A bit off topic, but you seem to be doing this kind of thing a lot: is there any trick for calibrating high-/low-probability events? I can see how to figure out whether my 50% is 50% or 40%, but I’d need to make a lot of predictions to get a statistically useful number of 1% predictions wrong, even if my 1% is really 2% (a serious error!)
You can know that your numbers were wrong, if many of the 1-2% predictions become true. But there is no way to find out (by looking at the outcome) whether it was 1% or 2% without several hundred predictions.
A bit off topic, but you seem to be doing this kind of thing a lot: is there any trick for calibrating high-/low-probability events? I can see how to figure out whether my 50% is 50% or 40%, but I’d need to make a lot of predictions to get a statistically useful number of 1% predictions wrong, even if my 1% is really 2% (a serious error!)
Are there any tricks? Base-rates/frequencies (plus Laplace’s law) and breaking down conjunctions (#2 and 3 in http://www.gwern.net/Prediction%20markets#how-i-make-predictions ).
You can know that your numbers were wrong, if many of the 1-2% predictions become true. But there is no way to find out (by looking at the outcome) whether it was 1% or 2% without several hundred predictions.