2012-03-24, 01:47 PM
bast963 Wrote:What I meant was that if the RNG is really random, and has an equal chance to roll between 0 to 100 on every roll, or if it's pseudo and it ends up picking a favored side and an unfavored side and switching them every X amount of rerolls like in tetris. Like the PRNG would pick (between A, B, C, and D) the C as the favored side, the B as the "slightly higher than usual" side, the A as ignored, and the D as the normal 25% chance of appearance, and keep this for like 10 rolls and then switch the sides for the next 10 rolls etc. Whereas a regular RNG would just have 25% chance to all four letters every roll.
There's probably a bigger argument to this that I'm not seeing here, isn't there?
A true RNG only exists conceptually. A computer uses no such thing. All things in a modern computer are deterministic and white box. Computers absolutely cannot be random. That's why only PRNGs make sense on computers. A true RNG may be possible with quantum computing, but that too only exists in concept at the moment.
All PRNGs that are worth anything are tested for randomness (that all numbers are suitably different from each other) and biases (that certain numbers in the output RNG are not prone to be certain numbers). For an example of a stream that cannot be a PRNG, look no further than RC4. The second and third bytes are heavily biased toward zero, so using RC4 is not suitable for a PRNG. However, you could use AES as a PRNG if you wanted to because all numbers are sufficiently random and have very small bias.