantitative coin grading as it correlates to subjective coin grading using the top TPG houses in the United States. What we can tell you from our preliminary results using our empirical data approach is that we stack up very well in terms of cross-over and repeatability percentages versus other TPG houses. Preliminary results of the study can be viewed on our Indiegogo Campaign site that will be launched this month which we will provide when the link to the site is available.

It seems to me that if the goal of your product is to analyze coins, you ought to have someone in the company who knows something about coins.

I'm not sure "knowing nothing about coins" is how I would describe the talent within QCG. In fact, I would go as far as to say we are some of the top experts in the world as it relates to reflections on different types of metallurgy and topography used when minting coins. I would also point out our vast knowledge of the type of lighting to use to extract certain attributes on a coin such as luster, wear, toning, balance, etc. In terms of our expertise in the field of coin grading, we can only say we were awarded a patent on our approach by the US government for coin grading. I would also point out, one needs to remember that quantitative coin grading does not tell you how, why, or where something happened to a coin, but rather measure the attributes that make up the Sheldon Grade and weigh those quantitative attributes using a methodology accepted by the profession. As for experts in the field of Numismatics, we are constantly engaging with and seeking advice from some of the top voices in the field.

So, a few minor questions remain:

Does it give market-acceptable grades?

I'm not sure we are the ones to answer this question. If you're asking, "would quantitative grading be considered technical grading or marketplace grading," I would say technical grading with a twist, as we do evaluate mint state coin criteria and weigh those attributes based on Jim Halperin's N.C.I. Grading guide and we also grade all those attributes on circulated coins. I guess another way to answer the question, is that I have heard it said that any coin can have 2 or three grades in it. It is our belief that the reason for this sentiment is weighed in the fact that beauty is in the eye of the beholder, in that one person might put more emphasis or bias toward luster than eye-appeal, thereby giving that coin a higher or lower grade. It is the goal of quantitative coin grading to measure those types of attributes with specificity to help guide the users in a transparent way so they can make a more educated decision based on the coin they are evaluating.

Does it do so consistently?

Much better than humans.

On what denominations and types has the software been trained?

Even though computer vision is considered a subset of Artificial Intelligence, we do not use machine learning to evaluate and grade coins, as coin grading is way too nuanced and complex for that type of approach. Instead, we use empirical data as the backbone of our databases, which just means we measure verified subjective grades evaluated by TPG experts on our device for any type of coin series and type. We then add statistics to that database and compare the images taken by the OCS200 after running specific algorithms for each attribute we are measuring. The results are then compiled, weighed, and documented to the user in 15 minutes.

When we took on this project, our goal is to grade all business strike coin series, type, and variety since the 1792-coin act. Initially the machine will ship grading Lincoln Wheat Cents, Buffalo Nickels, Mercury Dimes, Washington Quarters, and Morgan Silver Dollars, as new coins will be added every month.

How do the machine grades compare to grades assigned by advanced numismatists or third-party grading firms?

We are currently doing a validation study, as mentioned we have some preliminary results and insights which can be viewed on our Indiegogo Campaign, which can be accessed later this month. If you would like to learn about our technology, we would encourage you to visit this site, as we will provide a link when it is released to us.

Paul Hybert writes:

I would like to thank your readers for their insight and comments. Actually, the words used in your readers' comments are very important to us, as we will try to parse the facts.

First …. The article in The E-Sylum is an edited version of an article written in the SME Business Review, which was an edited version of our original press release, which you can access at the original press release link. As you are aware we cannot control the editions of other publications.

As for the facts, in our press release you will notice that we use the word quantitative nine times, which we further embrace in our name. We use the word subjective three times, and we use the word objective one time.

As for definition according to Webster's Dictionary ….

• Quantitative …. is related to measuring something, rather than its quality.
• Subjective …. Based on or influenced by personal feelings, tastes, or opinions.
• Objective … not influenced by personal feelings or opinions in considering or representing the facts.

As for what we are touting is the fact that we have built the world's first "quantitative coin grading microscope", which means we created a scientific instrument to measure and analyze things using N.C.I. Grading Guide methodology like wear, luster, toning, strike, etc., for the world of coin collecting. To get a better understanding of this technology we encourage you to visit our website at www.qcgcoins.com, or our Indiegogo Campaign, which will be starting in the next week or so.

As for context of how we used quantitative, subjective, and objective, we will refer to a couple of paragraphs in our original press release.

Interpretation: QCG is not using subjective data in our process

Interpretation: most accurate grading results are when you combine quantitative data and subjective data.

By combining objective data (QCG quantitative) with traditional evaluation (TPG subjective), collectors experience a more balanced, transparent, and trustworthy grading experience.

And, finally I will leave you with a question I asked Google AI:

What is the difference between quantitative data and objective data?

 

To read the original press release, see:
Quantitative Collectors Group (QCG): Revolutionizing the Way Collectors Assess Coin Value (https://drive.google.com/file/d/1kKdTJetD7n-7S3MryhjY19W5V9GoCcwm/view?usp=sharing)

To read the earlier E-Sylum article, see:
DIGITAL COIN GRADING MICROSCOPE LAUNCHES (https://www.coinbooks.org/v29/esylum_v29n05a08.html)

	NBS (coinbooks.org) Web