## High Short Possible Durations Long

The beta distribution was chosen for PERT instead of the normal distribution because it more closely resembles people's behavior when estimating. We are naturally optimistic, which skews the results to the left. Some call the beta distribution the "Murphy's Law distribution." Note that the tails of the distribution intersect the X-axis and do not trail off toward the sides (they are not asymptotic). This means that there is no chance at all of an actual duration being zero or infinity. If the actual duration is shorter than the most likely, it will not be much shorter (to the left), but if it is longer, then it could be a lot longer (to the right). The message is "when things go badly, they ga/ery badly." Actually, since we only use three estimates in PERT, you would get a triangular distribution if plotted. The PERT and triangular distributions are shown together in Figure 15-13.

Figure 15-13. Triangular and Beta Distributions Compared

Figure 15-13. Triangular and Beta Distributions Compared The formula for the weighted average of PERT is actually the mean of the triangular distribution used as an approximation of the beta, by experimentation. This is because in the 1950s, computing power to crunch the actual mean of the beta distribution was not available. The formula for computing the weighted average is:

PKRT Weighted Average = 