@yosclimber said:
Good work.
So it apparently got re-holdered, but not re-stickered?
Looks that way
The seller also replied that it just needs to be sent back in for sticker.
OK, good.
So what do we add for the CAC quality on a toned Morgan? 17-20%? The rest attributed to obverse vs reverse?
Good question.
I believe this is one of the many things that @treybenedict's tools will investigate,
using prices realized (sold) and the color scores.
@yosclimber said:
Good work.
So it apparently got re-holdered, but not re-stickered?
Yes it was in a non-TrueView holder that sold at GC. It also had a minor chip. Now it has a TrueView and no chip but would need to be sent back to CAC to get restickered.
Here’s the pricing info for the same Kennedy when I scanned it with the app. I clicked on the little rainbow on the top right to get the above color scan
That’s what I was able to do without knowing what I was doing and just guessing. That part was pretty intuitive. It also has a guide that I found that looks like it explains how to do more things with the app.
It looks like a useful inventory app for PCGS, NGC, CAC stickered coins. Not sure yet what all it can do.
I’m not sure what it can do for scanning toned coins that don’t have Trueviews, it might just be the pricing based on the cert number, might only give pricing for the grade and not toning info
For fun I tried scanning a couple of NGC coins to see what the app could do for a coin without a TrueView. It just gives the pricing info. The first NGC coin scanned just find, but the 2nd one which was a so-called dollar didn’t work. Here’s the scan reports:
First one worked
@Mr_Spud said:
That’s what I was able to do without knowing what I was doing and just guessing. That part was pretty intuitive. It also has a guide that I found that looks like it explains how to do more things with the app.
It looks like a useful inventory app for PCGS, NGC, CAC stickered coins. Not sure yet what all it can do.
I’m not sure what it can do for scanning toned coins that don’t have Trueviews, it might just be the pricing based on the cert number, might only give pricing for the grade and not toning info
After reading through the guide thingy in the app, I think I already figured how to do everything it can do. It’s easy to use and the scanning for pricing might really come in handy. The toning scanning seems to be a fun bonus feature that people will like using, but for now at least it only works with coins that have Trueviews which it accesses by looking it up online from the cert number. Overall I like it
@Mr_Spud said:
For fun I tried scanning a couple of NGC coins to see what the app could do for a coin without a TrueView. It just gives the pricing info. The first NGC coin scanned just find, but the 2nd one which was a so-called dollar didn’t work. Here’s the scan reports:
First one worked
The second one gave me this error message
We created a matching index for NGC & PCGS coins, and some coins have not been added yet, this is a known error but we will fix them as we see them since most are edge cases.
Could you show me what the coin is that did this?
Edit: I see, we don’t have matching setup for so-called dollars yet. Will update this ASAP
@Mr_Spud said:
After reading through the guide thingy in the app, I think I already figured how to do everything it can do. It’s easy to use and the scanning for pricing might really come in handy. The toning scanning seems to be a fun bonus feature that people will like using, but for now at least it only works with coins that have Trueviews which it accesses by looking it up online from the cert number. Overall I like it
Thanks! I’m happy to hear this. Hopefully we can continue to improve your user experience. I’d love to talk with you at Long Beach and get more into the details
@split37 said:
The app is running very smoothly. I really appreciate all the work put into this. It's a very handy tool to have on the go.
Thank you!
Amazing I’m glad to hear! I hope that it helps you in your near future buying & selling. We are hoping to add some cool features very soon and quickly so always open to hear suggestions of what you want to see!
Trey - One thing I previously suggested; Your qualitative analysis of toning for Peace dollars is skewed. The A,B,C grades for toned Peace dollars are being compared with the highest quality toned Morgans. The majority of my collection is Peace toners and I have amassed some of the best toned examples I've seen in 6 years. High-end rainbow Morgans compared to Peace dollars is not even a contest. I wonder if you can compile data by series for toning rather than compare toning on an "every-coin" scale?
I tried to download from the App Store to help the cause, but it says it requires iOS 16.0 or higher, and my iphone is older.
Oh, well. Most of my coins are raw and untoned anyway.
@split37 said:
Trey - One thing I previously suggested; Your qualitative analysis of toning for Peace dollars is skewed. The A,B,C grades for toned Peace dollars are being compared with the highest quality toned Morgans. The majority of my collection is Peace toners and I have amassed some of the best toned examples I've seen in 6 years. High-end rainbow Morgans compared to Peace dollars is not even a contest. I wonder if you can compile data by series for toning rather than compare toning on an "every-coin" scale?
Totally understand. We are going to segment by series in the future. This is just a public preview to test the waters. I created the model based on toned Morgan’s and understand that different coins can tone to different degrees depends on the series.
Apologies for this, but appreciate your feedback and will try to update accordingly to accommodate different series.
@split37 said:
Trey - One thing I previously suggested; Your qualitative analysis of toning for Peace dollars is skewed. The A,B,C grades for toned Peace dollars are being compared with the highest quality toned Morgans. The majority of my collection is Peace toners and I have amassed some of the best toned examples I've seen in 6 years. High-end rainbow Morgans compared to Peace dollars is not even a contest. I wonder if you can compile data by series for toning rather than compare toning on an "every-coin" scale?
Given that MS grades are relative to a series also, I wouldn't worry about comparing the toning grades from different series. It should all be relative.
@split37 said:
Trey - One thing I previously suggested; Your qualitative analysis of toning for Peace dollars is skewed. The A,B,C grades for toned Peace dollars are being compared with the highest quality toned Morgans. The majority of my collection is Peace toners and I have amassed some of the best toned examples I've seen in 6 years. High-end rainbow Morgans compared to Peace dollars is not even a contest. I wonder if you can compile data by series for toning rather than compare toning on an "every-coin" scale?
Given that MS grades are relative to a series also, I wouldn't worry about comparing the toning grades from different series. It should all be relative.
I can see this perspective too. However I agree that the scales should be adaptive to series and to overall performance. I think this could be interesting as a toggle option. Maybe some A/B testing to see how people like it!
Totally understand. We are going to segment by series in the future. This is just a public preview to test the waters. I created the model based on toned Morgan’s and understand that different coins can tone to different degrees depends on the series.
Apologies for this, but appreciate your feedback and will try to update accordingly to accommodate different series.
No apology necessary. I imagine with enough data, you will be able to systematically and categorically organize toners from best to worst per series. This is such an amazing tool and I can't emphasize how much value this brings to a toned collector like myself. Again, many thanks and keep up the good work!
Given that MS grades are relative to a series also, I wouldn't worry about comparing the toning grades from different series. It should all be relative.
100% disagree; SLQ's and Peace dollars do not tone like Morgans and Mercury dimes. I have input less than MS Mercury dimes and gotten higher toning scores than some of the most high-end toned Peace dollars in existence.
As far as analysis of toning, does it ignore non-US coinage? I ask because I added a Canadian medal certified by PCGS and the correct PCGS data came up, but there doesn't appear to be a way to analyze the color.
@TomB said:
As far as analysis of toning, does it ignore non-US coinage? I ask because I added a Canadian medal certified by PCGS and the correct PCGS data came up, but there doesn't appear to be a way to analyze the color.
Currently just US. However if you send me the cert for the Canadian coin I can ungate them to be run on the analysis.
It matches “reviewable” coins with their denomination so I believe that Canadian may come up as S$1 or C$1.
I checked out one of my coins. I’ll check many more tomorrow during the day after I get some sleep. I expected the pricing of toners to be flawed and I am not surprised but this one is kinda funny. You should go look at what this recently sold for at Great Collections recently and compare it to the PCGS price guide. Out of my collection, I have paid as much as 16x price guide. For some reason, I don’t expect your AI models to be in sync quite yet. Keep working at it though. I’ll play with the app more tomorrow night most likely. CAC posted this same coin on their Instagram today and mentioned their price guide for it as well. That too was funny…so it’s not just you. Instead of just taking the TPG population into consideration, I wonder if it would be helpful if you examined the CAC population. That could help produce some different information. Another very important factor is the rarity of excellent toning for a given date. Take Morgan Dollars as an example. An 1881-S with monster toning is nice but there are a ton of them. If you take that same toning and put it on another date and it becomes a gazillion dollar coin because great toning is so rare on some dates. The same applies for Mercury Dimes or any other series. Your app would have to somehow be able to have access to all True Views ever taken, score them and come up with a rarity rating for great toning on all dates. Something to think about. This is certainly something that is top of mind for collectors. My first experience with the app and I think it is pretty darn cool. I think you are definitely building on something neat. Keep it up!
"If it's not fun, it's not worth it." - KeyMan64 Looking for Top Pop Mercury Dime Varieties & High Grade Mercury Dime Toners.
@keyman64 said:
I checked out one of my coins. I’ll check many more tomorrow during the day after I get some sleep. I expected the pricing of toners to be flawed and I am not surprised but this one is kinda funny. You should go look at what this recently sold for at Great Collections recently and compare it to the PCGS price guide. Out of my collection, I have paid as much as 16x price guide. For some reason, I don’t expect your AI models to be in sync quite yet. Keep working at it though. I’ll play with the app more tomorrow night most likely. CAC posted this same coin on their Instagram today and mentioned their price guide for it as well. That too was funny…so it’s not just you. Instead of just taking the TPG population into consideration, I wonder if it would be helpful if you examined the CAC population. That could help produce some different information. Another very important factor is the rarity of excellent toning for a given date. Take Morgan Dollars as an example. An 1881-S with monster toning is nice but there are a ton of them. If you take that same toning and put it on another date and it becomes a gazillion dollar coin because great toning is so rare on some dates. The same applies for Mercury Dimes or any other series. Your app would have to somehow be able to have access to all True Views ever taken, score them and come up with a rarity rating for great toning on all dates. Something to think about. This is certainly something that is top of mind for collectors. My first experience with the app and I think it is pretty darn cool. I think you are definitely building on something neat. Keep it up!
Thank you for this amazing feedback. That’s an incredible coin as well.
Pricing for toners is not yet integrated yet. The prices show are base prices for certs that match the given specifications.
We are hoping to integrate that later though with more data and training!
Curious to hear more feedback as you use the app more
Given that MS grades are relative to a series also, I wouldn't worry about comparing the toning grades from different series. It should all be relative.
100% disagree; SLQ's and Peace dollars do not tone like Morgans and Mercury dimes. I have input less than MS Mercury dimes and gotten higher toning scores than some of the most high-end toned Peace dollars in existence.
I'm aware that they tone differently.
So would you like to see a scale for Peace dollar toning that assigns an "A" grade if a Peace dollar gets an absolute color score that is close to the maximum absolute color score observed across all Peace dollars?
I call that a relative score. Or one that is "normalized" relative to the max of the series.
Clearly, it would require more data, and it may not be stable.
What happens if a better toned Peace dollar comes along?
Does it get an A++?
Or do all the older scores get changed?
It's not so easy to implement a stable relative score because of this. @keyman64 already considered this, as he took it to the next level where he discussed toning by date within a series.
One simple solution is that the software could provide 2 color scores - one absolute and one relative to the max of the series. Or a third score for the date-mint mark.
In terms of how the color score ultimately affects the price equation, it should have a series-specific coefficient or coefficients. So it would not matter if the color score is normalized for a series. Any series-specific scaling would be handled by the series-specific coefficient. It makes no difference in the math. So don't worry - it should not undervalue your toned Peace dollars.
(price effect of toning) = Color*coef_Peace
= (Color/max_Peace_Color)*(max_Peace_Color*coef_Peace)
= (normalized Peace color score)*(coef of normalized Peace color score)
Note: the above is a linear price effect, as an example. It would be in addition to pricing by grade, date-mint mark, FSB, etc. It would likely be part of a proportional factor to a price that ignores toning, like
toned_price = untoned_price(grade,FSB,date,mint mark)*exp(Color*coef_series)
I'm not part of Trey's team, and I don't know the exact specification of how they might build their model to handle pricing for toned coins.
As a general suggestion, when presenting an estimated price based on a model, it is helpful and usually not difficult to also provide a standard error for that estimate. E.g. $1,077.77 +/- $82. There's plenty of screen real estate to display this, and it helps the user understand the degree of confidence there is in the estimate. It would not be so easy for, say, the PCGS Price Guide to provide standard errors for every value in one of their tables by date-mint mark and grade - it would double the size of that table.
I think the toning scores should be the same across all series. If a certain series doesn't have nice toning so be it. A nice toner is a nice toner independent of series. There will never be Peace dollars that tone as well as Morgans, but that doesn't mean Peace dollars should get an "A" score for subpar toning. Just like I shouldn't get an A in school if I'm a C student. As for value a C toned Peace may be very valuable thou and pricing can take this into account. The same for certain dates of a series. An 1881 S toned Morgan would be valued less than an equally toned tougher date. I think the pricing should reflect these things, not the toning grade.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
This app looks to have excellent potential! I loaded about 50 slabs this morning. Only issues so far are NGC tokens, USPI, and ultra modern coins. Can’t wait to see it develop.
Congrats on the release Trey, I find that the app works pretty seamlessly as far as loading inventory, scanning coins, and navigating through the app. I had no issues with anything.
That being said, I would like to see the pricing data rely on something other than stale comps. For example, my 1836 NCAC $5 in 58 is showing to have a value around $2000 in NCAC, but that info is based off of 10 year old comps. However, if I were to switch to any other pricing (P or N without CAC) it would be way more, and pcac is $4500. I recall you saying that the pricing data could pull comps from adjacent coins in the same series, with similar rarity, but it looks like it’s all based on whatever comps are available. There is no data for my 1804 $5 and the app won’t accept my Kellogg $20, so I’m guessing a future update may include pioneer gold?
@keyman64 said:
Pricing of toners will be very interesting and I am skeptical. Some great toners will sell near price guide, some 3-6 times price guide and others can go 20x price guide. There are so many variables and if you mess up on just one you might not even be in the ballpark. The rarity of toning based on the date and grade cannot be ignored. It seems to be a risky move where people might just call it another junk price guide.
I'm really interested to see how AI/ML can assist here. It seems like a perfect application given enough data.
@Mr_Spud said:
Here’s the pricing info for the same Kennedy when I scanned it with the app. I clicked on the little rainbow on the top right to get the above color scan
Very interesting. This looks like a huge price drop from 2005 which bears investigation.
It would be useful to see the Pure Estimates for the auction sold coins to see how close they are to the historical realized sale prices.
It also seems interesting that there are only 2 records of PCGS AU58 coins selling over the past 18 years. Will more records be added?
Great work on the app! These questions are only to help make it more useful
@AlbumAccumulator said:
This app looks to have excellent potential! I loaded about 50 slabs this morning. Only issues so far are NGC tokens, USPI, and ultra modern coins. Can’t wait to see it develop.
Congrats on the release Trey, I find that the app works pretty seamlessly as far as loading inventory, scanning coins, and navigating through the app. I had no issues with anything.
That being said, I would like to see the pricing data rely on something other than stale comps. For example, my 1836 NCAC $5 in 58 is showing to have a value around $2000 in NCAC, but that info is based off of 10 year old comps. However, if I were to switch to any other pricing (P or N without CAC) it would be way more, and pcac is $4500. I recall you saying that the pricing data could pull comps from adjacent coins in the same series, with similar rarity, but it looks like it’s all based on whatever comps are available. There is no data for my 1804 $5 and the app won’t accept my Kellogg $20, so I’m guessing a future update may include pioneer gold?
Kellogg cert 16223562
Correct, pioneer gold isn’t added yet on the NGC side of things. This will be coming in a future update.
Right now there are no indexes set up to be able to adjust present value within sales. So predicting based off old sales has not been integrated yet either. This will be coming in a later update, as I want to test a couple different methods to see which has the highest accuracy.
It’s also possible that there are no sales for those coins and they need to be compared to alike coins in the series. We are working on our own scarcity ratings as well for these use cases.
Really appreciate the through testing though!
Edit: Try doing manual entry on the Kellogg. It should work, just tested that cert. Likely the barcode isn’t registering correct.
@Mr_Spud said:
Here’s the pricing info for the same Kennedy when I scanned it with the app. I clicked on the little rainbow on the top right to get the above color scan
Very interesting. This looks like a huge price drop from 2005 which bears investigation.
It would be useful to see the Pure Estimates for the auction sold coins to see how close they are to the historical realized sale prices.
It also seems interesting that there are only 2 records of PCGS AU58 coins selling over the past 18 years. Will more records be added?
Great work on the app! These questions are only to help make it more useful
For coins like these, I believe that eBay and GC would be valuable pricing resources. We are working on integrating eBay into the platform, it’s just very difficult to register sales over 90 days old.
From my analysis, the pure estimate is typically on the conservative side.
@AlbumAccumulator said:
This app looks to have excellent potential! I loaded about 50 slabs this morning. Only issues so far are NGC tokens, USPI, and ultra modern coins. Can’t wait to see it develop.
Thank you!! We will have these pushed in the next update Are most of the tokens US Civil War?
Very cool app, I love it! The only issues are that proof/proof-like coins come up as having blue toning, while being blast-white. The other issue is much more serious, theres no “RickO” grade for the color scale!
Type collector, mainly into Seated. -formerly Ownerofawheatiehorde. Good BST transactions with: mirabela, OKCC, MICHAELDIXON, Gerard
@Ownerofawheatiehorde said:
Very cool app, I love it! The only issues are that proof/proof-like coins come up as having blue toning, while being blast-white. The other issue is much more serious, theres no “RickO” grade for the color scale!
Interesting. Not a lot of PL/DMPL coins and were used in training data so I will add some more today. This makes sense that it comes up as so, since PL has a blue-ish tint when Blast white. Will just have to account for normalization on my end.
@AlbumAccumulator said:
This app looks to have excellent potential! I loaded about 50 slabs this morning. Only issues so far are NGC tokens, USPI, and ultra modern coins. Can’t wait to see it develop.
Thank you!! We will have these pushed in the next update Are most of the tokens US Civil War?
Yes largely Civil War tokens and a few early 20th century merchant tokens.
How are these graphs actually read or scored?
I noticed when you scored a few on the other thread you came up with scores. Did you get those numbers based on these graphs?
@robec said:
How are these graphs actually read or scored?
I noticed when you scored a few on the other thread you came up with scores. Did you get those numbers based on these graphs?
Yes both are used in the score. To give an idea of how the score is calculated:
RGB & HSV analysis is compared to a large set of training data, it is then adjusted and normalized to fall within bounds. Then the model predicts where it would fall given all the rest of the data, giving a numerical identifier. This is then mapped back to the bounds of the letter grades A-F.
In the next update, we will have a version to toggle between obverse and reverse graphs, as well as score and letter grades.
Comments
Looks that way
The seller also replied that it just needs to be sent back in for sticker.
So what do we add for the CAC quality on a toned Morgan? 17-20%? The rest attributed to obverse vs reverse?
Thanks! Can’t wait to go live tomorrow and hear your feedback.
OK, good.
Good question.
I believe this is one of the many things that @treybenedict's tools will investigate,
using prices realized (sold) and the color scores.
App went live a bit early
Test and submit feedback here before leaving a review!
Will iterate and update based on feedback in this thread.
Thanks to over 100 who pre-ordered.
Yes it was in a non-TrueView holder that sold at GC. It also had a minor chip. Now it has a TrueView and no chip but would need to be sent back to CAC to get restickered.
I just scanned this Kennedy with the app
Mr_Spud
Here’s the pricing info for the same Kennedy when I scanned it with the app. I clicked on the little rainbow on the top right to get the above color scan
Mr_Spud
And I just scanned this dime with the app. I saved it in my inventory too by clicking on the down arrow next to the rainbow.
Mr_Spud
That’s what I was able to do without knowing what I was doing and just guessing. That part was pretty intuitive. It also has a guide that I found that looks like it explains how to do more things with the app.
It looks like a useful inventory app for PCGS, NGC, CAC stickered coins. Not sure yet what all it can do.
I’m not sure what it can do for scanning toned coins that don’t have Trueviews, it might just be the pricing based on the cert number, might only give pricing for the grade and not toning info
Mr_Spud
For fun I tried scanning a couple of NGC coins to see what the app could do for a coin without a TrueView. It just gives the pricing info. The first NGC coin scanned just find, but the 2nd one which was a so-called dollar didn’t work. Here’s the scan reports:
First one worked
The second one gave me this error message
Mr_Spud
Toning pricing is not yet implemented!
This will be in a future update.
After reading through the guide thingy in the app, I think I already figured how to do everything it can do. It’s easy to use and the scanning for pricing might really come in handy. The toning scanning seems to be a fun bonus feature that people will like using, but for now at least it only works with coins that have Trueviews which it accesses by looking it up online from the cert number. Overall I like it
Mr_Spud
We created a matching index for NGC & PCGS coins, and some coins have not been added yet, this is a known error but we will fix them as we see them since most are edge cases.
Could you show me what the coin is that did this?
Edit: I see, we don’t have matching setup for so-called dollars yet. Will update this ASAP
It was this one
Mr_Spud
Thanks!
Will try to get the whole so-called dollar series in as soon as possible.
Coins that come up with no-match error are a work in progress for us. 99% of all US Type coins & varieties should be included.
Happy tracking See you at Long Beach.
Thanks! I’m happy to hear this. Hopefully we can continue to improve your user experience. I’d love to talk with you at Long Beach and get more into the details
The app is running very smoothly. I really appreciate all the work put into this. It's a very handy tool to have on the go.
Thank you!
Amazing I’m glad to hear! I hope that it helps you in your near future buying & selling. We are hoping to add some cool features very soon and quickly so always open to hear suggestions of what you want to see!
Trey - One thing I previously suggested; Your qualitative analysis of toning for Peace dollars is skewed. The A,B,C grades for toned Peace dollars are being compared with the highest quality toned Morgans. The majority of my collection is Peace toners and I have amassed some of the best toned examples I've seen in 6 years. High-end rainbow Morgans compared to Peace dollars is not even a contest. I wonder if you can compile data by series for toning rather than compare toning on an "every-coin" scale?
I tried to download from the App Store to help the cause, but it says it requires iOS 16.0 or higher, and my iphone is older.
Oh, well. Most of my coins are raw and untoned anyway.
Totally understand. We are going to segment by series in the future. This is just a public preview to test the waters. I created the model based on toned Morgan’s and understand that different coins can tone to different degrees depends on the series.
Apologies for this, but appreciate your feedback and will try to update accordingly to accommodate different series.
Given that MS grades are relative to a series also, I wouldn't worry about comparing the toning grades from different series. It should all be relative.
I can see this perspective too. However I agree that the scales should be adaptive to series and to overall performance. I think this could be interesting as a toggle option. Maybe some A/B testing to see how people like it!
No apology necessary. I imagine with enough data, you will be able to systematically and categorically organize toners from best to worst per series. This is such an amazing tool and I can't emphasize how much value this brings to a toned collector like myself. Again, many thanks and keep up the good work!
100% disagree; SLQ's and Peace dollars do not tone like Morgans and Mercury dimes. I have input less than MS Mercury dimes and gotten higher toning scores than some of the most high-end toned Peace dollars in existence.
As far as analysis of toning, does it ignore non-US coinage? I ask because I added a Canadian medal certified by PCGS and the correct PCGS data came up, but there doesn't appear to be a way to analyze the color.
In honor of the memory of Cpl. Michael E. Thompson
Currently just US. However if you send me the cert for the Canadian coin I can ungate them to be run on the analysis.
It matches “reviewable” coins with their denomination so I believe that Canadian may come up as S$1 or C$1.
Thank you and PM sent.
In honor of the memory of Cpl. Michael E. Thompson
Bravo to PCGS for watching and letting this ride.
Keep up the great work Trey.
I checked out one of my coins. I’ll check many more tomorrow during the day after I get some sleep. I expected the pricing of toners to be flawed and I am not surprised but this one is kinda funny. You should go look at what this recently sold for at Great Collections recently and compare it to the PCGS price guide. Out of my collection, I have paid as much as 16x price guide. For some reason, I don’t expect your AI models to be in sync quite yet. Keep working at it though. I’ll play with the app more tomorrow night most likely. CAC posted this same coin on their Instagram today and mentioned their price guide for it as well. That too was funny…so it’s not just you. Instead of just taking the TPG population into consideration, I wonder if it would be helpful if you examined the CAC population. That could help produce some different information. Another very important factor is the rarity of excellent toning for a given date. Take Morgan Dollars as an example. An 1881-S with monster toning is nice but there are a ton of them. If you take that same toning and put it on another date and it becomes a gazillion dollar coin because great toning is so rare on some dates. The same applies for Mercury Dimes or any other series. Your app would have to somehow be able to have access to all True Views ever taken, score them and come up with a rarity rating for great toning on all dates. Something to think about. This is certainly something that is top of mind for collectors. My first experience with the app and I think it is pretty darn cool. I think you are definitely building on something neat. Keep it up!
Looking for Top Pop Mercury Dime Varieties & High Grade Mercury Dime Toners.
Thank you for this amazing feedback. That’s an incredible coin as well.
Pricing for toners is not yet integrated yet. The prices show are base prices for certs that match the given specifications.
We are hoping to integrate that later though with more data and training!
Curious to hear more feedback as you use the app more
I'm aware that they tone differently.
So would you like to see a scale for Peace dollar toning that assigns an "A" grade if a Peace dollar gets an absolute color score that is close to the maximum absolute color score observed across all Peace dollars?
I call that a relative score. Or one that is "normalized" relative to the max of the series.
Clearly, it would require more data, and it may not be stable.
What happens if a better toned Peace dollar comes along?
Does it get an A++?
Or do all the older scores get changed?
It's not so easy to implement a stable relative score because of this.
@keyman64 already considered this, as he took it to the next level where he discussed toning by date within a series.
One simple solution is that the software could provide 2 color scores - one absolute and one relative to the max of the series. Or a third score for the date-mint mark.
In terms of how the color score ultimately affects the price equation, it should have a series-specific coefficient or coefficients. So it would not matter if the color score is normalized for a series. Any series-specific scaling would be handled by the series-specific coefficient. It makes no difference in the math. So don't worry - it should not undervalue your toned Peace dollars.
(price effect of toning) = Color*coef_Peace
= (Color/max_Peace_Color)*(max_Peace_Color*coef_Peace)
= (normalized Peace color score)*(coef of normalized Peace color score)
Note: the above is a linear price effect, as an example. It would be in addition to pricing by grade, date-mint mark, FSB, etc. It would likely be part of a proportional factor to a price that ignores toning, like
toned_price = untoned_price(grade,FSB,date,mint mark)*exp(Color*coef_series)
I'm not part of Trey's team, and I don't know the exact specification of how they might build their model to handle pricing for toned coins.
As a general suggestion, when presenting an estimated price based on a model, it is helpful and usually not difficult to also provide a standard error for that estimate. E.g. $1,077.77 +/- $82. There's plenty of screen real estate to display this, and it helps the user understand the degree of confidence there is in the estimate. It would not be so easy for, say, the PCGS Price Guide to provide standard errors for every value in one of their tables by date-mint mark and grade - it would double the size of that table.
I think the toning scores should be the same across all series. If a certain series doesn't have nice toning so be it. A nice toner is a nice toner independent of series. There will never be Peace dollars that tone as well as Morgans, but that doesn't mean Peace dollars should get an "A" score for subpar toning. Just like I shouldn't get an A in school if I'm a C student. As for value a C toned Peace may be very valuable thou and pricing can take this into account. The same for certain dates of a series. An 1881 S toned Morgan would be valued less than an equally toned tougher date. I think the pricing should reflect these things, not the toning grade.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM
This app looks to have excellent potential! I loaded about 50 slabs this morning. Only issues so far are NGC tokens, USPI, and ultra modern coins. Can’t wait to see it develop.
My Jefferson Registry:
https://pcgs.com/setregistry/alltimeset/191115
@treybenedict
Congrats on the release Trey, I find that the app works pretty seamlessly as far as loading inventory, scanning coins, and navigating through the app. I had no issues with anything.
That being said, I would like to see the pricing data rely on something other than stale comps. For example, my 1836 NCAC $5 in 58 is showing to have a value around $2000 in NCAC, but that info is based off of 10 year old comps. However, if I were to switch to any other pricing (P or N without CAC) it would be way more, and pcac is $4500. I recall you saying that the pricing data could pull comps from adjacent coins in the same series, with similar rarity, but it looks like it’s all based on whatever comps are available. There is no data for my 1804 $5 and the app won’t accept my Kellogg $20, so I’m guessing a future update may include pioneer gold?
Kellogg cert 16223562
Founder- Peak Rarities
Website
Instagram
Facebook
I'm really interested to see how AI/ML can assist here. It seems like a perfect application given enough data.
Very interesting. This looks like a huge price drop from 2005 which bears investigation.
It would be useful to see the Pure Estimates for the auction sold coins to see how close they are to the historical realized sale prices.
It also seems interesting that there are only 2 records of PCGS AU58 coins selling over the past 18 years. Will more records be added?
Great work on the app! These questions are only to help make it more useful
Great feedback!
Correct, pioneer gold isn’t added yet on the NGC side of things. This will be coming in a future update.
Right now there are no indexes set up to be able to adjust present value within sales. So predicting based off old sales has not been integrated yet either. This will be coming in a later update, as I want to test a couple different methods to see which has the highest accuracy.
It’s also possible that there are no sales for those coins and they need to be compared to alike coins in the series. We are working on our own scarcity ratings as well for these use cases.
Really appreciate the through testing though!
Edit: Try doing manual entry on the Kellogg. It should work, just tested that cert. Likely the barcode isn’t registering correct.
For coins like these, I believe that eBay and GC would be valuable pricing resources. We are working on integrating eBay into the platform, it’s just very difficult to register sales over 90 days old.
From my analysis, the pure estimate is typically on the conservative side.
Thank you!! We will have these pushed in the next update Are most of the tokens US Civil War?
Very cool app, I love it! The only issues are that proof/proof-like coins come up as having blue toning, while being blast-white. The other issue is much more serious, theres no “RickO” grade for the color scale!
Type collector, mainly into Seated. -formerly Ownerofawheatiehorde. Good BST transactions with: mirabela, OKCC, MICHAELDIXON, Gerard
Interesting. Not a lot of PL/DMPL coins and were used in training data so I will add some more today. This makes sense that it comes up as so, since PL has a blue-ish tint when Blast white. Will just have to account for normalization on my end.
Hahaha, the RickO score is much needed for fun
Yes largely Civil War tokens and a few early 20th century merchant tokens.
My Jefferson Registry:
https://pcgs.com/setregistry/alltimeset/191115
Feel free to ping me when this is android ready or if you need android beta testers.
"It's like God, Family, Country, except Sticker, Plastic, Coin."
For sure! We should have web version up very soon.
How are these graphs actually read or scored?
I noticed when you scored a few on the other thread you came up with scores. Did you get those numbers based on these graphs?
- Bob -
MPL's - Lincolns of Color
Central Valley Roosevelts
Yes both are used in the score. To give an idea of how the score is calculated:
RGB & HSV analysis is compared to a large set of training data, it is then adjusted and normalized to fall within bounds. Then the model predicts where it would fall given all the rest of the data, giving a numerical identifier. This is then mapped back to the bounds of the letter grades A-F.
In the next update, we will have a version to toggle between obverse and reverse graphs, as well as score and letter grades.
Any questions just let me know!
I don't know if anyone has posted one yet, but I have achieved a Toning Grade of F for a coin. RickO would love it.
In honor of the memory of Cpl. Michael E. Thompson
@TomB I suspect that there are many more F coins out there than A coins.
Successful BST with ad4400, Kccoin, lablover, pointfivezero, koynekwest, jwitten, coin22lover, HalfDimeDude, erwindoc, jyzskowsi, COINS MAKE CENTS, AlanSki, BryceM