Using an Algorithm

Discussion in 'Coin Chat' started by Endeavor, Jan 10, 2014.

  1. thecoinlover

    thecoinlover Banned

    Oh, my bad. There is a saying that is kind of like that phrase that is a paradox.

    And to the OP, that sounds like that could be a really long process. However, I like coding, so I say go for it. :)
    Last edited: Jan 10, 2014
  2. Avatar

    Guest User Guest

    to hide this ad.
  3. rysherms

    rysherms Alpha Member

    to the OP - what is your mathematics and/or programming background/education?

    GDJMSP Numismatist Moderator

    I'd tell ya not to waste your time trying. You can easily determine the value of just about any coin, with the exception of a few absolute rarities, in less than 5 minutes. Assuming of course that you have the knowledge and experience to judge the quality of the coin and know where to look.
  5. Endeavor

    Endeavor Well-Known Member

    None really. I don't have coding or math education beyond some college. I do know excel and spreadsheet formulas. I taught myself.
  6. longnine009

    longnine009 Most Exalted Excellency Supporter

    "Well acually" -1 means that it got better with age and is better now than it was when it was struck-aka the very rare albino blacksheep error. And BTW, that 48.1 percent stat is a fabrication. All coin collectors are fabricators.
    Last edited: Jan 11, 2014
  7. Numis-addict

    Numis-addict Addicted to coins

    I forgot to carry the one:(
    Endeavor likes this.
  8. Numis-addict

    Numis-addict Addicted to coins

    I agree that this is a massive undertaking. My recommendation would be to take one series of coins, select about 10 dates and mint marks for it, and closely study past sales on the Ebay and Heritage sites, along with current ones as they come in. Then, attempt to create a formula that will work for these ten coins. If you can do that, then you could expand to a whole series, or attempt to create a formula that will work for a handful of coins from a couple of series. One important variable that you left out is hype. If a coin is being sold on HSN, how will you account for the hype applied to it? Or will your equation only be designed only for auction sales? Either way, the chance of you being able to accurately produce a formula for even a single series of coins is rather low. Sheldon tried it, and that didn't work out too well for him. That said, good luck. It would be amazing to see such a system developed.
  9. longnine009

    longnine009 Most Exalted Excellency Supporter

    How do you account for gradeflation in the pricing data?
  10. The current monetary value of any coin is simply the price that the next person is willing and able to pay for it. TC
    Endeavor likes this.
  11. Vegas Vic

    Vegas Vic Undermedicated psychiatric patient

    [quote=" Either way, the chance of you being able to accurately produce a formula for even a single series of coins is rather low. Sheldon tried it, and that didn't work out too well for him. That said, good luck. It would be amazing to see such a system developed.[/quote]

    It might be possible to make something for a single year specific mm but the real variable is demand is different even in a single series. Maybe something could be done for a specific coin but I don't think it would hold for a series for this reason. I am interested to see if you can pull this off. Keep us in the loop on your results please
    Endeavor likes this.
  12. rysherms

    rysherms Alpha Member

    No offense, but I would not invest any money or take a second mortgage out to buy coins once you complete this.

    Honestly I am trying, and it sounds like others on this site are too, to save you from a lot of work (of which the scope of I don't think you are aware of) and possible financial loss. I think it would be a safe bet that some very highly educated numismatists with mathematics and programming backgrounds have probably tried this and not had amazing success.

    Focus on collecting coins you love. Don't go into this trying to make money. Most fail.
    longnine009 likes this.
  13. Phil Ham

    Phil Ham Hamster

    I've created simple algorithm using Excel. One tab includes the updated melt value of US coins, US proof and mint sets, and Canadian coins using Coinflation to update the values. I link these prices to another tab, which includes all the coins in my collection. One column includes face value, another column is instrinsic value, another column is the upcharge or collectors value over instrinsic value (if any). Another column adds the collectors upcharge to the instrinsic value and compares it to the face value. It reports the highest of the two values using the IF function. I use ebay (sorted by sold items) to update the collector's upcharge.
  14. beef1020

    beef1020 Junior Member

    Someone else pointed it out earlier, but I will reiterate and add detail. The Sheldon grading system we use today was originally designed to work as a price system. Each variety had a 'basal value' which was the price for that coin in a grade of 1. So if the coin had a basal value of $1 and graded 12 then it was worth $12. If the coin had a grade of 40 then it was worth $40. This quickly fell apart as a useful tool to estimate the value of the coin, although we still use the underlying grading scheme.

    The major fault of the system was assuming a linear relation between grade and price, i.e. if a vf20 is worth $100 then a xf40 should be worth $200. The dynamics of the coin market are such that to the extent there exists a relationship it's exponential not linear.

    One concrete piece of advice I would offer is to choose a series that was not hoarded, so morgans are out, and look at auction records of the one year at varies grade levels. For instance, go to heritage and find out how much PCGS graded 1914 Lincoln cent go for in different grades, f12, vf20, xf40, au50, 64, 65, 66, 67. Your levels don't need to be too fine. Take the average of the last 3 coins or so which sold and plot those sale prices against the grade. You'll see the exponential nature of the curve, do a best fit and find an equation that does a reasonable job matching the data.

    Once you have this, you can test it against some other coins you are interested in. You will now have an equation which describes the curve of the price by grade scale and you will have an initial condition which you will need for each series, i.e. low grade indian head nickels sell for less than low grade bust halfs, but as long as the series or coin was not hoarded or otherwise had it's survival numbers messed with, the same equation describing the price to grade curve should be reasonable.
  15. Conder101

    Conder101 Numismatist

    I think the OP's scheme would work.......for widgets. Common always available in quantity average grade material. For anything else no.

    And Sheldon's "equation" wasn't completely linear either. It merely described the relationship between values and grades that existed at the time, and which had existed for some time. Where he went wrong was in not allowing for the fact that as the number of collectors increased and as they became more affluent, the demand would not increase evenly on all grades but would increase faster on the higher grades. So the nice neat original ratios would fall apart.
    Endeavor likes this.
  16. Endeavor

    Endeavor Well-Known Member

    I like this post.
  17. Endeavor

    Endeavor Well-Known Member

    That's a good question. The algo will have to account for changes in demand due to changes in everything (grading being one thing). I need to think about it more before I develop the algo.
Draft saved Draft deleted

Share This Page