This was part of my point.. without programming in the exact seed and algorithm, you don't know the chance of a candy appearing. You could expand the algorithm from my original post to view the change in line and "deduce" from historical information what might appear, but it's still just a guess and may be dead wrong.