Reply to thread

Message: <blockquote data-quote="aramis erak" data-source="post: 9480724" data-attributes="member: 6779310">A properly designed PRNG (Pseudo-Random Number Generator) and a properly designed die roller to use it will have a slight bias; the PRNG itself is never quite truly random, and so introduces a slight bias. The math to simulate a die also introduces a slight bias for at least some sizes...the ones I have written have tiny errors from the math for dice under d1000, dRes = 1+(randint_16 modulo dSides)... A randint_16 is a 16 bit output (0-65535), and so a d1000 has 66 chances each of 0-535, and 65 each of 536-999. we add one to the result to get the expected 1-1000 result.  For smaller dice, it's even less of an error. The PRNG adds way larger error than the math to turn the PRNG result into a die result.My current one, I just trusted the Python random module, then verified with large chi² test data - it's more random than my dice. Except for one player. He genuinely averages half a sigma below expected average. He also thinks all die rollers hate him. And comparing his results to everyone else's? Yeah, they do.</blockquote>