Expression of value for bullion coins

Hi and firstly thanks for an impressive and functional website.

Bullion prices have been on an excursion for a while now, which makes it difficult to evaluate a bullion coin's value over it's melt value. An auction value from a few weeks ago may not even be representative, and I have already spotted a few instances where the Numista estimated value (below AU grading) is less than melt value. In a downward bullion market trend, such anomalies won't be that obvious to spot.

I suggest that value estimates for bullion coins are expressed as [current bullion value + numistic value]. This is probably not as simple as saying it, and as I imagine it, what would be needed is;

• A database table recording/backfilling daily bullion prices
• A database table recording/backfilling daily forex rates
• Queries hinged to the above that estimate value by bullion value at date of sale, thus accurately isolating numistic value.

Maybe if a user enters a purchase date and price for their coin, that price remains private so that Numista can benefit from the data, but the user can seperately enter their desired premium over bullion value for swap/sell public display purposes. Eg. I bought a coin lot at a bargain, and i want to leverage my adavantage. I don't want my trade partner to see my purchase price. I can attach a copy of my invoice for an admin to verify, which would nudge the estimated value of a coin down nominally. The whole system wins in this scenario. At the moment a user would need to use the public message field to actively and regularly update their swaps around the price of bullion for coins without estimates, and for all coins if faith in Numista estimations erodes.

Outside of that, it should be mostly frontend website tasks to implement. 

Isn't that how it is done now?

https://en.numista.com/forum/topic145285.html#p1155781

I hadn't seen that thread, so thanks for the link.

In my opinion, % over bullion spot is an inaccurate and volatile metric. Consider a non bullion coin where the metal value is negligible. It could be extremely valuable depending on scarcity and condition, thus it's numistic value is absolute. The numistic market, not the metals market determines this. I feel that this principle should translate to bullion coin value estimates as well.

____Edit addition

Consider a freshly minted gold 1oz coin bought in 2023 for $2040 @ $2000 spot. It's numistic value is $40. I sell it in 2025 for $4050 @ $4000 spot. It's numistic value has appreciated by $10 outside of fluctuations in the gold price. % over spot valuation however says that I bought at 2% over spot and sold at 1.25% over spot, when in reality my coin has appreciated in numistic value. Invert the bullion market direction, and suddenly bullion coins become numistically much more valuable.

___

I also feel that Numista is missing a trick by not harvesting audited user purchase data. The inputs here however would need to be strictly private. I am experiencing that many modern collectable bullion coins, especially ones rare to Western markets, simply have no auction records, so no value estimation.

Edit addition:

Also, given enough quality current data, the effects of inflation should naturally dissipate.

I also wanted to throw in another factor that was on my mind in the last month: Exchange rates

They're not as volatile as the bullion prices lately, but still a bit more volatile than normal I think.

Let's take USD/EUR for example. USD got weaker over the last year, so it makes a difference if old auction results or user values from one year ago are calculated with the exchange rate from back then or now. I don't know how it's done currently. Might be already factored in.

It was discussed in the thread that the previous poster linked to, with no firm conclusion. Bullion is measured in USD as a standard, so IMO it becomes about the date of sale and the absolute USD numistic value over bullion on that day. When this is in the database, how you view it in your currency today is a factor of the fx rate today (same as how you would view the bullion price). Numistic value is thus treated as a seperated concern. If USD appreciates/depreciates in value, FX markets should compensate, so real value stays the same. If the USD ceases to be the global reserve currency, this would need to be re-evaluated. 

Just thinking my previous post through a bit more carefully…

I buy a coin at bullion value plus €10 @ €1 to $1 FX parity. Over the year, USD depreciates to $0.99  to €1. When I view the numistic value of my coin today, it will be recorded in the database at $10 and thus display as €9.90, which has exposed me to dollar depreciation.

The database should probably capture sales in the currency of sale, so that everything can be translated relative to the USD in the present day, eg EUR → USD → GBP, depending on your perspective. My €10 is still €10, but an American would have to pay me $10.10 (things got more expensive for them because their currency devalued). How much a Brit would need to pay depends on how the GBP has performed against the USD. An American bought at $10, and now I only need to pay them only €9.90, but their $10 is still $10 from their perspective.

This may not be perfect, but is less complicated than currency arbitrage, and tracking the entire FX web in the Numista database. If data is coming from multiple markets, things should balance out.

Just to illustrate below that on a common coin to Europe, that my AU value estimate is falling below bullion melt value. For an 1882 coin in this condition, I would expect to see reasonable numistic value, so it says to me that the valuation system as it is ATM is not working, and I cannot trust it.

Bullion value = €675.59

AU = €670

F to XF = €650

citkane

Just to illustrate below that on a common coin to Europe, that my AU value estimate is falling below bullion melt value. For an 1882 coin in this condition, I would expect to see reasonable numistic value, so it says to me that the valuation system as it is ATM is not working, and I cannot trust it.

 

Bullion value = €675.59

AU = €670

F to XF = €650

 

 

 

One thing to consider is that I believe the bullion value is almost “live”, while the other values are updated only once per day. So on a day like today with kind of an upswing in bullion price, it is more apparent.

That may be partially true if the scale of mis-estimation was not so extreme. It may be a common 1882 coin, but not undesirable. I would hazard a guess of at least €20 to €50 over bullion value. Compare this to the 1889 gold 1 sovereign, similar age, mintage and appeal, yet today the system is estimating that at €58 over bullion value for an AU specimen.

I am not trying to be critical, but I am pointing out what is obvious to me from my experience of working in data and systems. These kinds of inconsistencies (and I have noticed more)  point to an underlying failure in the functioning of the system, and if left unchecked, it will erode the trust of the users in this feature. From what I understand in the thread previously linked, in my opinion, the logic of estimation for the system is fundamentally flawed. I will be happy to put in my own time and effort to help fix it as best I can.

Hello,
 

I verified the calculation for N#19764 1880 R in grade XF. Numista shows a value of 590 € whereas the bullion value of the coin is 655.61 €.

The price of 580 € is calculated from one auction results and three collections.

• The coin was sold at auction on 5 Apr 2023 for 331 €. At that time, the bullion value for that coin was 344 €.  So it was sold 13 € or 4% below the bullion value.
• One member got the coin in 2019 for a price that was 62 € or 29% below bullion value.
• Another member got the coin in 2020 for 28 € or 10% below bullion value.
• The last member bought the coin in 2025 for 142 € or 26% below bullion value.

Based on these values, Numista determined a value of 590 €, i.e. 10% below bullion value.

I don't think that changing the calculation based on the same data could significantly improve the value. We may want to use the bullion value as the minimum value, but I believe it would distort the reality: if a coin is consistently sold below its bullion value, we should probably reflect this fact in the value.

I believe we have some room for improvements in two ways:

• We recently enabled members to enter the purchase date. We should use this date when available, instead of the date when the value was entered into the collection.
• Even if the hammer price for the auction above was 13 € / 4% below the bullion value, the buyer had to pay some fee to the auction house. The current buyer's fee at Katz auctions is 24.2%; assuming it was the same in 2023, the buyer actually paid 411 €, which was 66 € or 19% above bullion value. For most auctions, our database doesn't know the buyer's fee, but we could maybe make an assumption and add something like 20% to all the auction results that don't include the buyer's fee. This would make make the value less accurate for sellers, but more accurate for buyers.

----

Regarding the request in the third message of this thread, i.e. considering a fixed premium over bullion value (which captures the numismatic value) instead of percentage over the bullion value:

We currently apply the logic to calculate a percentage over bullion value only when the premium is quite small. The idea is to help providing an accurate price for coins that are traded as bullion investment with a low numismatic value. A percentage makes sense for those coins because the dealer spread typically moves as a percentage of spot. For semi-numismatic and numismatic coins, the premium is generally higher because of the numismatic value, and this value doesn't depend on the spot. A fixed premium also makes sense for low value coins (small silver coins, tiny gold bars), where the handling fee matters more than the variation of bullion value.

A good example can be found in the US mint reseller program: the US mint charges a percentage premium for gold bullion (from 3% for 1 oz gold Eagle to 9% for 1/10 oz) and a fixed fee for silver bullion (3.05 $ per coin)

We could consider adjusting the algorithm so that: if the premium is low (for example less than 20%), then calculate the premium as a percentage; if the premium is high, consider that the premium is due to numismatic value and calculate it as a fixed amount. Or maybe a mix of both models.

Regarding that last suggestion, I think it makes sense to differentiate that between the grades as well. There are a lot of 100+ year old coins without a lot of premium, but when they're in UNC condition, they go for a LOT of premium (especially when they're in slabs).

To stay with the Umberto example and Trooper's observation, PCGS reports that this coin is underappreciated and can often sell at melt. I however have a AU - UNC specimen for swap, and if someone wanted to buy it retail, they would definitely pay a premium if they could get it at all. I don't believe that there are any coins over 100 years old in that condition that will have no numistic value. This leaves me at a disadvantage in trading.

For another example, I intermittently buy investment bullion in bulk. With this I will include doubles of interesting and collectible silver coins. For modern UNC silver coins, the premium can vary significantly. For example, a 1oz Chinese panda of 10million mintage may retail at €15 over bullion, but a proof commemorative of 10thousand mintage may retail at €75 over bullion. Data and accurate estimation for modern bullion is most often missing or inaccurate on Numista, and a % over bullion estimation would have a dramatically different effect on these two coins.

I think that there are so many variables and ways to look at the market by different stakeholders that this needs a big detailed think. Also to consider is the technical aspects of caching and server workload if complexity increases. My gut feeling is that all is resolveable in an elegant way, so I will start work on a draft technical spec proposal for discussion.

Maybe to cherry pick just one point from Xavier; it is good that users can now input date of purchase and price of purchase. This is valuable data that can stabalise price estimation, especially for newer coins. I am however reticent to give the data because the purchase price is publicly displayed, and that can disadvantage me in a swap. If I got a bargain, it can move the overall estimation down a small bit, but I don't want my advantage for a specific coin to be seen by my swap partner.

As I mentioned, I have started a “living document” specification for valuation methodology:

https://docs.google.com/document/d/1LQy7-Ky4SLIu74YwYvoS5ASDYpONRBOIsOHyu6NulCA/edit?usp=sharing

Anybody can add notes, but if you want to edit it please PM your email and I will invite you as an editor.

@Xavier, please excuse my presumptuousness with this, it is just the way my psyche ticks. I can see and understand the beauty of what is already done with Numista, and would love to see an improved valuation system.