A Better Random Number Generator for Apple's Floating Point BAS

Random number generators (RNGs) are functions which produce pseudo-random numbers. Usually, the numbers produced are fractions between 0 and 1. Ideally, every fraction a computer language can represent will be as likely to be generated as every other by that language’s RNG, and the order of numbers will be unpredictable to the user. Slightly more formally, the RNG should produce sequences of numbers which, so far as standard statistical tests can tell, behave in the same way as “truly random” number sequences would behave. RNGs are used to simulate imperfectly predictable real life events (computer games use them, for example), to provide test data for input to complex computer programs (often exposing bugs that a series of test inputs missed), to evaluate mathematical functions or solve equations for which no theoretical solution exists, or similarly, to provide estimates of a solution against which a new theoretical solution can be checked. Randomization, or randomly rearranging the order of things, is the basis of a wide class of increasingly popular, very sensitive statistical tests. Randomization of the order of events in experiments is a necessary part of the design of every experiment that we have run. The better the generator, the better the simulation, solution, test, or experimental control.