“Estimates are filled with more uncertainty than confidence.” -Dominica DeGrandis
Summarising, in a nutshell, the pointlessness of estimates. Much better to make the work as small as possible (to reduce inherent variability and uncertainty), and then use data and the law of large numbers to get realistic ideas about how long something is likely to take. Keeping in mind, of course, that it’s all useless if you’re doing something completely novel.
Dominica’s suggestion is to use Flow Time to provide the data needed to improve probabilistic planning. Makes sense, no?