Maybe I'm the bot.
Also, ninja'd by Josh Himself on how pseudorandom number generators work.
Just to add to that, though, it would actually be perfectly OK for a random number generator (RNG) function to produce 0 or 1 as its first result, as long as the many hundreds or thousands of subsequent calls to the RNG return different and impossible-for-a-human-to-guess result numbers.
As long as your program doesn't insert a new seed value in between calls to the RNG, then you get two useful results:
1. Each subsequent number is virtually impossible to predict from the previous number. (This is what makes these numbers appear to be "random.")
2. As long as you start with the same initial seed, the sequence of random numbers generated from each call to the RNG will be exactly the same each time.
These two features are what allow every new system generated from a different starting seed value to look very different (because the pseudorandom numbers are different on each run), but for every system generated from the same seed value to look identical (because the
sequence of random-looking numbers is always the same).
Here's a version of the Park-Miller & Schrage RNG method I use:
Code: Select all
seed = 123459876 + Date.getSeconds() * 0xFFFFFF + Date.getMinutes() * 0xFFFF;
A = 16807;
M = 2147483647;
Q = floatFix(M / A, 0);
R = M % A;
oneOverM = 1.0 / M;
mask = 123459876;
seed ^= mask;
hi = floatFix(seed / Q, 0);
lo = floatFix(seed - hi * Q, 0);
seed = A * lo - R * hi;
if (seed < 0)
seed += M;
result = oneOverM * seed;
seed ^= mask;
I defy you to eyeball that and tell me whether 0, or 1, or 63285462855, or (the real problem) multiple input seed values will all produce the same result number.
Surely you guys want to see some AI and weapons and stuff
Still no idea what floatFix is though. I see INT_MAX and a high power of seven in there though, which I'm not particularly fond of (why not primes?) But anyway. Probably better than rand()'s LCG.
