as posted on the atari boards: I have to say, bugs and all, the most consistently frustrating thing about this game are the die rolls. I knew it wasnt my imagination, so I tracked 300 consecutive die rolls by my party (I got too bored noting the exact number,so I only noted either BELOW 10 or ABOVE 10). The results? 211 rolls BELOW 10. 75 ABOVE 10 14 EXACTLY 10. Also, noted 20's and 1's. Result? 22 1's, 9 20's. Now, this is statistically bogus. Someone rolling that bad would say DICE FIXED in a minute. I am wondering if they did this to better balance the game? I am not a prop/stat guy, but my buddy told me the odds of this happening on this # of rolls are small,something like 1 in 8 or 1 in 9. So tomorrow I startagain. He says 200 consecutive 300 roll sessions with this disbursement is like 1 in 50. (note - i dont cound the creature rolls. only my own, combat and non-combat. (most were specllcraft rolls).
you gotta count ALL rolls not just your own, the random number generator doesnt distinguish between pc and npc rolls, or else this isnt an accurate test
Actually, it is still an accurate test, as it represents a sample for the population we're studying. He isn't worried about monster rolls - he's worried about PC rolls. You can't find out about randomness in PC rolls by looking at monster rolls. Regardless, it's generally accepted for experimental evidence that we should try to get a probability of the result being random chance at less than 5%. I'm not sure what test he used, and my stat-fu is weak since getting out of grad school. I'd like to see if others have done the same thing. -O
There's a learned discussion regarding this with some developer input as well here: http://www.ataricommunity.com/forums/showthread.php?threadid=324381 Uniformity of rolls seem Ok but uncorrelatedness is doubtful with the RNG they use, from what I gather.
According to a troika programmer: "Our random number generator is a slightly modified version of the Quick and Dirty random number generator described in Numerical Recipes in C. It is a easy to use, reasonable RNG that uses VERY little cpu time. As far as I know it is the same RNG used in Arcanum and maybe other titles." The Quick and Dirty alogrithm is quick, but it is dirty, which explains the correlation problems some people are having.
