3 ms·
For large enough runs, 0, 1, 2, ..., max, 0, 1, ... etc exhibits that behaviour. Thus, a PRNG that fails your proposal is bad, but passing it doesn't meanthat
by dbaupp 13y ago
For large enough runs, 0, 1, 2, ..., max, 0, 1, ... etc exhibits that behaviour. Thus, a PRNG that fails your proposal is bad, but passing it doesn't meanthat the PRNG is good. A test suite like diehard (or the moderner version dieharder) runs tests for uniform distribution as well as a lot of others, but even then, a pass is only saying "any glaring nonrandomness can't be detected by this suite".
- adsr 13y agoThanks, that's true but my comment was mostly a response to the "three million zeroes and a single four" as a valid random sequence comment. I wasn't suggesting it as a generic method to test the quality of a PRNG.