Podfrey wrote:Sorry Mike, but no. The vulnerability in itself is a "bad thing", but the bigger issue is that the Random Number Generator (RNG) it ISN'T random.
<long sigh>
Podfrey wrote:There are far too many times when the RNG "sticks" on one number and repeats it. Taken over a long enough sample the total output looks OK, hence why Cyanide defend it. But sometimes it's too often statistically off the scale over batches of, say, 20 consecutive rolls.
I have, in fact, looked at the RNG across decent-sized samples. Then I went into the program and found the RNG itself and duplicated it, then ran it through several million iterations to look at number distribution, just in case the modifying numbers they put into the mersenne twister were mucking it up. Then I went into the program and found the routines that convert the 32-bit number into a dice value, duplicated that, and ran
every possible RNG output through it.
Also note - don't talk statistics in terms of 20 dice rolls. Statistics are about large samples. It'd be like saying that "statistically" one dice roll NEVER results in the expected value of 3.5... well no shit, that's because that value is based on an infinitely large sample. In small sets your data points will be all over the place, but the real test is whether or not they work across a long period or not. If the statistical patterns were true on a fractal level (which is what you're complaining they're not), you could
then talk about the data not being random.
The only thing broken is player perception - computer games have taken to ensuring that players don't fail too many times in a row (usually giving an automatic success after two failures) to better match people's expectations of success. Those systems are what aren't random - the dice generation in BB, however, is.