Gregory Benford popularized the observation that sets of numbers which span multiple orders of magnitude exhibit digits proportionally to their logs. This has become a basis for determining non-expected digit representation distributions, and has been associated with fraud detection.

I observed that if one calculates the benford coefficients over a range of values, base 2 to bse 36 being convenient in java, one would have a tool for not only detecting the presence of non-expected digit representation, but one would have an indicator of which numbers were the most likely contributors to the non-expected digit representations. The stinkers would be present over a higher number of multiple base values, and over a higher number of their digits. One would score a number as to it's normanity, and those numbers which were not truely from a uniform distribution over several magnitudes would score higher. One thus has a useful indicator of an entity being fradulent.

I have intentions of folding this into the webGLTutorials. Progress to date is that I have calculated the benford coeficients for base 2 thru 36, and for digits the 1'st thru the 7'th. The resulting set of numbers makes a nice collum grid array, resembling a 1/X distribution. I'm using it as my demo object to allow mouse pointer designation for movement and rotation of a rendered object scene.

project is being developed by Thomas P. Moyer