Wednesday, January 02, 2013

Skiltao's Dice: Return of the Comebackening

Some weeks ago, I went over some numbers that Skiltao mentioned on his blog (Part 1 & Part 2). I didn't do any rigorous mathematical analysis on themwhat I did was more akin to recognizing the geometric qualities of dice probabilities—but I did manage to illuminate the qualities of the dice rolls and why certain relationships existed between them. Whenever I was done, Skiltao asked about alternative dice relationships, for example, if different-sized dice would have similar relationships or how dice which have repeated numbers would be affected.

The first question is straightforward. The odds of a tie between, say 1D10 and 3D4 being expressed as a certain value on a 1D10 + 3D4 roll is easy to test.

1D10 ranges from 1 to 10, while a 3D4 ranges from 3 to 12. The odds of a tie are the compounded probabilities of a set of values on the table below:

1D10      1D4 #1        1D4 #2        1D4 #3
[ 3]      (1)           (1)           (1)
[ 4]      (2)           (1)           (1)
[ 5]     (2,3)         (2,1)          (1)
[ 6]    (2,3,4)        (2,1)         (1,2)
[ 7]     (3,4)        (3,2,1)        (1,2)
[ 8]     (3,4)         (3,2)         (1,2)
[ 9]     (3,4)         (3,4)        (3,2,1)
[10]      (4)          (3,4)         (2,3)

I'm sorry that's not as clear as the 2D6 table from part one, but you get into matrices after a while.

None the less, the 1D10 is as fully reversible as 1D6. You can flip the range of probabilities, [3-10], to get an equal set of probabilities, [8-1]. The summation for [8-1] and (3-10), yields 11 by matching [3, 8), [4, 7),...[8, 3).

So yes, as long as the lower set can be flipped, the relationships between the probability of a tie between two sets of dice will be equal to the odds of a particular summation of their results.

Now, whenever we compare the odds of 1D10 beating 3D4, we have a different relationship.
Again we're doing pairs of odds:
1D10  3D4  1D10  3D4  1D10  3D4  1D10  3D4  ... 1D10  3D4
(4)   [3]  (5)   [3]  (6)   [3]  (7)   [3]  ... (10)  [3]
           (5)   [4]  (6)   [4]  (7)   [4]  ... (10)  [4]
                      (6)   [5]  (7)   [5]  ... (10)  [5]
                                 (7)   [6]  ... (10)  [6]
                                            ... (10)  [7]
                                            ... (10)  [8]
                                            ... (10)  [9]

Again-again, we can turn that into a set of probabilities of getting a results equal to less than, say 10, on the summation between 1D10 and 3D4.
 1D10        3D4
 [1]   (3,4,5,6,7,8,9)
 [2]    (3,4,5,6,7,8)
 [3]     (3,4,5,6,7)
 [4]      (3,4,5,6)
 [5]       (3,4,5)
 [6]        (3,4)
 [7]         (3)

Because, for a 1D10 [1-7] = [3-10], we can see that the odds of getting a ten or less on 1D10 + 3D4 = the odds of 1D10 beating 3D4.

Now, for irregularly numbered dice, 2,2,3,4,5,5--known as XD<NNNNNN>, where 1D6 = 1D<123456>--it's a different matter.

The odds of 1D<112344> tying 2D<112344> is equal to:

1D<1...4>  2D<1...4>
 [2]        (2)
 [3]        (3)
 [4]        (4)

Now, the odds are swankier, but still symmetrical. [1-3] is still equal to [2-4]. So the odds of 1D<112344> equaling 2D<112344> is equal to getting a sum of 5 on 3D<112344>. Special note: I don't have to crunch the actual odds. This is all geometry of probabilities, which is something I'm making up as I go and is probably wrong. If you were hoping for actual numbers and formulas and equations you are shit out of luck.

Right, so part two subsection two: are the odds of 1D<112344> beating 2D<112344> equal to the odds of rolling less than or equal to N on 3D<112344>? Probably.

1DN*  2DN* 1DN*  2DN* 
[3]   (2)  [4]   (3)
           [4]   (3) 

Again, we can turn that into a set of probabilities of getting results equal to less than, say 4, on the summation between the two.

 1DN*  2DN*
 [1]  (2,3)
 [2]   (2)

Brace yourselves: it's the same fucking thing, so yeah.

But I'm not sure if those answers are in keeping with the spirit in which the original questions were asked. I took a moment and reconciled the 1D<112344> and 1D6 intersection, and because the geometry of the probabilities are both symmetrical, I'm pretty sure the relationships would hold true.

---
*Seriously, fuck that.

No comments: