Predicting The Cut Of Fancy Diamonds For Your Diamond Appraisals
Fancy cut diamonds (or non-rounds) have differences in the quality of their cut; however no independent lab grades with that in mind.
I have been developing statistical models that try to incorporate the quality of cut into the pricing models that I already have.
Round diamonds have establish cut parameters as laid out by the GIA. These parameters have been well studied and established. The jewelry industry by and large accepts the research that GIA has completed. Here is the most in-depth paper I have read on the cut of round diamonds. Diamond appraisals for round diamonds often note the quality of the cut.
Fancy diamonds, however, are tricky. Remember fancy cut diamonds are any non-round diamond including heart, emerald, pear, oval, marquise, asscher, radiant or princess. GIA has not release standards by which we can grade the quality of the cut for these non-round diamonds. My research has shown that cut is a very important determinant of how expensive a diamond is despite not being specified by GIA. With that information, it is difficult to complete a diamond appraisal accurately.
Excluding cut as a predictor of how expensive fancy cut diamonds are means the estimated diamond appraisal values that I give are not perfectly correct. This is one reason why I give a range. Including it, however, has been difficult. I have developed models that include variables similar to cut such as polish or symmetry but find them be any more accurate, yet significantly more confusing.
Now, I have been working on a model that predicts cut then includes the predicted cut variable in another model that prices the price you should receive. From doing that, I have learned two things. First, cut is really hard to predict without seeing the diamond. A trained gemologist can tell a well-cut fancy diamond quickly; a computer model cannot. Second, depth is by far the most important determinant of cut, followed by symmetry. Table, length to width ratio, and polish follow up last. Click here to see the graph showing the variable importance plot from a random forest model. Depth is obviously furthest to the right indicating meaning it is the most powerful predictor.