Information on calculator Random Number Generators from PPC Journal articles

[Dieter]

(11-13-2016 07:27 AM)Namir Wrote:  Regarding the range of Chi-Square values, the results are:

Alg. 1 has the range (2.878, 16.79)
Alg. 2 has the range (7.48, 24.206)
Alg. 3 has the range (8.302, 14.976)

Thus Algorithm 1 seems to yield better distributed numbers. Algorithm 2 seems to yield the worst distributed numbers.

FTR, I have done a few hundred runs with 10.000 random numbers each, using the 9821 and (x+pi)³ RNG. The Chi² values came out very similar. For instance, a substantial part for the third RNG is below 5, values like 2 or 3 appear quite often, sometimes even less. That's why I do not think that we can determine this way if one of these two RNGs is "better" than the other.

For the R-D generator I got Chi² values that sometimes were significantly higher (like 30 or 40, in single cases even much more). So this seems to be not on par with the other two. But honestly... for calculator games: who cares?

Dieter