The proposed definition of PERT CRITICAL PATH does not do justice to the research in 1962 which shows that PERT ignores the "merge bias", or that PERT has the capability to use statistics (variance, standard deviation, probability distributions) to calculate the likelihood of the scheduled finish date or identify the finishing date that provides the desired confidence of not further overrunning, e.g., the P-80 date will overrun only 20% of projects like this. In other words, without the following changes both limits and strengths of PERT are poorly represented. It will appear, with some justification, that AACE does not understand PERT at all.
So, we need to include "so a PERT Analysis does not evaluate the risk to the schedule that is caused when slack paths that merge with the PERT Critical Path drive the finish date if some risks occur. This effect is called the "merge bias," and ignoring this effect underestimates schedule risk relative to other schedule risk analysis techniques (e.g., Monte Carlo). Once the PERT critical path is identified and its expected value duration computed, the analyst calculated the standard deviation of the PERT crtitical path to discover results including: (1) the probability of finishing on the scheduled date, and (2) a finish date that provides an acceptable probability of being overrun."
PERT (Performance Evaluation and Review Technique) is a similar methodology to that of CPM (Critical Path Method) analysis with the most significant difference being that rather than based upon activities of estimated duration starting and ending at points in time, PERT is based upon milestone (or points in time) separated by logic restraints of estimated duration. The estimate of duration of an activity of usually well known scope is typically more certain than the estimate of a logic restraint and less well developed scope between two milestones. For that reason the estimate for CPM is usually a set duration while for PERT the estimate is a range, although this aspect is subject to variation in either format.