Determining True Value of Diamonds!

[valeria101]

Date: 2/13/2006 11:21:30 PM
Author: Richard Sherwood


Get a copy of Rapaport and input the different deductions or premiums for fluorescence, and then get the DiamCalc program and input the different deductions or premiums for the different cut categories.

By the time you finish you will no longer be a technician/statistician only, but a diamantaire coming out of the cocoon as well. A weary one, granted, but one just the same...

Wouldn''t the results of this be more ''what should be'' than what is out there? ...
For example... the simpler AGA cut grades do not seem to correlate with prices all that well. Perhaps one could wish they did, but... perhaps that grading system is not widely recognized enough to become an effective price coordinating tool. Like GIA grades have. Or the Rap.

Speaking of AGA and prices: it does seem to relate to prices (taking the database here) in a similar way other normative scores I work with relate with descriptive statistics: the extremes work allot better than the mid range - i.e. the lowest and the highest scores tend to be effective price factors (to some extent, sometimes not even), but nothing in-between does.


[disclaimer: this isn''t saying something bad about AGA - quite on the contrary in fact, and I would be happy to elaborate on that, as an admirer of the system. This is only a trivia matter of empirical research.]

 

[Paul-Antwerp]

I made the effort of quickly digesting your list of example-stones.

Number 21 and 22 are not overpriced, these prices are typing mistakes.

All in all, I had the impression that number 16 might have some hidden value, while number 3, 4, 5, 6, 9 and 11 are overpriced in the sense, that they are very hard to sell.

My constructive criticism is that the basic philosophy of the tool is incorrect. Not all diamonds are equal if you only consider weight, colour and clarity. If an ugly stone has a GIA-report, it does not make it less ugly than the beautiful stone without a report. You are entering depth and table, while these do not seem to have any effect on the recommendation. Cut is far from being entered into the model.

Extreme fluorescence has no impact, while it could make a difference of 40% in value.
Extremely large culet has no impact, while it could make a difference of 70%.
Depth percentage has no impact, while it could make a difference of 30%.

Somehow, I think that the model as such has a circular formula. You are entering a whole database with parameters, prices and so on, but I cannot see how you are able to put a resulting value on each stone. If you are refering back to price as a comparative price of stones with the same parameters, this is not value of that stone. You do not have have sufficient data to deduct a value for a stone.

The basis of this criticism is not on the fact that it does not work for the best 5% of the stones. It just does not work, period.

Live long,

 

[denverappraiser]

Why don''t you simply share your formula''s with us? You seem to be asking for advice on how to tweak your model to reflect the appropriate variables and how they should relate to one another. You might find more constructive criticism of the model if you gave a bit more information and didn''t require people to desconstruct the thing based on your web interface as Paul has started to do.


Neil Beaty
GG(GIA) ISA NAJA
Professional Appraisals in Denver

 

[JohnQuixote]

AA,

It's never easy to have your work criticized. Your polite and inquisitive tone is appreciated.

2 cents: You can overcome lab 'bias' by incorporating Rich Sherwood's suggestions... Of course, then it's just Rich's lab valuation and not yours (though my valuations agree with his).

To me the prime concern with your model is the tangibles it excludes.

Cut is now graded by both the AGS and GIA. GIA has some issues to work out, but perhaps your model would function better - at least for consumers who are becoming more cut-oriented - by incorporating a cut grade as well as color/clarity/polish/sym.

Perry suggested you review the HCA, which has been a reliable tool for rejection of undesirable proportions combinations. If you are not going to incorporate 5 measurements, especially the all-important crown and pavilion angles, in your pricing metric - perhaps the simple incorporation of a cut grade would help.

 

[Richard Sherwood]

Date: 2/14/2006 9:49:20 AM
Author: valeria101

Date: 2/13/2006 11:21:30 PM

Author: Richard Sherwood

Get a copy of Rapaport and input the different deductions or premiums for fluorescence, and then get the DiamCalc program and input the different deductions or premiums for the different cut categories.

By the time you finish you will no longer be a technician/statistician only, but a diamantaire coming out of the cocoon as well. A weary one, granted, but one just the same...

Wouldn't the results of this be more 'what should be' than what is out there? ...

For example... the simpler AGA cut grades do not seem to correlate with prices all that well. Perhaps one could wish they did, but... perhaps that grading system is not widely recognized enough to become an effective price coordinating tool. Like GIA grades have. Or the Rap.

Speaking of AGA and prices: it does seem to relate to prices (taking the database here) in a similar way other normative scores I work with relate with descriptive statistics: the extremes work allot better than the mid range - i.e. the lowest and the highest scores tend to be effective price factors (to some extent, sometimes not even), but nothing in-between does.

[disclaimer: this isn't saying something bad about AGA - quite on the contrary in fact, and I would be happy to elaborate on that, as an admirer of the system. This is only a trivia matter of empirical research.]

Hi Ana. The DiamCalc lists several cut grades systems other than the AGA in helping understand cut. AGS, GIA, HRD, Russian, etc.

I do find the AGA system to be very realistic though in relating to pricing structure. The DiamCalc program incoporates it nicely into a realistic discount structure for varying degrees of cut.

If AA could input a similar structure into his software it would go a long way towards making his data more realistic.

Something along the lines of:

Super Ideal Make (Ideal make with H&A optical symmetry & outstanding Idealscope/ASET)
Ideal Make (AGA 1A or AGS 0 or equivalent cut grading system)
Premier Make (AGA 1B...)
Fine Make 1 (AGA 2A...)
Fine Make 2 (AGA 2B...)
Commercial Make 1 (AGA 3A...)
Commercial Make 2 (AGA 3B...)
Below Average Make 1 (fair make) (AGA 4A...)
Below Average Make 2 (poor make) (AGA 4B...)

The DiamCalc sofware gives a realistic discount for these descending grades of make. The AGS system could be used, or the GIA, just so long as some system is used.

 

[Research]

Neil Beaty, in regard to your question about the formula. It is actually a combination of formulas as I believe was stated previously.

However I can tell you that in general, after Carat, clarity tends to be next most important, and about twice as important as color. Difference of depth from ideal depth is also a fairly important variable. Girdle, culet, flr tend to be less important. Again these differ depending on what type of diamond you are looking at. And we looked only at the carat ranges we specified.

It seems though that price of larger diamonds can be predicted more accurately than smaller ones, indicating that there is a lot more unexplained price variance in smaller diamonds. it would have been fun to look at stones over 6.0 carat, but we had much less sample here.

-AA

 

[Paul-Antwerp]

I understand that clarity is your main indicator, colour your second, depth your third, and so on.

Now, take two GIA-EX-EX-stones, medium girdle, pointed culet, no fluorescence, both 1.01 Cts.

Stone 1, E-VS2, with 64.7 depth and 65 table
Stone 2, F-VS2, with 60.5 depth and 57 table

If both stones are priced 6000$, apparently 1 is better according to your model.
If the second stone is priced 5998$, your model prefers stone 2.

In reality, stone 1 is between 15 and 30% less valuable, because of its horrible cut.

Sorry for sounding a bit harsh in my criticism. I appreciate your openness and your willing ear. But as long as this model, which does not work, is available on the net, it will lead some consumers into purchasing the wrong stone, and possibly overpaying enormously for it. Therefore, you should remove it.

Live long,

 

[Research]

Paul from Antwerp, the problem with your example is that it is an unrealistic one. Certainly stone 1 is a very bad cut. In fact so bad, that I was not able to find a stone with a depth and table anywhere near those numbers. Therefore it is a moot point, the model looks at actual retail diamonds on the interent. I assume an e-VS2 would have either been cut better, or that this combination between depth and table is not even possible/probable.

The model proved to be a statistically good fit the first day in terms of predicting actual retail internet prices very accurately. As to how these are set, that is something you and your colleagues are responsible for. But if you are to give examples, please do use real stones we can find on the web, not made up ones.

Also, as explained earlier, this is indeed a screener tool. Obviously other details need to be considered, and using the tool does not mean one should not educate oneself about diamonds as much as possible first. Though I wonder if that is what you want your customers to be?

Sorry for sounding harsh, but it seems you are too eager to disprove something by any means possible.

 

[denverappraiser]

OK, so it’s a complicated formula or set of formulas. We can handle it. It’s the root of your model. You are taking a set of attributes from the user and comparing them with some stones that you consider similar that are supposedly being offered for sale somewhere online. This is a reasonable enough approach but the details of how that comparison is being made are pretty important to deciding if it’s appropriate. Don’t you agree?

Neil Beaty
GG(GIA) ISA NAJA
Professional Appraisals in Denver

 

[Paul-Antwerp]

Date: 2/17/2006 8:34:17 AM
Author: Research

Paul from Antwerp, the problem with your example is that it is an unrealistic one. Certainly stone 1 is a very bad cut. In fact so bad, that I was not able to find a stone with a depth and table anywhere near those numbers. Therefore it is a moot point, the model looks at actual retail diamonds on the interent. I assume an e-VS2 would have either been cut better, or that this combination between depth and table is not even possible/probable.

The model proved to be a statistically good fit the first day in terms of predicting actual retail internet prices very accurately. As to how these are set, that is something you and your colleagues are responsible for. But if you are to give examples, please do use real stones we can find on the web, not made up ones.

Also, as explained earlier, this is indeed a screener tool. Obviously other details need to be considered, and using the tool does not mean one should not educate oneself about diamonds as much as possible first. Though I wonder if that is what you want your customers to be?

Sorry for sounding harsh, but it seems you are too eager to disprove something by any means possible.

Hey, no problem.

Sorry to say, but such a stone is technically possible.

You pointed out that depth is a pretty important parameter in your model. I just showed that it makes absolutely no difference. If the model does not recognize that 64.7 depth has less value, how can you expect it to see that 62.5 depth has?

Live long,

 

[Research]

Neil,

Sort of, we took a sample of just over 44,000 internet diamonds, and observed the relationships between the various variables on price. Statistically we have already confirmed that it is a sound model. And we have tested it successfully as a screening tool (we have 2 people that have purchased diamonds using it this way who are very happy). Though they obviously considered other factors as well. However, we do believe that feedback is a good thing and that improvement is always possible.

For instance, the depth and table variables mentioned earlier. Our approach/assumption was to take the absolute value of the difference between whatever the values are of the diamonds being entered into the tool, and the table value of 54.5 and depth value of 60.5. We understood that slight variations from these two points are acceptable, but too far and they start affecting price negatively (though we never saw any values as crazy as the ones Paul came up with).

So for instance, are the “ideal” - “average” values for width and table appropriate?

 

[denverappraiser]

Here are 2 stones being offered by the same dealer from the pricescope database.

Stone A
http://www.whiteflash.com/diamonds/diamond_Details.aspx?itemcode=1971882
Stone B
http://www.whiteflash.com/diamonds/diamond_Details.aspx?itemcode=AGS-5884901

Not surprisingly, your model prefers the first one. Many people here would disagree with this preference but I’ll be happy to call it a matter of taste. It still prefers stone A even after I raise the price by $2000! Do you feel that a user can correctly conclude that this is a great bargain, that stone B is a terrible rip off, or both?

Neil Beaty
GG(GIA) ISA NAJA
Professional Appraisals in Denver

 

[denverappraiser]

Neil Beaty
GG(GIA) ISA NAJA
Professional Appraisals in Denver

 

[Research]

"Many people here would disagree with this preference but I’ll be happy to call it a matter of taste"

Could you explain, if this is sarcasm it''s lost on me, sorry. Are you saying that B is clearly a better deal, or are you saying it is just a matter of taste?

Also, what woudl you value each of these at $ wise?

 

[denverappraiser]

Actually, that wasn't being sarcastic. There's a huge taste component and I didn't want to convert this into a discussion of what constitutes a good cut. There's plenty of other threads on that topic. Other than cut and choice of grading lab, the stones look to be superficially similar. Stone A isn't way out on the fringes like Pauls example and these stones differ in price by 50% from the same dealer. I think most the people here would solidly agree that stone B is 'better' and that this they are willing to pay the premium to get it even when they are fully aware of both offers.

I don't assign values on stones that I haven't seen and without clearly defining what market is being discussed.

Do you stand by the position that stone A is a better deal, even at a $2000 higher price?

Neil Beaty
GG(GIA) ISA NAJA
Professional Appraisals in Denver

 

[WinkHPD]

Date: 2/13/2006 4:55:15 PM
Author: Research

David,

We understand from professionals we have spoken too that the average consumer normally can not tell a glass diamond from a real one, never mind a really good one from a bad one. Obviously no model can take the place of the trained eyes of a professional looking at a diamond through a loop or under a projecting microscope. Instead effort is intended to give the average consumer another tool in order to feel more comfortable making a purchase which for most is the only diamond purchase they will make in their life, and which represents a considerable financial investment.



This is the intent behind this effort, a screening tool as you say, and your input is appreciated.

-AA

While I can appreciate your intent, it is obvious that you also know little or nothing about diamonds and the way they are priced and sold in the real world, therefore I find it not surprising that you have gone so far from bad to worse.

AGS may comprise only a small part of the diamond certificates available in the market, but only because it is relatively new. It is however highly sought after by those who wish to buy well cut diamonds that will look better than the average cut diamond. It has been successful and accepted by the trade from day one because of its high standards and its transparent peer reviewed scientific studies. To lump it with other or none is ridiculous and makes you look very foolish to the very people you are asking to bless your project here.

It is precisely because. "no model can take the place of a the trained eyes of a professional" that you should not make a tool that gives absolutely wrong advice and post it to those non professionals as a value judging tool. The sheer absurdity of your attempting to provide such a poorly designed "tool" leads me to believe that those who sell only GIA certs will gladly post links to it so their stones can be proclaimed possibly underpriced when ever compared to another diamond without a GIA certificate. Your "tool" then will be used to actually defraud those very people you are claiming to want to help.

Please come and play with us for a while, learn a little about what you are wanting to do, then design a tool that has value. Then perhaps you can begin to design something that has some merit rather than this currently farcical abomination.

Wink

P.S. This is in no way a condemnation of GIA or GIA Diamond Grading Reports. Gia has set the standards for generations and hopefully will continue to be a valuable leader and partner to jewelers and the jewelry industry. This is only a condemnation of this very poorly thought out "tool".

 

[Regular Guy]

Caught up with this discussion today. I''d like to share a few observations, but also a question:

I''ve trained in both sales and college publishing, which -- as I''ve heard it described -- makes my approach to info a mile wide and an inch deep. On this board, that could qualify me as just a troublemaker. Instead, I call myself Regular Guy. I think that works.

In college sales, I became aware of an approach to statistical analysis that''s been called: Resampling. It''s described here. (Click on the link to go to the link).

Is this, do you think, similar to your approach? I''ve really only been cursorily introduced to it, but the idea seems to be that, if given enough data, you don''t need to make very many assumptions, but just follow it''s natural grain.

Now, my observation(s). In the introduction to your model, you write:

"...This allows you to compare two diamonds even if they are totally different, and determine which is a better value for the price. You can also compare 2 diamonds from different stores and understand which store is offering better values....This model is intended to help you make a purchase decision by helping you understand the relative value of diamonds and their attributes."

It seems from the flack you''ve already gotten here, that your proactive decision to move from, in the best of cases, an observation about the characteristics of diamonds at certain price points, to a recommendation about a shopper''s approach to valuation, and that stretch, until better educated, may be too far.

Taking Neil''s most recent, still unchallenged recent example for a point of discussion...Neil actually understates his case. In your model, not only would you value the more generally poorly regarded diamond better at a lower price point...you would value that "generic" one carat as a better value for even as much as $143 more in cost. So, your model shows to purchase the "generic 1.02 carat" diamond at $7778 over the...yes...clearly better diamond at its real price of $7635. Never mind that the generic diamond actually costs $5390, is also available through the same vendor, and if you follow my protocol under my signature, you''ll see I readily recommend shoppers to make exactly the kind of comparison, in the real world, that Neil describes, anticipating that many of these shoppers will still go for the option at $2400 more (let alone $143 less).

A realistic question, then...allowing there may be some wisdom to the "random walk" of purchasing as described by the model you''ve tried to capture...can you tweak it, goose it up with knowledge from experts, such that it will still conform to its real world behavior, after providing the enhancements necessary to make an expert feel as though your judgements are reasonable.

That''s much too technical for me. I hope it won''t be for you, because, failing the meeting of this challenge, you''ve also received another suggestion from Paul here, which is "you should remove it."

Clearly, the intention of your model is to bring help, and not cause harm. Hopefully you can demonstrate an ability to do this.

Regards,

 

[Richard Sherwood]

Kind of reminds me of the Hippocratic oath physicians take:

"First, do no harm..."