Just picked up Chances Are: Adventures in Probability, Michael Kaplan and Ellen Kaplan. It's a book about probability and its impact on everyday life. Chapters on gambling, health, judging, fighting, and predicting.
One of the predicting aspects I'm most interested in is predicting cost and schedule completion. It continues to amaze me how often I encounter a project - a troubled project - that does not have a probabilistic project plan. Without a probabilistic estimate to completion and the associated costs, the project is doomed from day one. Without this information, the project managers and their "customers" are literally driving in the dark. They have no clue about how reliable the plan is in estimating completion, where the programmatic risks are, or even where opportunities are for improving the schedule. They know only that the path to the end is described by the connected tasks and their durations. But they know nothing about the dynamic behaviors of these tasks from a probabilistic model point of view.
Why is this the Case?
I think because it is actually hard work to put together a network of tasks, do the analysis, build the probabilistic model and develop an understanding of how the probabilities and their related statistics interact with this network. It is much easier to simply start working on the delivery of products or services with the VERY naive assumption that things will work out in the end, or our method can deal with the pending changes, or simply that "what we don't know can't hurt us."
In the end though it is dumb luck if the project shows up on time, on budget and on specification.