In a schedule, calculating overall duration based on probabilities can be very counterintuitive.
For example lets say there are few activities in parallel, and lets assume they all have same start date and same most likely duration. And also let’s assume that their most likely duration to finish is 10 days, minimum expected duration is 5 days or maximum in 15 days. So they all have a distribution of probability that will look like a triangle, raising from 0 from 5, topping at 10 and decreasing to 15 again. we can all say at once that most likely duration to finish altogether is 10 days but that is wrong. Because if we calculate the probability, we find that finishing at 10 days has only 0.5×0.5×0.5×0.5 = 12.5% chance. It will take more than 10 days by 100-12.5 = 87.5 % chance. Why? Because even if only 1 activity delays, the whole set will be considered late, just like a whole dinner party is late, when only 1 person is late.
In real life, this method does not need to be used often however. This is because of the way we do our schedules. The logic links we place, the durations we assign, will automatically take care of the sequencing of activities and the assigned durations will emerge most likely as from experience. The method above however describes the theoretical way. In construction things are rarely this precise.