2 May 2018 - predict probabilities 1
< yesterday --
tomorrow >
If you make enough predictions, “an x% chance of y,” and keep track
of them, then you have the tools you need to make demonstrably
statistically sound predictions—you can promise that the percentages
you give are accurate. It’s a simple matter of applying formal
methods, or in other words, it is utterly unintuitive. Virtually
nobody does it outside of specialized areas like meteorology and
election forecasting, which are disliked and distrusted because they
honestly predict probabilities, instead of falsely claiming to predict
events. Now you understand why fields like sociology and political
science have lies in their names.