Numista values - is this really working?

With some trepidation, I want to raise the topic of Numista values again. Naturally, I am appreciative of the work done by the Numista team although I feel sad that I am compelled to say so explicitly in order to avoid remonstrations from other members.

Here are the values, in South African Rands, for the coin at https://en.numista.com/catalogue/pieces883.html. It is a very common coin with a Numista Rarity Index of 5, so a fair number of users have entered values.



These values are not really helpful because they are contradictory. Coins in VG condition are valued higher than coins in VF condition for one year, and the other shows coins in F condition as more valuable than those in XF. I think that because people grade coins subjectively, and buy coins in vastly different markets around the world, the Numista values are not a reliable field guide to the approximate value of a coin.

The Numista Rarity Index is often dismissed as not useful when people ask about it, although I find it a very useful guide to the general availability of types. But these values are also not what we had hoped, for exactly the same reasons that are used to dismiss the NRI.

I know Xavier put a lot of work into this feature and, honestly, people don't seem to dislike it as much as I thought they would. Even I have switched to the new layout as my default in the hope of contributing useful information when I record the values of the coins I buy. I'm very disappointed that the results are so irregular.

I don't want to ruffle feathers by suggesting a solution, but I would like to ask some questions for the community's consideration:

1. When we see skewed values like in the example I've given - the sort of example that is now commonplace on the site - is the value feature really worth the cost of the additional complexity of entering coin data?

2. Are these values expected to even out over time and, if so, over what time period can we expect to see reasonable values for different grades of common coins?

3. Is it possible for the administrators to pull a report indicating what percentage of users have selected the old vs. the new interface? Also, is it possible to pull this report for Numista users from before the changeover, because some new users might not even know that the old interface still exists. (I don't expect the data to be shared with me - I simply suppose that this might be a good indicator to the team of the popularity of the new feature.)

4. Is there a possibility of smoothing out the values by segmenting differently, or by weighting more definitive values entered by administrators or referees? Would this be something referees should be tasked with? (One idea here might be to generate a value for each coin from all values entered by users, irrespective of grade, and then algorithmically adjust that value for each grade based on parameters like the spread between maximum and minimum values, and a logarithmic adjustment to the average value for each grade. This would give us values for more coins than we have at the moment, and prevent this sort of contradictory grade value.)

5. Is it possible to provide a one-click mechanism to switch between the old and new views to allow rapid access to both mechanisms of adding coins?

These are the questions that come to my mind when I consider ways that Numista can move forwards, rather than just fall back to old features. They are far from a coherent solution to the problem and I leave them here for quiet consideration rather than as actual requests. I hope, though, that we can all agree that this is actually a problem and that the values we see in this example don't do Numista, or the community, any favours.

Honestly, I don't enjoy Numista as much since I've switched to the new interface. It is less readable than the old one and my heart drops a little each time I open a coin page and see nonsensical values, and the multiple line display. I realise that it works for the way that some people collect but it doesn't work for me. I'm very worried that the old interface will be discontinued at some point, and I'm also concerned since it was said somewhere on the forum that no new work would be done on the old interface. For some of us, the old interface was the thing that made Numista the best choice for coin collectors on the Internet.

Please speak up if you agree with me. Please speak up if you don't agree with me. I'm not saying that one interface is better than the other for everybody because I can only speak about how I collect, and how I prefer to record my coins. I think, however, that some honest feedback might be useful to Xavier, to the team, and to the community.

A numista rarity of 5 doesn't necessarily mean that lots of people have added the values. Lots of older, now non-active members may have entered the coin before values were implemented and then not gone back in to enter the values. It will even out eventually.

Perhaps a temporary fix would be to discount the highest and lowest values entered and take the average of the rest. This would knock out any extreme prices that may be driving up or bringing down the average.

Απόσπασμα: "neilithicman"​A numista rarity of 5 doesn't necessarily mean that lots of people have added the values. Lots of older, now non-active members may have entered the coin before values were implemented and then not gone back in to enter the values. It will even out eventually.

​Yes, I understand that an NRI of 5 doesn't mean that lots of people have added values. Primarily, the fact that there is a value for each grade means that more people have added values for this coin than for many others, and that is vaguely related to the NRI.

I'm not confident that this will even out eventually. Aside from the problem of overvaluing coins in their local markets, it is a major problem when Numista ever shows a VG coin with a higher value than the equivalent F or VF coin.

I actually edited my post after your message arrived and included a suggestion to address this. This involves aggregating the values of each coin irrespective of grade, and then apply a logarithmic grade adjustment to the average value algorithmically. The curve of this adjustment could be based on the spread between the highest and lowest value. That is just the approach I would take to avoid the problem of contradictory grade values, and it still doesn't give accurate market values, but it might be a useful starting point for a more thorough approach.

The values are supposed to reflect what people are paying for the coins, Jarcek has said in another thread about values that there is a new feature coming that hasn't been implemented yet that will make it so lower grades can't be a higher price than higher grades

https://en.numista.com/forum/topic75026.html

Απόσπασμα: "neilithicman"​The values are supposed to reflect what people are paying for the coins, Jarcek has said in another thread about values that there is a new feature coming that hasn't been implemented yet that will make it so lower grades can't be a higher price than higher grades
​
​https://en.numista.com/forum/topic75026.html
​
​

​It makes sense to me that the numbers are reflecting what people are paying for coins. I imagine that there are plenty of people who "pay more" for certain types of coins because they are in a remote area, in a part of the world that doesn't have as easy access to another countries coin, etc.

If I am in the United States I can easily collect 1 cent coins from circulation in good condition. If I am on the other side of the world, you are paying a premium for the supply chain. I see some Russian collectors comment that I pay too much for Soviet coins at times, because they are "common coins." They maybe common in Russia but not here.

I suspect that after a decade or more there will be enough data to come up with an algorithm to smooth the data into what people are actually paying as market value. You can't do the analysis until the data exists though.

Your example looks really bad, but the truth is there isn't much of a difference between those values! And they are actually pretty accurate if you just take the lowest value at the lowest grade and the highest at the highest;
About €0.50-€2.00 or $0.50-$2.00, because the South African Rand isn't worth much...
When I say accurate, I mean in relation to NGC and the standard world coin catalog

Απόσπασμα: "jadejackal"​​It makes sense to me that the numbers are reflecting what people are paying for coins. I imagine that there are plenty of people who "pay more" for certain types of coins because they are in a remote area, in a part of the world that doesn't have as easy access to another countries coin, etc.

​Yes I suspect that is probably what has happened in some cases. For instance, I can easily get the lower grade NZ silver pre-decimal coins over here for melt value sometimes, or at least close to melt value. But all it would take is for one person who doesn't have access to NZ coins very often to pay an exorbitant price for a VG coin and it would skew the catalogue price of that year well above the price for a coin in F grade.

Well, probably not just one person; I think the extreme values on either end aren't counted, so it would take 2 or more such persons for the values to start getting skewed. Don't quote me on that, though.

Απόσπασμα: "King"​Your example looks really bad, but the truth is there isn't much of a difference between those values! And they are actually pretty accurate if you just take the lowest value at the lowest grade and the highest at the highest;
​About €0.50-€2.00 or $0.50-$2.00, because the South African Rand isn't worth much...
​When I say accurate, I mean in relation to NGC and the standard world coin catalog

​Well, for the 1948 coin, the VG value is 201.66% higher than the F value. Because these numbers are small, it doesn't mean that the values are not heavily skewed.

Any way how I could send a post that keeps blanking out in the edit window?

I have the text saved, but so far hadn't been able to send it.

Hi,

I'll never give values into numista, numista is not made for that, so please forget about it. Believe your own data, but not any statistics derived from them...... if you didn't make them yoursef!

Ole

Honestly, I think it does work.

This has been suggested by me and implemented by Xavier in the following manner:
https://en.numista.com/forum/topic74862.html

Principally:
1) the MEDIAN, not the average of the values are calculated for a Numista value, which I believe is a great solution.
2) You do not need to enter a matching value, you can enter your own value and trust your OWN VALUE... eventually it will regress to the average, if enough people add values.
3) I agree with above, that the Numista value is not my guidance, but I still enter my own, regardless. Hoping that it will eventually change the public opinion... if not, I agree with Sjoelund, I stick with my own.

In principle, it is a guidance, not a must!

I am still unable to post the comment that I intended to; but TL/DR is that I have found a logical explanation for why skewed values like the OP example occur.

I don't know what is the reason for its inability to be posted, however, so just to be safe, I'm only going to link to the version I posted four weeks ago on Twitter.

Απόσπασμα: "January First-of-May"​I am still unable to post the comment that I intended to; but TL/DR is that I have found a logical explanation for why skewed values like the OP example occur.

​If your explanation is correct (that is, if lower grades data comes from foreigners and higher grades data from locals), then you can use the values displayed by Numista as a price guide. For example, if you live outside South Africa, then for South African coins you should look only at the values displayed for lower grades and ignore other values.
But I do not know whether your hypothesis about coin buyers and grades is correct, and how to verify it.

Απόσπασμα: "numinis"​

Απόσπασμα: "January First-of-May"​I am still unable to post the comment that I intended to; but TL/DR is that I have found a logical explanation for why skewed values like the OP example occur.

​​If your explanation is correct (that is, if lower grades data comes from foreigners and higher grades data from locals), then you can use the values displayed by Numista as a price guide. For example, if you live outside South Africa, then for South African coins you should look only at the values displayed for lower grades and ignore other values.
​But I do not know whether your hypothesis about coin buyers and grades is correct, and how to verify it.

​I think I should clarify that this is intended for common coins - i.e. where locals don't really care about lower-grade examples because it's routine bargain bin stuff for them, and if they do buy any, don't really remember exactly how few cents they paid (and/or buy them as part of a lot, making the individual coin's price meaningless).

For any coins scarce enough that the locals would care about lower-grade examples (and/or for any coins that circulated little enough that they hardly ever show up below XF/AU/UNC in the first place - in particular, any non-circulating types), the relationship breaks down, resulting in values that correspond neither to local prices nor to foreign prices; admittedly, this is probably the system working as intended, and most coins that scarce probably don't yet have enough entered prices for a displayed value anyway.

(And of course for any coins that have no or very few local collectors on Numista - this would include, for example, nearly everything from Africa outside the RSA - there was no relationship in the first place; the values being shown are always the foreign prices, because the local prices, if they even exist, are entirely irrelevant.)

I think I should also add that one more exception is euro coins - enough people actively collect euro coins from circulation that the value of any dates and types that aren't especially uncommon in circulation is going to be their face value (= local price), even in lower grades. (Though obviously not in AU and UNC.)

I suspect (though have not checked) that the same is probably true for state quarters, and any other series that is being actively collected from circulation; my conclusion should still hold for common coins that aren't actively collected from circulation, e.g. due to having been demonetized (which is the case for the example in the OP).