Login Register



[Reply]
Forum Index > AIM/ICQ Discussion
Dice Roller RNG
Posted on: 03-06 3:59 am
metaphysician

Just as a heads up, my gaming group has been using the restored client and its dice roller for our gaming. I think there might be something wrong with the RNG. I don't have a full statistical analysis or anything, but it seems to skew direly low, with far more 1s than probability would indicate.

For reference, we do a lot of 10 sided dice ( White Wolf games ), so it could be connected to the system handling larger dice size numbers incorrectly.

Hope this is at least helpful information.
Posted on: 03-08 2:10 am
Wildman

Beta Tester

hmm the internal class rnd() is used which is a windows API random number generator.. maybe more randomness is needed

Hououin - your friendly neighborhood AIM server operator
Posted on: 03-19 2:00 am
metaphysician

Beta Tester

Okay, on further play/examination, it looks like the dice roller function is using the time to seed the RNG, because rolls done at roughly the same time are producing the same dice outcomes for two or more players. However, they clearly aren't providing *much* randomness, because the same strings of numbers are turning up repeatedly within the same session ( we're talking "identical 10 digit string recurring five minutes later" ). I don't know how its producing the numbers, but it must not be using nearly enough randomness.
Posted on: 03-19 2:19 am
cglasgow99

Beta Tester

The problem may be that you're using rand instead of rand_s, as mentioned in this article.

https://en.wikipedia.org/wiki/CryptGenRandom

Basically, original Windows API random # generator is kinda weaksauce. They made a stronger one.
< - 1 - >

[Reply]