Quote:
Originally Posted by NutzyClutz
What are odds of going on 154 roll run?
It is easy to solve this problem exactly, and you just need a program that can multiply 4x4 matrices. Excel can do this. I found the exact probability of rolling at least 154 times before missing a point to be about 1 in 5.6 billion. The wizardofodds website is reporting the number 1 in 5.3 billion (not 3.5 billion as Mizzles has posted), but this is based on a monte carlo simulation rather than an exact calculation, and he admits that he did not run enough samples for accuracy at 154 rolls, and he is extrapolating his error. This isn't the first time that the wizard has relied on simulation for something that can be computed exactly with relative ease, and then reported more digits than could be justified by the size of the simulation. This problem cries out for a state space solution as shown below.
I heard the number 1 in 1.56 trillion in the media, but this is simply (30/36)^154 which is the probability of no sevens in 154 rolls. This is not sufficient since it is OK to have sevens on the come out roll. It is also OK to LOSE on the come out roll (with 2,3, or 12) according to the article at wizardofodds.com, as the dice don't get passed in this case. If you wish to know the probability of at least 154 rolls before losing a point or losing the come out bet, I can compute this by only changing a single number in my matrix, and this gives about 1 in 260 billion.
For the original problem, the solution is as follows:
Code:
[ 12 3 4 5 ]
A = 1/36 * | 6 27 0 0 |
| 8 0 26 0 |
[ 10 0 0 25 ]
[1] [p0]
A^153 * |0| = |p1|
|0| |p2|
[0] [p3]
P(154 or more rolls before losing point) = p0+p1+p2+p3
Explanation
Define 4 states:
State 0 is the come out roll.
State 1 is rolling for a point of 4 or 10, both of which are 6:3 underdogs
State 2 is rolling for a point of 5 or 9, both of which are 6:4 underdogs
State 3 is rolling for a point of 6 or 8, both of which are 6:5 underdogs
The 4x4 matrix A above completely characterizes the game of craps, as far as the pass line bets are concerned. The numbers in the matrix are state transition probabilities, with the 1/36 factored out for convenience. The number in (ROW, COLUMN) is the probability that we transition TO state ROW, FROM state COLUMN. So for example, numbers in the first column give the probabilities when we are on the come out roll, on which we have a 12/36 probability of rolling a 7/11 or 2/3/12, either of which would bring us to another come out roll. There is a 6/36 probability that we roll a 4 or a 10 which would put us in state 1. There is an 8/36 probability that we roll a 5 or a 9 which would put us in state 2. And there is a 10/36 probability that we roll a 6 or an 8 which would put is in state 3. The columns for the other states each give 2 non-zero probabilities; one for making the point and returning to the come out state 0, and the probability for any number besides the point or 7, which leave us in the same state. Note that we do not require a state for rolling a 7 which would terminate the player's rolls.
Now the column vector [1 0 0 0] represents the state before the first come out roll, where we are in state 0 with probability 1. When the transition matrix multiplies this vector, the elements of the resultant vector indicate the probabilities of being in each state after the roll. For each roll this matrix is multiplied by the resultant vector from the last roll, until after multiplying it 153 times, the resultant vector will give the probabilities of being in each of the 4 states after 153 rolls, each of which satisfy the conditions of the problem. Note that these will not sum to 1 since the cases where we seven out are not represented. We simply need to sum these 4 elements to give our desired probability.
EDIT: Points 8 and 10 were swapped in the state definition and explanation.
EDIT: Factor of (1/36) is part of matrix which gets raised to power of 153.
Last edited by BruceZ; 06-01-2009 at 01:28 PM.