i am trying to work out the number and sides of dice to use in my RPG for different things, but i want to see the distributions of the rolls so i don''t have to trial-and-error it [as much]... right now i have a little program i wrote, which takes the number of dice and how many sides they have for input, and "rolls" them a bunch of times and draws a distribution graph of it for me. however, this is slow and innaccurate unless i use a huge number of iterations (i use 10000 usually, and even then the graph isn''t as smooth and symmetrical as it should be). is there a formula for this? so i could draw the graphs as they "should be"? thanks in advance. --- krez (krezisback@aol.com)

Maybe the binomial distribution can be used for what you're trying to graph.

Edited by - grady on February 20, 2002 5:51:02 PM

with my google search i found out that that won't work (here), as well as a bunch of pages telling me how to use various dice-rolling software for various statistics classes around the world...
thanks for the attempt though!

EDIT -- wow, i left a quote out of the link and the forum got real thin...

Edited by - krez on February 20, 2002 6:34:42 PM

You can use brute force:

  int Count(int iDice, int iSides, int *piCount){   int iDie[iDice];   int i, iSum;   char bDone = 0;   for (i = 0; i < iDice; i++)      iDie[i] = 1;   for (i = 0; i < iDice * iSides; i++)      piCount[i] = 0;   while (!bDone)   {      int iSum = 0;      for (int i = 0; i < iDice; i++)         iSum += iDie[i];      piCount[iSum]++;      for (i=0, iDie[0]++; (i < (iDice - 1) && (iDie[i] > iSides); i++)      {         iDie[i] = 1;         iDie[i+1]++;      }      if (i == (iDice - 1))         bDone = 1;   }   return(pow(iSides, iDice));}  

There might be a few errors, but it at least gives the general idea. The return value is the total number of possible outcomes. It is the same value you would get from summing the piCount array. The probability of a number, i, occuring is piCount/iTotal where iTotal is the value returned by the function. The only reason for actually using a generator is to verify that the numbers generated are actually distributed as they should be.

I assume there are less computationally intensive methods to find the probability of a given sum, but right now I''m at a loss as to what that might be. Since it is semetric you can cut it in half by stopping when the last digit passes the halfway mark. So with a 3d6 you could stop when the 3rd die rolls over to 4.

Well, it seems it would be a bit challenging for him to figure out exactly how he is going to use that to find the probabilities of a given total. It isn''t quite as straight forward as you make it sound since you have to find all the sequences and then calculate the probability of each sequence. With a 10D10 that is at least just around 10^5 instead of 10^10. It is still pretty much a brute force approach since you with a 10D10 there are only 91 totals, but around 10^5 sequences.

I don''t feel up to showing the actual code, but you would generate the sequences by generating all sequences where each number is less than or equal to all the numbers before it. Once you have a given sequence you have to count how many times a given value occurs, i.e. how many dice came up one. Then you plug those counts in for the x''s in the multinomial probability density function. Assuming all your dice have the same number of sides then all the p''s are equal to 1/sides. You have to sum all the dice to figure out what you are calculating the probability of and sum all the probabilities for all the sequences that sum to the same value.

If you do it correctly then when you are done if you sum all the probabilities for all the totals it should equal one. You most likely will not be exactly equal to one due to the number of computations. You should be very close to it though.

You lost me there. You don't need to know how many ways to get x1 A1's and so forth. The multinomial calculates the probability of getting any of them. Getting the xn's is from my view the challenge. Now perhaps combinatoric provides a method for finding the xn's, but you don't say that.

Here is a link that explains how to calculate the probability of a given total. The P(p,n,s) function is a function taking the total of interest, p, the number of dice, n, and the number of sides, s.

Edited by - LilBudyWizer on February 22, 2002 10:14:02 AM

ha ha ha.
thanks yous guys, i really do appreciate it... but apparently i''m not as good at mathematics as i once was (or thought i was?)...
and this is not to say that i am uninterested in learning... but really, to even come close to getting this i''d have to reread a few (several?) of those books i still have from school, and all i really want is a few nice graphs to compare when i''m deciding if some stat should be 3d6 or 6d3.
perhaps some day i will want to become learned in these things, but i do know enough probability to say it isn''t likely
brute force it is i guess... dammit...

--- krez (krezisback@aol.com)