50g: an interesting RAND anomaly

[DavidM]

(03-17-2018 06:41 PM)Claudio L. Wrote:  ...It seems you found that the last 3 bits (hard to define bits on a decimal number, but bear with me) have no bias and good uniform distribution (you got a nice 50/50/50/50), but have a short period, and you even happened to find out what that period is.

Using a different nibble will give you different properties, not necessarily any better.

I actually convert the native BCD seed to binary form before taking the MOD in the Saturn version, so it does actually end up being the last few bits in this case. And I'm obviously no expert in PRNGs, either, but I believe your assessment is spot on.