So reference-class forecasting is better on biases. It’s better on unknown unknowns. It’s simple and easily done. And it has a proven track record of delivering more accurate forecasts. I’m happy that it has been taken up to the extent that it has by different organizations around the world—much more than I thought would ever happen when I first developed the method for Gordon Brown—but I wouldn’t blame anyone for wondering why, given all its strengths, it isn’t used even more than it is, across the board.
