Features include:
• Dual Sided Toning Analysis (Obverse/Reverse segmented and ran independently)
• Toning score is now tappable to use in both Letter Grades and Number Scores.
• Drop down for all grades in all sales to quickly query another grade of a coin.
Bug fixes:
• NGC Matching Error
• Improved Matrix Scan
• Updated No Data Icons
@ElmerFusterpuck said:
Looks like a fun app so far! As an Android user, I'll just sit back, sip on some tequila and wait for that version.
Will there be an option in the app to analyze non TrueView photos? There are lots of us who take really good quality photos of coins - and quite a few them don't have TrueViews.
Yes. They’ll have to meet certain qualifications however for our storage purposes. I explained a bit more as to our idea on self-serve photos in the comment above.
It gets a little more tricky with matching photos and keeping usability on the model when new outside factors are added in.
Likely will be sooner than Android when we push a web versioning!
Would it be easier if you made "custom" photos have to have a solid black or white background?
Those that take their own pics can usually do a photo version of their coin with that type of solid background very easily. I know that, when I was doing my own pics, I could, and I am a neophyte compared to MANY on here that take pics.
@ElmerFusterpuck said:
Looks like a fun app so far! As an Android user, I'll just sit back, sip on some tequila and wait for that version.
Will there be an option in the app to analyze non TrueView photos? There are lots of us who take really good quality photos of coins - and quite a few them don't have TrueViews.
Yes. They’ll have to meet certain qualifications however for our storage purposes. I explained a bit more as to our idea on self-serve photos in the comment above.
It gets a little more tricky with matching photos and keeping usability on the model when new outside factors are added in.
Likely will be sooner than Android when we push a web versioning!
Would it be easier if you made "custom" photos have to have a solid black or white background?
Those that take their own pics can usually do a photo version of their coin with that type of solid background very easily. I know that, when I was doing my own pics, I could, and I am a neophyte compared to MANY on here that take pics.
I’ve figured out a way it could work, but I want to test with some early users here on the forum. I’ll likely be reaching out to some of you guys soon!
As a result of the new toning analysis update. It will likely take 1-2 minutes to be able to generate a report for both the obverse and the reverse. I've tested numerous amounts of times and it seems to be all fine.
1.2 update already has plans for some great new features. Excited to share soon!
Remember that sharing with your friends helps us tremendously in our startup journey.
Hey Dan, thanks for the response! Really appreciate these insights as well.
I think that the pure estimate is conservative. I understand that it’s likely below what most people typically pay for their coins at shows or in private sales, however I wanted to be a good baseline for users when looking into purchasing their items.
.
Previously I had questioned/asked what the Pure estimate was supposed to represent:
"Is it a prediction of what that same coin would sell for if put into an auction today? Does it represent a dealer wholesale, retail or what one might be able to sell it for to a dealer, on their own via eBay or other?"
With the above response "a good baseline for users when looking into purchasing their items" has it been determined what the Pure estimate is representing?
Also what is 'when future value is incorporated' going to be doing?
Future value/Present Value creates an index for alike coins in the series and tracks multiple grades, services, dates and varieties to see how they change over time. Very common in economic theory to give a nominal and a real value, similar application here.
As for the pure estimate, a larger more in depth post is coming to describe and explain more around the background of how it is derived and what it estimates.
Thanks!
That is what I thought of when you said future value (economics) but was not sure if this was related to the previously mentioned ratio from other examples or referring to something else.
Okay (coming later) on the what the program estimates represent. Today it is more difficult to determine (at least for me) what an auction result actually represents due to todays variables. I use auction results occasionally to try and figure values if I am searching around or interesting in something but I do a lot of scrubbing of the data so to say for what I believe is my specific case. But even then I can not be sure and figure a range and this is for a single coin.
https://youtube.com/watch?v=2YNufnS_kf4 - Mama I'm coming home ...................................................................................................................................................................... RLJ 1958 - 2023
@ElmerFusterpuck said:
Looks like a fun app so far! As an Android user, I'll just sit back, sip on some tequila and wait for that version.
Will there be an option in the app to analyze non TrueView photos? There are lots of us who take really good quality photos of coins - and quite a few them don't have TrueViews.
Yes. They’ll have to meet certain qualifications however for our storage purposes. I explained a bit more as to our idea on self-serve photos in the comment above.
It gets a little more tricky with matching photos and keeping usability on the model when new outside factors are added in.
Likely will be sooner than Android when we push a web versioning!
Would it be easier if you made "custom" photos have to have a solid black or white background?
Those that take their own pics can usually do a photo version of their coin with that type of solid background very easily. I know that, when I was doing my own pics, I could, and I am a neophyte compared to MANY on here that take pics.
I’ve figured out a way it could work, but I want to test with some early users here on the forum. I’ll likely be reaching out to some of you guys soon!
Wondering if circle cropping would work better, like this? That would eliminate background clutter like slab edges and other stuff.
@ElmerFusterpuck said:
Looks like a fun app so far! As an Android user, I'll just sit back, sip on some tequila and wait for that version.
Will there be an option in the app to analyze non TrueView photos? There are lots of us who take really good quality photos of coins - and quite a few them don't have TrueViews.
Yes. They’ll have to meet certain qualifications however for our storage purposes. I explained a bit more as to our idea on self-serve photos in the comment above.
It gets a little more tricky with matching photos and keeping usability on the model when new outside factors are added in.
Likely will be sooner than Android when we push a web versioning!
Would it be easier if you made "custom" photos have to have a solid black or white background?
Those that take their own pics can usually do a photo version of their coin with that type of solid background very easily. I know that, when I was doing my own pics, I could, and I am a neophyte compared to MANY on here that take pics.
I’ve figured out a way it could work, but I want to test with some early users here on the forum. I’ll likely be reaching out to some of you guys soon!
Wondering if circle cropping would work better, like this? That would eliminate background clutter like slab edges and other stuff.
It’ll be a form of something like this. There’s a couple good ways to automate the process to be able to give a good slab shot and a good finished crop.
Resolution then becomes a bigger issue. This photo is fine, but the one taken off my iPhone probably isn’t high quality enough.
I have scanned my slabs, the pricing seems to (off) toward the low side compared to resources such as the gray-sheet, 3rd party (website) reference and Redbook. I'll check again since you have updated your system. Also, old/outdated auction references.
@Joe_360 said:
I have scanned my slabs, the pricing seems to (off) toward the low side compared to resources such as the gray-sheet, 3rd party (website) reference and Redbook. I'll check again since you have updated your system. Also, old/outdated auction references.
Features include:
• Dual Sided Toning Analysis (Obverse/Reverse segmented and ran independently)
• Toning score is now tappable to use in both Letter Grades and Number Scores.
• Dropdown for all grades in all sales to quickly query another grade of a coin.
Bug fixes:
• NGC Matching Error
• Improved Matrix Scan
• Updated No Data Icons
• Updated Sales Data (Not all are fixed)
@treybenedict said: Update 1.1 is Live! Update on the app store here
Features include:
• Dual Sided Toning Analysis (Obverse/Reverse segmented and ran independently)
• Toning score is now tappable to use in both Letter Grades and Number Scores.
• Dropdown for all grades in all sales to quickly query another grade of a coin.
Bug fixes:
• NGC Matching Error
• Improved Matrix Scan
• Updated No Data Icons
• Updated Sales Data (Not all are fixed)
New Countries Added:
Australia
New Zealand
More improvements and updates coming soon!
So I’ve tried but can’t get a toning score because the TPG called it cleaned?
Was hoping the engine would scan the image and give me feedback irrespective of what the TPG said?
@treybenedict said: Update 1.1 is Live! Update on the app store here
Features include:
• Dual Sided Toning Analysis (Obverse/Reverse segmented and ran independently)
• Toning score is now tappable to use in both Letter Grades and Number Scores.
• Dropdown for all grades in all sales to quickly query another grade of a coin.
Bug fixes:
• NGC Matching Error
• Improved Matrix Scan
• Updated No Data Icons
• Updated Sales Data (Not all are fixed)
New Countries Added:
Australia
New Zealand
More improvements and updates coming soon!
So I’ve tried but can’t get a toning score because the TPG called it cleaned?
Was hoping the engine would scan the image and give me feedback irrespective of what the TPG said?
We are still gathering more data as to the AT/QC analysis. In a further update, we will push this!
I just checked out the latest version of the app.
I hope you can eliminate the need for 4 (or 5) different screens to show the color analysis. There seems to be plenty of room on 1 page and it can be quite useful. I could see myself taking a screenshot and using it as a second image upload within a registry set or showcase. If I want to show to a friend or potential purchaser of a coin what the app produces for a coin, I would currently need to share 4 screen shots when only one is needed.
What’s the obverse grade?
Oh, what’s the numeric score for the obverse?
What’s the reverse grade?
Oh, what’s the numeric score for reverse?
See, you have to share four screenshots for what could be viewed in just one screenshot. It would be good to think about how the information will be used and it would be nice to just have all the information on a single screen. There seems to be plenty of room to accomplish this.
Taking this one step further, I think we need an additional measurement. We need an overall score/grade. Using this coin as an example, I recommend using a 65% or 70% weighted score for the obverse and then a corresponding remainder weighted score for the reverse. There have been discussions about grading or strength of pricing (for toners) favoring obverses and something similar could be applied here.
The math in this example:
(78 x .65) + (76 x .35) = Combined Score of 50.7 + 26.6 = 77.3 (insert monster/t-rex here)
All 5 scoring pieces could be all on a single screen in a single box. The box can be made bigger because you can get rid of the obverse and reverse buttons that take up valuable real estate.
When all the data is on one screen then it becomes a valuable tool people will want to use, screenshot and share the images, be it among friends, in the registry/showcases or when trying to buy/sell a coin and make comparisons to other coins. This will make your app much more user friendly and popular.
A tiny step further would be the ability to rotate the phone and getting a horizontal screenshot of the analysis in the same aspect ratio of a True View. That way, when a person uploads a TrueView to their Instagram, the second image in the same post that gets uploaded is a screenshot from your app with the same aspect ratio. The third element in the Instagram post would be a horizontally oriented video of the coin. Your app then becomes even more popular because it is more user friendly.
Another thought is instead of people having to take a screen shot, they can click a tiny save button and it will download the screenshot to the person’s phone minus the cell phone time, wifi or cell phone signal and battery charge percentage etc.
Thanks for a neat app! A lot of work has gone into it and it is appreciated.
The analysis of this coin seems off. This coin is a monster and I am not talking about the reverse. The app seems to think the tiny sliver of reverse toning is much nicer/significant than the obverse.
Here’s another one that doesn’t right, showing a much stronger score for the obverse. Do you now see the value in having all analysis captured in a single screenshot? I wouldn’t have to share four images.
@keyman64 said:
The analysis of this coin seems off. This coin is a monster and I am not talking about the reverse. The app seems to think the tiny sliver of reverse toning is much nicer/significant than the obverse.
The obverse should in no way be a B-.
There’s a known issue with the scores being inverted for Obverse and Reverse we are looking into the fix for this.
@keyman64 said:
Here’s another one that doesn’t right, showing a much stronger score for the obverse. Do you now see the value in having all analysis captured in a single screenshot? I wouldn’t have to share four images.
Also yes. I totally see the value in having all scoring in one place.
Still trying to figure out the best way to proceed in terms of user experience. I have some ideas and would love to chat about them
_______________________________________________________________________________________________________________________________________
This is how I would lay out the results.
One screen, landscape, following @keyman64's suggestion.
This emphasizes that you provide a numeric color score for each side.
The % is a concise way to suggest the range is 0-100.
I don't know if that is the actual theoretical range, but we can see it goes to at least 78.
I can imagine "bragging rights" to the highest known score for a type.
Probably the [nFB] would not appear for the 1939.
If the coin was FB, FH, etc. that would appear to the right of the grade.
The die variety ID could go in that lower right corner position, when applicable.
When comparing these 2 coins, I personally like the 1939 "57" obverse more than the 1942 "73" reverse.
This is because I like the large areas of red and blue-green on the 1939.
Besides the potential subjectivity of assigning a color score,
I believe a factor which limits the usefulness of assigning color grades (that might reflect price above that of an untoned coin)
is that the grade might already encompass the eye appeal of the color pattern.
I have observed that several of the beautifully toned seated half dimes have risen in grade over time,
presumably to reflect prices realized and the changing preference of collectors for coins with attractive toning.
.Question for @treybenedict
I scan the following 5 coins into my inventory:
Cert. Pure Est.
42773188 - $1348.22
34791590 - $1937.82
06627143 - $982.23
37876088 - no data
40781467 - no data
1348.22 +1937.82 + 982.23 = $4268.27
App shows "Total Inventory is $6320.15 for 5 coins"
How is the $6320.15 calculated ?
thanks
edited to say I guess I figured it out, there was no PCGS-CAC data for 37876088 so the app used ALL ($2051.88)
and there was no data at all for the modern 40781467 so the app used $0.
P.S. I think the best way to choose a function that computes the color score from the pixel colors is to pick the one that is best at explaining observed prices, controlling for the other measurables like populations and grade (see below).
So I think it is best to do that statistical analysis first.
Maybe you already did that, but since you said your price model does not adjust for color, it sounds like you have not done this, or maybe you did most of it, but it is not finalized.
How about some econometrics.
If I was doing it, I'd do a linear regression of log(price) on:
log(2+population_higher) [I use log(2+pop_higher) because I don't want this to equal 0 when pop_higher = 0]
log(2+population_higher+population_at_grade)
dummies for each grade (constrain so that coefficients rise with grade)
dummy for FB/FH
dummies for PCGS, NGC, CAC
obv_color_score
rev_color_score
log(PCGS_3000_index) [to control somewhat for time effects]
and interactions of the above variables.
So you are looking for the highest R-squared value across specifications with different color score functions.
If you are varying the number of explanatory variables, then you compare SBIC instead of R-squared.
One of the practical problems is that there is a limited number of observations on prices for toned coins.
This limits the number of interacted variables.
This regression should be run separately by type of coin, which also limits the number of observations.
42773188 - $1348.22
34791590 - $1937.82
06627143 - $982.23
37876088 - no data
40781467 - no data
1348.22 +1937.82 + 982.23 = $4268.27
App shows "Total Inventory is $6320.15 for 5 coins"
How is the $6320.15 calculated ?
thanks
edited to say I guess I figured it out, there was no PCGS-CAC data for 37876088 so the app used ALL ($2051.88)
and there was no data at all for the modern 40781467 so the app used $0.
It will default to the "all" sales price if there is no data for a specific service. More data is being implemented as we speak so I hope that 40781467 will have some populated data within the next two weeks
@yosclimber said:
P.S. I think the best way to choose a function that computes the color score from the pixel colors is to pick the one that is best at explaining observed prices, controlling for the other measurables like populations and grade (see below).
So I think it is best to do that statistical analysis first.
Maybe you already did that, but since you said your price model does not adjust for color, it sounds like you have not done this, or maybe you did most of it, but it is not finalized.
How about some econometrics.
If I was doing it, I'd do a linear regression of log(price) on:
log(2+population_higher) [I use log(2+pop_higher) because I don't want this to equal 0 when pop_higher = 0]
log(2+population_higher+population_at_grade)
dummies for each grade (constrain so that coefficients rise with grade)
dummy for FB/FH
dummies for PCGS, NGC, CAC
obv_color_score
rev_color_score
log(PCGS_3000_index) [to control somewhat for time effects]
and interactions of the above variables.
So you are looking for the highest R-squared value across specifications with different color score functions.
If you are varying the number of explanatory variables, then you compare SBIC instead of R-squared.
One of the practical problems is that there is a limited number of observations on prices for toned coins.
This limits the number of interacted variables.
This regression should be run separately by type of coin, which also limits the number of observations.
The model for pricing color is still being built. I've run similar methods that you've mentioned and it is a good approach but is missing some key variables that are used for indication of rarity of toning. This is still under production, but I love how you are thinking it makes me happy to see.
@yosclimber said:
_______________________________________________________________________________________________________________________________________
This is how I would lay out the results.
One screen, landscape, following @keyman64's suggestion.
This emphasizes that you provide a numeric color score for each side.
The % is a concise way to suggest the range is 0-100.
I don't know if that is the actual theoretical range, but we can see it goes to at least 78.
I can imagine "bragging rights" to the highest known score for a type.
Probably the [nFB] would not appear for the 1939.
If the coin was FB, FH, etc. that would appear to the right of the grade.
The die variety ID could go in that lower right corner position, when applicable.
When comparing these 2 coins, I personally like the 1939 "57" obverse more than the 1942 "73" reverse.
This is because I like the large areas of red and blue-green on the 1939.
Besides the potential subjectivity of assigning a color score,
I believe a factor which limits the usefulness of assigning color grades (that might reflect price above that of an untoned coin)
is that the grade might already encompass the eye appeal of the color pattern.
I have observed that several of the beautifully toned seated half dimes have risen in grade over time,
presumably to reflect prices realized and the changing preference of collectors for coins with attractive toning.
Nice coins, by the way!
PCGS will not allow for the modification of TrueViews and I will respect their word on this.
@ProofCollection said:
Do you have a website where we can use this app? My phone screen is too small and my thumbs too big to do anything but make phone calls with it.
Of course PCGS would need to be happy with it.
I don't know if they would view this the same as a modified TrueView JPG.
I have their permission to modify their TrueViews in my seated half dime attribution guide.
I use their white background versions like the above.
The agreement is that I watermark each one with the text "photo provided courtesy of PCGS".
Of course, your usage is in a different "live" situation, while mine is static and imbedded in PDFs.
Of course PCGS would need to be happy with it.
I don't know if they would view this the same as a modified TrueView JPG.
I have their permission to modify their TrueViews in my seated half dime attribution guide.
I use their white background versions like the above.
The agreement is that I watermark each one with the text "photo provided courtesy of PCGS".
Of course, your usage is in a different "live" situation, while mine is static and imbedded in PDFs.
We are working on a share feature that can be exported to Instagram as well as to the forums that incorporate the majority of what you suggested. I'll display some templates and mockups later. The true view will stay unharmed.
The last 5 auctions (PCGS) shown below, if I throw the high, $780, the average for the other 4 is $501
See data below, seems that your ap is on the low side,, I will check some more. Thank you!
@treybenedict
Example#2, 1890 CC VAM-4 TB XF-40
Pure $247.35 (last auction was from 2014) - could be not looking for Tail Bar?
NGC Est price $480
Host, last auction for the XF-40 $480, price guide is $475 - Also auction pay an extra 18% to 22% on ave.
See eBay sale for an VF-25 that got $455 (are you pulling in eBay data?)
See data below:
@Joe_360 said: @treybenedict
Example#2, 1890 CC VAM-4 TB XF-40
Pure $247.35 (last auction was from 2014) - could be not looking for Tail Bar?
NGC Est price $480
Host, last auction for the XF-40 $480, price guide is $475 - Also auction pay an extra 18% to 22% on ave.
See eBay sale for an VF-25 that got $455 (are you pulling in eBay data?)
See data below:
Hey Joe!
Will Address this right now and see if that item was a Tail Bar. The updated pricing for the present value has not yet been implemented.
Thanks for your patience!
Also note: You are on the NGC tab, and showing PCGS results in the thread. Those results are under the PCGS tab in the app.
The last 5 auctions (PCGS) shown below, if I throw the high, $780, the average for the other 4 is $501
See data below, seems that your ap is on the low side,, I will check some more. Thank you!
Will check on this one as well.
EDIT: This is a problem with the model. Simple one-line fix. Apologies I see the problem here. Will be deployed in the 1.2 update.
Does the app combine results for VAM designated and non-VAM designated variants of the same coin? As in the example above, an 1890-CC Tailbar is known as VAM 4 (coin #87198) and just coin #133934.
@ProofCollection said:
Does the app combine results for VAM designated and non-VAM designated variants of the same coin? As in the example above, an 1890-CC Tailbar is known as VAM 4 (coin #87198) and just coin #133934.
Just made an Instagram and will update a feature to allow sales & toning reports to be generated as IG posts through the app. Follow us here: https://instagram.com/purenumismatics
Thanks Experimenting with design for the IG layouts so people can share their reports. Curious as to how it'll shape up but excited to be back on IG for coins!
Thanks Experimenting with design for the IG layouts so people can share their reports. Curious as to how it'll shape up but excited to be back on IG for coins!
I just followed you as well. If you like, you can look at how I have my feed set up. I would likely post a pure color analysis as a second image to keep the look of my feed consistent so I would create a pure graphic that includes the TV, and analysis including both the letter grade and numeric grades. As we know, all letter grades are not equal. Any third slide I might do would be a video, more than likely.
Thanks Experimenting with design for the IG layouts so people can share their reports. Curious as to how it'll shape up but excited to be back on IG for coins!
I just followed you as well. If you like, you can look at how I have my feed set up. I would likely post a pure color analysis as a second image to keep the look of my feed consistent so I would create a pure graphic that includes the TV, and analysis including both the letter grade and numeric grades. As we know, all letter grades are not equal. Any third slide I might do would be a video, more than likely.
Sounds great! I'm trying to figure out a way to fit both sides in one photo. & Yes! Letter grades are not always the same, I will include the number system when production hits. Going to look at your IG now.
Comments
Update 1.1 coming out on Monday.
Features include:
• Dual Sided Toning Analysis (Obverse/Reverse segmented and ran independently)
• Toning score is now tappable to use in both Letter Grades and Number Scores.
• Drop down for all grades in all sales to quickly query another grade of a coin.
Bug fixes:
• NGC Matching Error
• Improved Matrix Scan
• Updated No Data Icons
New Countries Added:
Australia
New Zealand
More improvements and updates coming soon!
Thanks
Would it be easier if you made "custom" photos have to have a solid black or white background?
Those that take their own pics can usually do a photo version of their coin with that type of solid background very easily. I know that, when I was doing my own pics, I could, and I am a neophyte compared to MANY on here that take pics.
I've been told I tolerate fools poorly...that may explain things if I have a problem with you. Current ebay items - Nothing at the moment
I’ve figured out a way it could work, but I want to test with some early users here on the forum. I’ll likely be reaching out to some of you guys soon!
As a result of the new toning analysis update. It will likely take 1-2 minutes to be able to generate a report for both the obverse and the reverse. I've tested numerous amounts of times and it seems to be all fine.
1.2 update already has plans for some great new features. Excited to share soon!
Remember that sharing with your friends helps us tremendously in our startup journey.
That is what I thought of when you said future value (economics) but was not sure if this was related to the previously mentioned ratio from other examples or referring to something else.
Okay (coming later) on the what the program estimates represent. Today it is more difficult to determine (at least for me) what an auction result actually represents due to todays variables. I use auction results occasionally to try and figure values if I am searching around or interesting in something but I do a lot of scrubbing of the data so to say for what I believe is my specific case. But even then I can not be sure and figure a range and this is for a single coin.
https://youtube.com/watch?v=2YNufnS_kf4 - Mama I'm coming home ...................................................................................................................................................................... RLJ 1958 - 2023
Wondering if circle cropping would work better, like this? That would eliminate background clutter like slab edges and other stuff.
10-4,
My Instagram picturesErik
My registry sets
It’ll be a form of something like this. There’s a couple good ways to automate the process to be able to give a good slab shot and a good finished crop.
Resolution then becomes a bigger issue. This photo is fine, but the one taken off my iPhone probably isn’t high quality enough.
1.1 Update Bumped to Tuesday.
Apple wanted further review about what “toning analysis” is haha!
Back under review and update coming tomorrow!
Need more help testing toning with Obverse/Reverse segmentation.
I'm searching for 2-3 people to do user interviews with.
Email me @ treyb@berkeley.edu if you'd like to be a part of the experimental process. Have goodies for you guys as well!
User interviews filled. Thanks for all that reached out.
Going to conduct 2-3 each week, maybe one per day from here on out until the next month.
Update still pending review. Also known error where some toning analysis have no graphs. Patch coming tomorrow.
I have scanned my slabs, the pricing seems to (off) toward the low side compared to resources such as the gray-sheet, 3rd party (website) reference and Redbook. I'll check again since you have updated your system. Also, old/outdated auction references.
Could you show me which certs? Will take a look
Update 1.1 is Live! Update on the app store here
Features include:
• Dual Sided Toning Analysis (Obverse/Reverse segmented and ran independently)
• Toning score is now tappable to use in both Letter Grades and Number Scores.
• Dropdown for all grades in all sales to quickly query another grade of a coin.
Bug fixes:
• NGC Matching Error
• Improved Matrix Scan
• Updated No Data Icons
• Updated Sales Data (Not all are fixed)
New Countries Added:
Australia
New Zealand
More improvements and updates coming soon!
@treybenedict
I'm trying out the app, so far, I like it
boston
Successful transactions with : MICHAELDIXON, Manorcourtman, Bochiman, bolivarshagnasty, AUandAG, onlyroosies, chumley, Weiss, jdimmick, BAJJERFAN, gene1978, TJM965, Smittys, GRANDAM, JTHawaii, mainejoe, softparade, derryb
Bad transactions with : nobody to date
Thanks! I'm glad to hear. Any feedback is much appreciated.
So I’ve tried but can’t get a toning score because the TPG called it cleaned?
Was hoping the engine would scan the image and give me feedback irrespective of what the TPG said?
We are still gathering more data as to the AT/QC analysis. In a further update, we will push this!
@treybenedict
I just checked out the latest version of the app.
I hope you can eliminate the need for 4 (or 5) different screens to show the color analysis. There seems to be plenty of room on 1 page and it can be quite useful. I could see myself taking a screenshot and using it as a second image upload within a registry set or showcase. If I want to show to a friend or potential purchaser of a coin what the app produces for a coin, I would currently need to share 4 screen shots when only one is needed.
What’s the obverse grade?
Oh, what’s the numeric score for the obverse?
What’s the reverse grade?
Oh, what’s the numeric score for reverse?
See, you have to share four screenshots for what could be viewed in just one screenshot. It would be good to think about how the information will be used and it would be nice to just have all the information on a single screen. There seems to be plenty of room to accomplish this.
Taking this one step further, I think we need an additional measurement. We need an overall score/grade. Using this coin as an example, I recommend using a 65% or 70% weighted score for the obverse and then a corresponding remainder weighted score for the reverse. There have been discussions about grading or strength of pricing (for toners) favoring obverses and something similar could be applied here.
The math in this example:
(78 x .65) + (76 x .35) = Combined Score of 50.7 + 26.6 = 77.3 (insert monster/t-rex here)
All 5 scoring pieces could be all on a single screen in a single box. The box can be made bigger because you can get rid of the obverse and reverse buttons that take up valuable real estate.
When all the data is on one screen then it becomes a valuable tool people will want to use, screenshot and share the images, be it among friends, in the registry/showcases or when trying to buy/sell a coin and make comparisons to other coins. This will make your app much more user friendly and popular.
A tiny step further would be the ability to rotate the phone and getting a horizontal screenshot of the analysis in the same aspect ratio of a True View. That way, when a person uploads a TrueView to their Instagram, the second image in the same post that gets uploaded is a screenshot from your app with the same aspect ratio. The third element in the Instagram post would be a horizontally oriented video of the coin. Your app then becomes even more popular because it is more user friendly.
Another thought is instead of people having to take a screen shot, they can click a tiny save button and it will download the screenshot to the person’s phone minus the cell phone time, wifi or cell phone signal and battery charge percentage etc.
Thanks for a neat app! A lot of work has gone into it and it is appreciated.
Greg
https://greatcollections.com/Collections/1120/The-Keyman64-Mercury-Dime-Collection/2024-07-07
The analysis of this coin seems off. This coin is a monster and I am not talking about the reverse. The app seems to think the tiny sliver of reverse toning is much nicer/significant than the obverse.
The obverse should in no way be a B-.
https://greatcollections.com/Collections/1120/The-Keyman64-Mercury-Dime-Collection/2024-07-07
Here’s another one that doesn’t right, showing a much stronger score for the obverse. Do you now see the value in having all analysis captured in a single screenshot? I wouldn’t have to share four images.
https://greatcollections.com/Collections/1120/The-Keyman64-Mercury-Dime-Collection/2024-07-07
There’s a known issue with the scores being inverted for Obverse and Reverse we are looking into the fix for this.
Also yes. I totally see the value in having all scoring in one place.
Still trying to figure out the best way to proceed in terms of user experience. I have some ideas and would love to chat about them
Just wanted to clarify there’s a known issue with Obverse and Reverse scores being inverted.
A patch for this will be fixed tomorrow apologies.
_______________________________________________________________________________________________________________________________________
This is how I would lay out the results.
One screen, landscape, following @keyman64's suggestion.
This emphasizes that you provide a numeric color score for each side.
The % is a concise way to suggest the range is 0-100.
I don't know if that is the actual theoretical range, but we can see it goes to at least 78.
I can imagine "bragging rights" to the highest known score for a type.
Probably the [nFB] would not appear for the 1939.
If the coin was FB, FH, etc. that would appear to the right of the grade.
The die variety ID could go in that lower right corner position, when applicable.
When comparing these 2 coins, I personally like the 1939 "57" obverse more than the 1942 "73" reverse.
This is because I like the large areas of red and blue-green on the 1939.
Besides the potential subjectivity of assigning a color score,
I believe a factor which limits the usefulness of assigning color grades (that might reflect price above that of an untoned coin)
is that the grade might already encompass the eye appeal of the color pattern.
I have observed that several of the beautifully toned seated half dimes have risen in grade over time,
presumably to reflect prices realized and the changing preference of collectors for coins with attractive toning.
Nice coins, by the way!
.Question for @treybenedict
I scan the following 5 coins into my inventory:
42773188 - $1348.22
34791590 - $1937.82
06627143 - $982.23
37876088 - no data
40781467 - no data
1348.22 +1937.82 + 982.23 = $4268.27
App shows "Total Inventory is $6320.15 for 5 coins"
How is the $6320.15 calculated ?
thanks
edited to say I guess I figured it out, there was no PCGS-CAC data for 37876088 so the app used ALL ($2051.88)
and there was no data at all for the modern 40781467 so the app used $0.
Successful transactions with : MICHAELDIXON, Manorcourtman, Bochiman, bolivarshagnasty, AUandAG, onlyroosies, chumley, Weiss, jdimmick, BAJJERFAN, gene1978, TJM965, Smittys, GRANDAM, JTHawaii, mainejoe, softparade, derryb
Bad transactions with : nobody to date
P.S. I think the best way to choose a function that computes the color score from the pixel colors is to pick the one that is best at explaining observed prices, controlling for the other measurables like populations and grade (see below).
So I think it is best to do that statistical analysis first.
Maybe you already did that, but since you said your price model does not adjust for color, it sounds like you have not done this, or maybe you did most of it, but it is not finalized.
How about some econometrics.
If I was doing it, I'd do a linear regression of log(price) on:
log(2+population_higher) [I use log(2+pop_higher) because I don't want this to equal 0 when pop_higher = 0]
log(2+population_higher+population_at_grade)
dummies for each grade (constrain so that coefficients rise with grade)
dummy for FB/FH
dummies for PCGS, NGC, CAC
obv_color_score
rev_color_score
log(PCGS_3000_index) [to control somewhat for time effects]
and interactions of the above variables.
So you are looking for the highest R-squared value across specifications with different color score functions.
If you are varying the number of explanatory variables, then you compare SBIC instead of R-squared.
One of the practical problems is that there is a limited number of observations on prices for toned coins.
This limits the number of interacted variables.
This regression should be run separately by type of coin, which also limits the number of observations.
It will default to the "all" sales price if there is no data for a specific service. More data is being implemented as we speak so I hope that 40781467 will have some populated data within the next two weeks
The model for pricing color is still being built. I've run similar methods that you've mentioned and it is a good approach but is missing some key variables that are used for indication of rarity of toning. This is still under production, but I love how you are thinking
it makes me happy to see.
PCGS will not allow for the modification of TrueViews and I will respect their word on this.
Do you have a website where we can use this app? My phone screen is too small and my thumbs too big to do anything but make phone calls with it.
Coming end of next week!
For those using the app right now. GitHub went down
Some parts may be broken due to the outage. Fixing ASAP when GitHub comes back online.
Everything is fixed now.
If you had run split analysis, please run again and it should be corrected
________________________________________________________________________________________________________________________________
If you put the overlay text in a separate transparent image,
you should be able to overlay it in a browser without modifying the original TrueView JPG.
https://stackoverflow.com/questions/20276107/how-can-i-overlay-two-images-in-html5
Presumably you can do something similar in an app.
Of course PCGS would need to be happy with it.
I don't know if they would view this the same as a modified TrueView JPG.
I have their permission to modify their TrueViews in my seated half dime attribution guide.
I use their white background versions like the above.
The agreement is that I watermark each one with the text "photo provided courtesy of PCGS".
Of course, your usage is in a different "live" situation, while mine is static and imbedded in PDFs.
We are working on a share feature that can be exported to Instagram as well as to the forums that incorporate the majority of what you suggested. I'll display some templates and mockups later. The true view will stay unharmed.
@treybenedict
Pricing, example #1, 1883 CC Morgan
See data below, seems that your ap is on the low side,, I will check some more. Thank you!
@treybenedict
Example#2, 1890 CC VAM-4 TB XF-40
Pure $247.35 (last auction was from 2014) - could be not looking for Tail Bar?
NGC Est price $480
Host, last auction for the XF-40 $480, price guide is $475 - Also auction pay an extra 18% to 22% on ave.
See eBay sale for an VF-25 that got $455 (are you pulling in eBay data?)
See data below:
Hey Joe!
Will Address this right now and see if that item was a Tail Bar. The updated pricing for the present value has not yet been implemented.
Thanks for your patience!
Also note: You are on the NGC tab, and showing PCGS results in the thread. Those results are under the PCGS tab in the app.
Will check on this one as well.
EDIT: This is a problem with the model. Simple one-line fix. Apologies I see the problem here. Will be deployed in the 1.2 update.
Does the app combine results for VAM designated and non-VAM designated variants of the same coin? As in the example above, an 1890-CC Tailbar is known as VAM 4 (coin #87198) and just coin #133934.
Understood, but I include the estimated NGC pricing set at $480. So, our host and NGC are the same. Thanks, Joe
It does combine them to one.
Fixed the problem. This coin slipped through and was messing up the price. Should be updated now.
Thanks so much for catching these two bugs!
Pushing some updated sales data later today. Will update you guys when we do!
Just made an Instagram and will update a feature to allow sales & toning reports to be generated as IG posts through the app. Follow us here: https://instagram.com/purenumismatics
Just followed you out there!
10-4,
My Instagram picturesErik
My registry sets
Thanks
Experimenting with design for the IG layouts so people can share their reports. Curious as to how it'll shape up but excited to be back on IG for coins!
I just followed you as well. If you like, you can look at how I have my feed set up. I would likely post a pure color analysis as a second image to keep the look of my feed consistent so I would create a pure graphic that includes the TV, and analysis including both the letter grade and numeric grades. As we know, all letter grades are not equal.
Any third slide I might do would be a video, more than likely.
https://greatcollections.com/Collections/1120/The-Keyman64-Mercury-Dime-Collection/2024-07-07
Sounds great! I'm trying to figure out a way to fit both sides in one photo. & Yes! Letter grades are not always the same, I will include the number system when production hits. Going to look at your IG now.
Thanks!