How Good is CoinTalk at Grading?

Discussion in 'US Coins Forum' started by physics-fan3.14, Jul 27, 2019.

  1. physics-fan3.14

    physics-fan3.14 You got any more of them.... prooflikes?

    Thank you very much. As a grader and collector who's opinion I respect very highly, this is extremely nice to hear.
     
    Lehigh96 likes this.
  2. Avatar

    Guest User Guest



    to hide this ad.
  3. messydesk

    messydesk Well-Known Member

    One other thing that comes to mind, is that this would make for an interesting article in The Numismatist, once the last coin is posted for the first experiment, especially since it represents original work on a facet of popular subject not yet rigorously explored.
     
    Lehigh96 likes this.
  4. Lehigh96

    Lehigh96 Toning Enthusiast

    Agreed
     
  5. GDJMSP

    GDJMSP Numismatist Moderator

    Regarding your 2nd paragraph in your response, and this is just a suggestion. By doing this - I will attempt to pick coins which I think are accurately graded, - my guess is you'll unintentionally skew the results of your efforts.

    I believe picking the coins at random would be a much better option and provide much more honest and accurate results.
     
    TypeCoin971793 and Lehigh96 like this.
  6. Lehigh96

    Lehigh96 Toning Enthusiast

    I completely agree with Doug. Someone mark down the date.
     
  7. GDJMSP

    GDJMSP Numismatist Moderator

    LOL ! C'mon now Paul, you and I have agreed on a great many things over the years, and on a pretty regular basis.

    But yeah, I get what ya mean :)
     
    Lehigh96 likes this.
  8. ddddd

    ddddd Member

    7/29/2019
    :D
     
  9. messydesk

    messydesk Well-Known Member

    The purpose of the test is to see if a group can arrive at the same conclusion as a TPG. If this test is only going to be done for 25 coins, then the coins need to be selected to be representative of a typical grading event, with outliers or grading anomalies excluded. If a random selection of a coin shows one that looks way off, it should be rejected, as the results of the study will show that for that coin, the hive mind and TPG didn't match, but not give insight as to why. If we were using 2500 coins, we might find that the number of coins for which this happened told us something about TPG tendencies (or Heritage's photography department), but that won't be possible with 25, nor is it the intended purpose of the study. Perhaps rather than stating that he'd be picking coins which he though were accurately graded, he should have said that coins that seemed inaccurately graded or had uncharacteristically poor photos would be excluded.
     
    Paul M. and Seattlite86 like this.
  10. GDJMSP

    GDJMSP Numismatist Moderator

    I could easily agree with automatically excluding coins with poor pics. But by picking coins, for any reason, is going to skew the results because it will bias the test. One way or another, it will bias the test.
     
  11. Lehigh96

    Lehigh96 Toning Enthusiast

    It begs the question, is the TPGs “screwing the pooch” 1 out of 25 times a typical grading event? A sample this small is bound to have a higher deviation, that doesn’t mean you should introduce bias to reduce it.
     
  12. messydesk

    messydesk Well-Known Member

    A typical grading event is a coin that is graded more or less correctly. If there is a 50% chance that a group of 25 randomly selected coins will have one that is way off, then there is a 2.7% chance of an atypical grading event for any given coin.

    For this study, we're going by photos, so we have an incomplete picture. If Jason grabs a random coin and sees that it doesn't make any sense, he won't know if that's on Heritage photo people, the TPG graders, or a "mechanical error." What's to gain by introducing it into the study? It would seem better (necessary, actually) to disclose the selection methodology, including reasons for excluding a coin and how many randomly selected coins were rejected.
     
  13. Lehigh96

    Lehigh96 Toning Enthusiast

    So you are saying that any photo that doesn’t mesh with the assigned grade may not be accurately depicting the appearance of the coin and should therefore be excluded. While that seems more reasonable than excluding coins that are definitive outliers, I still don’t think it is necessary. I think he should just pick 25 random examples and live with the results.
     
  14. Santinidollar

    Santinidollar Supporter! Supporter

  15. messydesk

    messydesk Well-Known Member

    Yes, that's what I'm saying. And while it isn't really all that necessary, I'd rather have 25 good data points than 24 and one suspicious one. At the end of the day, however, I'd be surprised if 25 randomly chosen coins had more than one such coin in it.
     
  16. LuxUnit

    LuxUnit Well-Known Member

    Very good read!
     
  17. C-B-D

    C-B-D Well-Known Member

    Next start a thread called, "How bad is CoinTalk at handling being wrong?"

    :p I kid, I kid.
     
    Paul M., CircCam and messydesk like this.
  18. messydesk

    messydesk Well-Known Member

    Another thing we may observe is that The Hive converges on TPG grades with increasing accuracy as this experiment goes on.
     
    TypeCoin971793 and Paul M. like this.
  19. micbraun

    micbraun coindiccted

    No need for such a thread, we’re always right.
     
    RonSanderson and C-B-D like this.
  20. GDJMSP

    GDJMSP Numismatist Moderator

    The TPGs have already proved that, several times ;)
     
  21. Eric Babula

    Eric Babula Active Member

    I am not a statistics junkie, but this is an excellent thread! And, I thank you for your efforts on this subject. Very nice work! Great reading!

    One thing you mention is that we should not only post our grade guess, but an explanation of why we came to our conclusion. Then, you also mention that we get better samples and more accurate results if the GTG thread is a poll. The fear I have with the polls is that more people might be inclined to just click the button of their choice, and move on, and we will have even less people give detailed explanations for their grades. If that is the case, we who read these GTG threads will have less analysis to read and learn from. What are your thoughts about that? Maybe that's another study that someone needs to undertake! LOL!
     
Draft saved Draft deleted

Share This Page