the problem is only asking about the probabilities from the final state on, yeah? at that point your selections are practically reduced to a 50/50. i'm sure it'd be much different if it were asking about the probabilities of each state change in the run. i'm pretty sure you're all over complicating this.