I'm fascinated to know the author's opinion about Joel Spolsky's "Evidence-Based Scheduling" approach. The author mentions that breaking the schedule into the smallest possible pieces can lead to "overfitting," but Evidence-Based Scheduling uses previous programmer estimates vs. actual time to give a probabilistic schedule that changes with new events. Instead of a rigid ship date, Joel's method can give you a probability of shipping on a given date with the current information. Is Evidence-Based Scheduling still harmful, or is it basically the equivalent of the relative points-based estimation the author brought up at the end of the article?
EBS looks to me like an elaboration of the old-school PERT 3-point estimation technique. By the Laws of Agile, everything invented before 2001 is busted and useless, so I guess EBS would go out with the bathwater for being tainted by association.
