A bell-curve random distribution function?
I thought I had come up with an approximate normal-distribution substitute for Scratch's linear random but when I added a frequency graph to plot the results it became clear that it was NOT a bell-curve :-( I've asked in the AT forum for better suggestions. Preferably something that doesn't do an arbitrary number of rand() calls per result.
HERE'S THE BEST SOLUTION SO FAR:
My attempt took two factors in the range of 0..sqrt(high) and multiplied them together. It turned out to be two back-to-back sqrt() curves rather than a normal curve.
It's quite surprising that Scratch has been around this long without a gaussian distribution random function available, They're tremendously useful in all sorts of applications.