> On Oct 17, 2016, at 1:10 PM, Kelly St. Clair <xxxxxx@efn.org> wrote:
>
>  And when computers WERE brought to bear on the issue, the programs used were primitive and error-prone; consider the first Atlas of the Imperium, and its notoriously bad dataset, with whole sectors of bad output.

That was fixable even then; the Apple BASIC RAND() function was actually pretty good.  (iirc an Apple II was what was used to generate the Atlas) IF you didn't commit the rookie error of re-seeding it with the same value every time :-/ Also I do believe that someone sold a card that sampled background radiation for a true random seed. I know I considered getting one at one point, just couldn’t justify the $150-ish cost.

I did a bunch of rand() testing with my ][+ back in the day; I repeatedly ran out to (iirc) 500,000 rolls of 2 and 3D6 and got a nice smooth distribution curve that looked properly bell shaped. Took 4 or 5 hours to do each set, of course. A 1kHz clock speed’ll do that for yah :-)

--
Bruce Johnson
University of Arizona
College of Pharmacy
Information Technology Group

Institutions do not have opinions, merely customs