Most states print a warning on lottery tickets: “Should not be used for investment purposes.” With weighted-average and PERT-based estimates artificially bounded by what we choose to be relevant, our projects should probably carry the same warning.
Look at the picture below. What patterns do you see in it? This is an unfair test for any human because humans are pattern-seeking and pattern-finding organisms. Give us a cloud, and we'll see something in it. We even see a man in the moon! In fact, this picture is composed of random dots, quite deliberately pattern-free! Yet I see beautiful woman dancing in a flowing gown there! Don't you?
Nassim Nicholas Taleb has been writing about a phenomenon he refers to as being Fooled By Randomness. He claims that, unlike bounded samples such as the distribution of the heights of students, many events in this world are better described by what are called power curves. These distributions have "fat tails" rather than the fat middles we expect a "normal distribution" to exhibit. So, rather than cluster around a middle, likely events can and do appear at the extremes.
What’s the problem with that? Well, for one, when we calculate the probability of an event occurring, Taleb says that we invariably assume a fat-middle distribution, so we discount the likelihood of what in a fat-