Number of people who own each year of the coins

I would like to see for all coins / tokens / medals the number of people who own each year.
Referees can already see that for their Referee countries so why not give this opportunity for everyone? It should be a quick fix for Xavier to do as referees can see it. Probably only one little setting to change. It would make the rarity number much more meaningful. This was brought up before but seems to have been forgotten.

I think this should be a relatively easy upgrade which should be given high priority.

Ernie

Numista needed this feature five years ago. With the values system, it is even more important.

Definitely supported, how did I not see this when you first posted it?

Looking more long-term, I think the Numista Rarity Index number could be abolished altogether (as I've heard some folks say), or at least we could make the NRI a thing for each year line. If I have a proof issue of only a few thousand mintage, I don't want it to be lumped in with the extremely common business strike with millions of examples.

Since we are removing the rare comment from coins, we need this feature even more.
Like the suggestion so we can get this feature implemented.

Hello,
I agree it would be nice to have an indication of how many people own each year, to evaluate which years are more or less common.

Thank you Xavier
This feature will be a huge help for everyone to figure out which coins are the rarest.

Have we implemented this?

Bueller?


Is this improvement in effect, but I cannot see it?

I do believe we are owed an answer to the question whether this has been implemented or not. Or at least an approximation of when it will be implemented.

Ernie

It was not implemented yet. As for estimated date, that is a question for Xavier.

…perhaps as a Christmas gift for all? (No answer from Xavier after 30 days)

Hello and Merry Christmas!

There is now an indicator of the frequency of each year/mint/variety to help identifying which years are more or less common. It shows the percentage of users who own each year or variety among all the users who own the coin. Note that, since some users own several versions, the sum may be greater than 100%.
See for instance: https://en.numista.com/catalogue/pieces6834.html

I don't think it's meaningful to display the exact number. The rarity index and this new frequency indicator already give an idea.
Frequencies are displayed only if there are enough users.
I hope you like it

Yes, this is a great addition. Unfortunately, the total should be around 100% because of rounding. If someone has two or more coins of a date, just include them in the total for that date.Statistics are easier to absorb when using a base of 100. Do you think the British would want to go back to 12 pennies to a shilling?

Ernie

Quote: "erniemix"​Yes, this is a great addition. Unfortunately, the total should be around 100% because of rounding. If someone has two or more coins of a date, just include them in the total for that date.Statistics are easier to absorb when using a base of 100.

​If you are asserting that the totals for the page should add up to 100%, they almost never will, and it is not the intent, based on Xavier's explanation. It is meant only to tell you "what percentage of people that own this type own this particular year." So there is a 100 base, but only in a year-to-type ratio. (This may be your point.)

So in a type with only years 1995, 1996 and 1997:
-If everybody owned and cataloged 1997, and nobody owned or cataloged 1996 and 1995, the percentage would be 1997-100%, 1996-0% and 1995-0%.
-If everyone owned 1996 and 1997, an nobody owned 1995: 1997-100%, 1996-100%, 1995-0%.

In my opinion, it is useful only for recognizing which years are more common than others where the mintage of each year is unknown. Example: https://en.numista.com/catalogue/pieces3777.html
Clearly 1989 is the most common coin cataloged on Numista, at 71% for that year.

It can also point out abnormalities: A high percentage on a low mint year, or a low percentage on a high mint year can indicate the mintage listed is in error, or coins were recalled.

I would like to see the total for each year, but it's not my decision, and Dad always told me, "Don't look a gift horse in the mouth."

After reviewing a bunch of numbers, a lot of the coins with low mintage can not be right.

Using the following as an example:

United States
1913 Liberty nickel mintage 5 pieces 0.4%
1894 S 10 cent mintage 24 pieces 0.2%

Prices for the above coins are in the millions of dollars.

Ernie

Either some people are entering coins they do not have or the calculations are incorrect.

I checked some Finnish coins and either people have entered wrongly or many Numista members have fake Finnish coins. For example markka 1870 or penni 1864. This could be a feature to identify fake coins also. I have many 1 penni 1864 fakes and none real ones.

Quote: "erniemix"​After reviewing a bunch of numbers, a lot of the coins with low mintage can not be right.
​
​Using the following as an example:
​
​United States
​1913 Liberty nickel mintage 5 pieces 0.4%

[...]
​
​Either some people are entering coins they do not have or the calculations are incorrect.
​

​While this is certainly an interesting feature, taking that piece as an example, this is the reason why I was not overly excited for this feature.

The calculation is correct--there are 14 members who claim to own that 5-mintage year-line. And a big reason is claiming to own a year-line they do not actually have. I recognize a few names as being members who just mark every year-line as owned (I count five). I have never understood why some members do that...

And of course, there are mis-identified coin or counterfeits registered like actual coins.

Whether we see the number of members who own a year-line or the percent frequency, I imagine this issue will always be around. And because we cannot possibly ask every member for proof of their coin to ensure the frequency is accurate, I think the prices are still a better indicator of rarity.

From previous research regarding a coin (see https://en.numista.com/forum/topic59033.html ) many people intentionally check all years in the type for all the coins they own, even though they may only hold one year. They do this because they collect by type, and yes it greatly skews the data.

Amazing new feature! Congrats!

Funny, when I suggested it, the idea was shot down by the same admin who now suddenly decided that it's a good idea. I'm glad this was finally implemented, though.

Quote: "PianoMan96"​Funny, when I suggested it, the idea was shot down by the same admin who now suddenly decided that it's a good idea. I'm glad this was finally implemented, though.

​People change

Thanks for the implemented feature.

Just on question to Xavier. Are swapping coins taking into account as well? Thanks.

Yes, members are counted independently whether their coin are for swap or not.

I'm loving this feature, thanks!

Just a small suggestion, though. I think it would improve the ease of use to have a balloon explaining the concept when you hover over "frequency", in the same way that the title of the catalog appears when you hover over "KM" in the infobox.

There is already an explanation below the table of years. Obviously it's not visible enough. I'll make sure it's more easily discoverable.

Maybe if we had a condition grade labeled "fake" and then those coins would not be counted.

Hi
I guess this index behaves well as long as the number of total Numista users who own at least one coin from the whole mintage is lower than specific year mintage.
Question arises, how would this percentage figure behave, when (lets say 500 coins of specific year are minted) all coins are owned by Numista users, and we know the Number of Numista total users if far over 500.
LP

​Thank you Xavier! This is a really nice feature and looks very cool!

Quote: "Xavier"It shows the percentage of users who own each year or variety among all the users who own the coin. Note that, since some users own several versions, the sum may be greater than 100%.
​

Only thing is, I am struggling a bit with this above 100% perk. Is the purpose to show how many people own duplicates? It is a bit difficult to figure it out on coins that have a lot of date lines. I think a normalised value is more intuitive, or at least displaying the total % somewhere?

Best
strato

Hello,

Great feature. However I agree with people who suggest the total should be 100% for a coin. It would be easier to understand.

As for the problem of unrealistic owners, I create this for a discussion
https://en.numista.com/forum/topic90524.html#p764148

I actually like that the percent totals can be more than 100 percent.

Here is an obvious example (110 percent total):
https://en.numista.com/catalogue/pieces78794.html

Altogether, there are ten different members who claim to own a year-line. All of them own 1648, whereas only one owns 1647 (in addition to 1648). By allowing the totals to be over 100 percent, all year-lines are independant of each other.

And the lines being independant of each other would certainly help on pages with many year-lines. If we switch to a 100-percent-total system, and one particular page has dozens of year-lines, the highest number may be quite low because all year-lines would be dependant on each other. With the current system, I imagine the frequency numbers are more accurate.

If any of that makes sense...

For this coin with hundreds of date lines:

1950F :: 450 pcs. :: 0.5%
1969F :: 5,000 pcs. :: 0.1%
1996J :: 50,000 pcs. :: 0.2%
1987D :: 500,000 pcs. :: 0.4%
1994G :: 5,000,000 pcs. :: 0.9%
1950D :: 100,000,000 pcs. :: 28%

For me, this Frequency % would tell how rare a date is, compared to the other dates. Useful when mintage is not known. Or when mintage is known, but some coins were melted later.

The current % is not very useful for this purpose. I suspect not counting duplicates also contribute to this. I would be very curious to know, but only someone with access to the numbers could check.

I do agree that "independent" lines are great. The most "popular" coin could be normalised to 100. e.g. the most popular date has 2,000 owners, it gets 100. Then a date with 40 owners gets 2 (40/2000) and finally, a date with only 1 owner would get < 0.1.

Quote: "Sulfur"​I actually like that the percent totals can be more than 100 percent.
​
​

​ Sulfur, my big plus to your comment! It is bright advantage of this calc. methodology over others.

If you really want, as stated, to evaluate the rarity of a coin, you should count coins, not users.
If I understand well, if a member has 3 of the same coin, it counts only 1. So it underestimates the frequency.

The index to be accurate should be independent of the identity of the owners. We are talking of coins, not users. So it should be number of coins of this line/total number of coins of the whole page in %, no ?

it's better to see how many of each coin are owned, and the number of users/owners.

I love this feature. It is the best new addition to Numista since I've been a member.

Also, I'm very much in favour of the > 100% situation. It makes perfect sense if you think of it in terms of "25% of people who own this coin have this date". That is a much more useful metric than a normalised version, which would be "25% of the owned coins of this type are of this date." The latter would give much smaller numbers for coins with lots of year lines, and we would therefore lose considerable "resolution" to rounding.

I think this is perfect as it is and I hope we don't change it.

Thank you, Xavier!

I still dont know exactly what the procentage means??

Quote: "gyoschak"​what percentage of people that own this type own this particular year
​
​So in a type with only years 1995, 1996 and 1997:
​-If everybody owned and cataloged 1997, and nobody owned or cataloged 1996 and 1995, the percentage would be 1997-100%, 1996-0% and 1995-0%.
​-If everyone owned 1996 and 1997, an nobody owned 1995: 1997-100%, 1996-100%, 1995-0%.

​

I agree with Ecapoe. We're interested in the rarity of a date. If we don't count duplicates, this won't be useful.

In Sulfur's example, there could be 1 x 1647 and 10 x 1648s. But there could also be 1 x 1647 and 10,000 x 1648s distributed between 10 owners. I would see the same %, despite a big difference in actual rarity.

If small numbers are a problem, they can be normalised. But if the date's frequency is 1 x 1647 for every 10,000 x 1648's, I want to see the small number, rather than 10% which is misleading.

Quote: "stratocaster"​I agree with Ecapoe. We're interested in the rarity of a date. If we don't count duplicates, this won't be useful.
​
​In Sulfur's example, there could be 1 x 1647 and 10 x 1648s. But there could also be 1 x 1647 and 10,000 x 1648s distributed between 10 owners. I would see the same %, despite a big difference in actual rarity.
​
​If small numbers are a problem, they can be normalised. But if the date's frequency is 1 x 1647 for every 10,000 x 1648's, I want to see the small number, rather than 10% which is misleading.

​I see your point, Strat, but I'm concerned that this isn't going to give you the results that you expect for a number of reasons.

1. Numista is not a scientific census of coins and each collector has a different methodology when it comes to entering their collection. I, for instance, only list the best coin I have for each year in Numista, so I might have 10 x 1648s, but I'd only list one. That is just how I have chosen to use the system but, in your example, it would lead to an inaccuracy of about 9%, which is quite a lot.

2. People's collections are also different. Some people only collect until a specific date, so if you have a coin that was minted from 1939 to 1970, for example, and some collector only collects up to 1950 overall, then you would be skewing the numbers towards the earlier dates. (This argument holds for the current system too, but Xavier is not claiming to present actual rarity information - just the relative number of people who have a coin in their collection.)

3. And actually, I was surprised to discover that it is fundamentally flawed to extrapolate the actual rarity of a coin from the numbers held in Numista collections! I was actually trying to do precisely that using the frequency data for US Nickels a few nights ago and I discovered that the results were the opposite of what I expected. The higher the mintage for each year, the smaller fraction of the overall number of coins produced are recorded in Numista. As soon as I thought about it, this made perfect sense because, although there are a lot of us, the number of collectors for each type of coin is limited.

So, I would guess that Xavier has actually thought carefully through these scenarios and that what we have is intended to be taken at face value, and might be extrapolated to hint at the availability of coins for each year. But it isn't, and it can't ever be, a scientific measurement of the actual rarity of different dates within a type.

Thanks andrewdotcoza. I am aware of all these points. I don't enter all the coins I own either. Also sometimes people check all datelines. As you've acknowledged, these things affect both systems so 1 and 2 are irrelevant.

Re. 3, I am not interested in a scientific survey. I am just interested strictly in how many coins are logged in Numista, however people want to list them.

I think user-related stats could fit better in the swap mechanics, wish lists etc. I am not very familiar with this though.

Best
strato

About the presentation of the column.

It is not good to make a balloon explaining the concept when you hover over "frequency": it is too much to explain and to understand.

But if the explanation of meaning of colors be added to the text already written, together with the colors themselves, then the lines about frequency will attract attention. So it will become 'visible'.

Something like that:

Frequencies show the percentage of Numista users who own each year or variety among all the users who own this coin. Since some users own several versions, the sum may be greater than 100%.
(Color) for frequency from ...% till ...%, (color) for frequency from ...% till ...%, (color) for frequency from ...% till ...%, (color) for frequency from ...% till ...%. No numbers mean too little information for statistics.

Hello,

I just changed the frequencies so that the total equals to 100%, as per the usual statistical definition, so that it should be more straightforward to understand.

The new definition is: number of Numista users owning that year/variety among the total of coins registered in the collections of Numista users (counting only one coin per user and per year/variety).

For example, if a coin type was issued in 2 years

• 2000: owned by users A, B and C
  
• 2001: owned by users A and D
  

Then the frequencies are the following:

• 2000: 3/5 = 60% (instead of 3/4 = 75% with the previous formula)
  
• 2001: 2/5 = 40% (instead of 2/4 = 50% with the previous formula)
  

...Huh. That's actually a little sad - I liked the old version much better.

Would it be possible to switch to the old version in some way? A button in the settings or something.

Quote: "Xavier"​​For example, if a coin type was issued in 2 years
​

• 2000: owned by users A, B and C
  ​
  
• 2001: owned by users A and D
  ​
  

​Then the frequencies are the following:
​

• 2000: 3/5 = 60% (instead of 3/4 = 75% with the previous formula)
  ​
  
• 2001: 2/5 = 40% (instead of 2/4 = 50% with the previous formula)
  ​
  

​

​Question for clarity:
Is this a “member-to-year ratio”, or a “total coin-to-year ratio”? (Is the total number of coins counted, or just one coin per year per user?) Of course some people have multiple copies of a single year. It sounds like a user-to-year ratio, and some people will say total coins is more important.

Thanks for running Numista!!!

Quote: "January First-of-May"​...Huh. That's actually a little sad - I liked the old version much better.
​
​Would it be possible to switch to the old version in some way? A button in the settings or something.

​Yeah, I'd also vote for the old version of this feature if I had a choice.

Sorry, this won't be an option.

I confirm that quantities entered by the users are not taken into account. Whether you enter 1 or 10,000 coins for the same type and same year won't change the frequencies.

We should probably have a poll as to which option is actually preferred - when it was the old way people complained that it should change, now that it changed people complain that they prefer the old way.

I certainly admit that I'm willing to accept either option if I actually got proof (i.e. a poll result) that this is actually the preference of the majority.

Also a vote for the old system, mainly because it is much faster to see the very common years (always 80% - 100%), than to have to compare the percentages against each other when there are many lines.

​​​​​​With the new system, percentages depend on the number of lines per cointype.

Let's vote!
http://www.easypolls.net/poll.html?p=5e07cbc2e4b02c7a707a97e9

Formula 3?:

Let's take the example of a coin type which was issued in 2 years with 10 total coins
2000: 8 owned by 3 users A, B and C
2001: 2 owned by 2 users A and D

Forumula 3: divide year by the total of coins
2000: 8/10 = 80%
2001: 2/10 = 20%

Quote: "January First-of-May"​We should probably have a poll as to which option is actually preferred - when it was the old way people complained that it should change, now that it changed people complain that they prefer the old way.

​The poll is up and I've cast my vote! :-)

But I would like to point something out to everyone in the thread: There isn't really any reason to force the total of all percentages on a coin page to 100%. If you gave an exam to a class of 20 people, they would each get a score for the exam like 70%, 85%, 90%, etc. You wouldn't expect the total of everyone's scores to be 100%.

Similarly, in a calculation that shows the relative availability of different variants of a coin type, there is no reason to expect the total to be 100%. In fact, like with the students, it would be a bad thing if it did.

Quote: "gyoschak"​Formula 3?:
​
​Let's take the example of a coin type which was issued in 2 years with 10 total coins
​2000: 8 owned by 3 users A, B and C
​2001: 2 owned by 2 users A and D
​
​Forumula 3: divide year by the total of coins
​2000: 8/10 = 80%
​2001: 2/10 = 20%
​

​But isn't this Formula 2?

[Edit: My apologies. It definitely isn't.]

Is there a mistake in the poll?

"Forumula 1: divide by the total of users
2000: 3/4 = 75%
2001: 2/5 = 40%
The total is more than 100% due to the user A owning both years."

Why is the divisor 5 and not 4 for the 2001 coin? I think the result for Formula 1 should be:

2000: 3/4 = 75%
2001: 2/4 = 50%

No, formula 3 is total coins, the other formulas only look at the number of users. Formula 3 doesn’t care how many users there are.

Quote: "gyoschak"​No, formula 3 is total coins, the other formulas only look at the number of users. Formula 3 doesn’t care how many users there are.

​Yes, I understand. :-) My mistake, and I apologise.

I do actually think this is a better than Formula 2. However, it feels like your idea is to try and guess the relative number of coins minted from the relative number of coins in Numista, and that isn't ever going to give a useful result because the number of collectors in the world is limited. So if we were optimistic and imagined there was one coin that everyone had, the result of this formula would be the same if 100 or 1 million of those coins had been minted.

Formula 1, however, gives us something much more useful: a prediction of how likely we are to find a particular coin variant based on the shared experience of all Numista users. It also allows us to measure how lucky we have been when we get a less common variant. I really like that!

Quote: "andrewdotcoza"​it feels like your idea is to try and guess the relative number of coins minted from the relative number of coins in Numista

That ​is not my idea (if I understand you correctly). I simply want to know how hard it is to collect a year.

Total coin formula 3:
If 100 coins are entered in Numista, and 99 are from 2000 and 1 is from 2001, we can see 99% on 2000 and 1% on 2001.

Member formula 2:
But if one member has the 99 coins from 2000, and the other member has the 1 coin from 2001, the member formula shows 50%/50%. That would lead someone to believe each year is equally hard to find. But clearly it is not, because there are 99 times more 2000 coins found than 2001 coins found.

Also vote for formula #3. Counting coins, not users

Some people enter quantities of coins which have no link with the real number of coins they own, sometimes thousands. By counting only the users, these users can only modify the counts by 1, so it limits their negative impact. If quantities were taken into account, a single user could totally modify the frequencies by themself.

Quote: "andrewdotcoza"​Is there a mistake in the poll?
​
​"Forumula 1: divide by the total of users
​2000: 3/4 = 75%
​2001: 2/5 = 40%
​The total is more than 100% due to the user A owning both years."
​
​Why is the divisor 5 and not 4 for the 2001 coin? I think the result for Formula 1 should be:
​
​2000: 3/4 = 75%
​2001: 2/4 = 50%

​Indeed, my mistake. Here is the correct text for the poll:

Which formula do you prefer for frequencies of years and varieties?

Let's take the example of a coin type which was issued in 2 years

• 2000: owned by 3 users A, B and C
  
• 2001: owned by 2 users A and D
  

There are 4 distinct Numista users and a total of 5 user/years pairs.

Formula 1: divide by the total of users

• 2000: 3/4 = 75%
  
• 2001: 2/4 = 50%
  

The total is more than 100% due to the user A owning both years.

Formula 2: divide by the total of pairs user/year

• 2000: 3/5 = 60%
  
• 2001: 2/5 = 40%
  

The total is guaranteed to be 100%. Frequencies can be quite low in case the coin type has many years.

In both formulas, the quantities owned by the users are not taken into account. Only the number of users owning each year is considered.

Fine. Voted for the "forumula" I preferred

Anyone else would like to vote?
http://www.easypolls.net/poll.html?p=5e07cbc2e4b02c7a707a97e9

There is no difference between formula 1 and formula 2 obviously. That is why it does not seem to be the question, what someone prefers. There is no neediness for any pull.

Let us name 'position' a line in mintage section.
Let we have N people, who marked their positions, and let total amount of marked positions is M. Let Ni is the marked quantity of position number i.
Then, the frequency fi for position i is
fi=Ni/M

The first formula gives exactly the same result
result=Ni/N
and again
result=M/N · Ni/M=M/N · fi
We get, what should be proved. The first formula is just a multiplication of the standard definition of frequency fi on the same factor M/N for every position.
It means, that there are no more information in the first formula, neither more information in the second. The only difference is, as Xavier mentioned already, the second formula is the standard definition of frequency. Being so, the second formula
fi=Ni/M
should be in the column.

Then, as it was already stated, this frequency does not mean much: it already shows, that rarities are overpopulated, many collectors indicate all lines, just not to bother about them, when they have one sample. I met with a profile, when the collector informs, that he put the coins he has in the mintage section arbitrary, and only in case a specific year has meaning for you, then, he will look for your requested year manually. This policy adds extra confusion.

The topic of this discussion is
Number of people who own each year of the coins

So in fact we get what we wanted: according to definition above
Ni is Number of people who own each year of the coins,
and the column frequency now gives the exact answer:
fi=Ni/M
to this question.

The whole discussion above shows that there is not much trust in Ni, that is why fi is good indication in swaps, but it has nothing to do with mintages, population of coin, rarities and even fashion.

Here are the results of the poll:

• Formula 1: 55% (21 votes)
  
• Formula 2: 34% (13 votes)
  
• I'm fine with either formula: 11% (4 votes)
  

I understand there is a clear (yet not unanimous) preference for the formula 1, so I just brought back the initial version.

As mentioned by Cyprusalexander, both formula give the same information, since there is a fixed ratio between both. The advantage of formula 1 is that we avoid very low figures for coins with a lot of years/varieties.
I'm still a bit annoyed that we call "frequency" something that doesn't sum up to 100%. Feel free to suggest other names.

ownership?

Yay for formula 1.

Quote: "Xavier"​I'm still a bit annoyed that we call "frequency" something that doesn't sum up to 100%. Feel free to suggest other names.

​Prevalence?

Just "presence" or "presence in collections" would clearly define what is is.

A win for Formula 1 is a win for usability!

Ownership isn't that bad.
It's the chance of a member owning a certain year.
These values are good if you wantt to trade with a year collectors, the chance to swap a coin with an owner ship near the 100 will be quite challening.

The aroused question about new creation is exactly what was mentioned before. The first approach being expressed in mathematical form was named just 'result' (see above some posts), for there is no strict name for such procedure. The simple way is just to use the already existed term frequency. Every one knows it and can check its meaning in any mathematical textbook; there is explanation in addition at every article (being explained with colors it will be complete and visible). It means, that there is no neediness 'to invent a bicycle', just use the frequency
fi=Ni/M
and that is all. It is correct from all points of view and what is most important it is simple. Even those, who prefer to know how much this variant is under the common one in the year, must divide one number on the other. They do not in fact get what they need directly from numbers, they still calculate. It means, that no advantages at all are gained with the first approach.

The example with a list of students, who get their marks in % is presented by mistake, for it is not relevant to the case. Here we divide the amount of position Ni into whole number, it is the demand of procedure; in case of students there is no division into amount of students in class by any means, it is the demand of that procedure.

All these mean, that the only solution left is standard definition of frequency with the name of column 'frequency'.

Quote: "Xavier"​Hello and Merry Christmas!
​
​There is now an indicator of the frequency of each year/mint/variety to help identifying which years are more or less common. It shows the percentage of users who own each year or variety among all the users who own the coin. Note that, since some users own several versions, the sum may be greater than 100%.
​See for instance: https://en.numista.com/catalogue/pieces6834.html
​
​I don't think it's meaningful to display the exact number. The rarity index and this new frequency indicator already give an idea.
​Frequencies are displayed only if there are enough users.
​I hope you like it
​

​Hi Xavier

Sorry if you have answered this before, but what does the red dot mean in the frequency column?

Aaron

Hello Aaron,

More reddish --> Higher percentage, If I am not mistaken, and vice versa -).

Thanks afpcoin