Are the dice rolls rigged?

MrDamage · February 2018

I’ve always wondered this about the game. I’ve sat there as a low level pc and watched double 1’s, really low and crappy rolls. Playing on a PW, I failed a save vs slow trap with a 3 like about 5 times in a row at different times. Then other times at higher level I get more crits and stuff. Or so it seems. Does anyone know if the engine actually uses a random generator for rolls or perhaps a group of number arrays set to pc CR or something?
If I charge into battle with KD, I almost always seem to succeed if I get initiative.

FreshLemonBun · February 2018

Random number generators are usually not random and they can suffer from some quality issues. There are methods for testing the fairness of dice rolls that are sometimes used to determine if a conventional die is no longer fair ie weighted or warped with time. It would be interesting to see something like that used in NWN and also some overall roll tracking, like tracking streaks and histogram of each value since after a while you expect it to even out.

While not NWN I've seen a lot of statistically impossible streaks happen several times a session while playing NWN2, but that was with circumstances generating many multiple saves.

NeverwinterWights · February 2018

This topic came up quite a bit on the old forums. Some people who wanted supreme random superness said there were issues with the seed but most agreed that Random() in NWN is random enough. Sometimes it can seem like you are just getting crappy rolls but if you run a couple script tests in game, like one that rolls a d4 4,000 times you will see that for the most part you get each number about 25% of the time:


void main()
{
    object oPC = GetLastUsedBy();
    int iOnes;
    int iTwos;
    int iThrees;
    int iFours;
    int iCount;
    for (iCount = 0; iCount < 4000; iCount++)
    {
        int iRandom = Random(4)+1;
        if (iRandom == 1) iOnes++;
        if (iRandom == 2) iTwos++;
        if (iRandom == 3) iThrees++;
        if (iRandom == 4) iFours++;
    }
    SendMessageToPC(oPC, "Ones: "+IntToString(iOnes)+"\nTwos: "+IntToString(iTwos)
                    +"\nThrees: "+IntToString(iThrees)+"\nFours: "+IntToString(iFours));
}

On occasion just because of chance you will roll each number 4 times in a row..or even possibly more. But over time and the more iterations it will all even out.

MrDamage · February 2018

Hmmm yeah. What sparked my suspicions initially was literally years ago when my low lvl character was bashing something as I wanted to watch the rolls. I noticed that every say 2 or so minutes, I would get a double 1 roll. Very odd. I suddenly realised that here and now with all the devs lurking about and stuff is probably the best chance I’d ever get at clearing this up lol.
Or as my old maths teacher once said.
Today we’re going to study the science of probability, although some would argue there’s no such thing.

NeverwinterWights · February 2018

I wish I remembered more about how they determined what the seed was. It's been so long and I didn't understand half of what they were talking about. Maybe something about the milliseconds and the area? But yeah might be cool if someone like maybe @niv could re-shed some light on that.

Sherincall · February 2018

Random() is random in NWN. It just uses the underlying libc rand() function.

There is a tiny bias towards lower numbers, but you'll notice it once in 4000 rolls at worst, and likely once in 200 million rolls. Not something you should be concerned about.

There were indeed problems with random seed. The client re-seeds the RNG based on static objects and tiles in the area. This is done so that the randomly placed grass is placed in the same spot for all clients. In versions prior to 1.65, the game would then forget to reset this seed, resulting in repeatable random sequences. That has since been fixed.

MrDamage · February 2018

@Sherincall
Wow! Thankyou very much for the explanation. That is really interesting to find out. I’m really tempted to copy and paste that into the ‘Did you know’ thread!

TarotRedhand · February 2018

One of the measures of "quality" of a pseudo random number generator is its periodicity. That is how many "random" numbers can be generated before the same sequence of numbers are repeated. If you are interested and you can find it, there is ANSI C code on the net for a prng that has a periodicity of 2 to the power 144 (i.e. an astronomically big number). I know this because I converted it to modula 2 a long time ago now. Suffice to say this is overkill for gaming purposes.

TR

mlkent22 · February 2018

Yes they are rigged.

Are the dice rolls rigged?

Comments