3 ms·
Presenting the election result as 98% certainty is either fake or insanity (for context, that would mean (on average) only 1 wrong prediction in 200 years of US
by tomp 10y ago
Presenting the election result as 98% certainty is either fake or insanity (for context, that would mean (on average) only 1 wrong prediction in 200 years of US presidential elections).
Edit: admittedly, that was Huffington Post, hardly a standard for good journalism. E.g. NYT had 80%, 1 wrong in 20 years, which is still quite optimistic IMO.
- dllthomas 10y agoIsn't 98% certainty "1 wrong in 50 years", and 80% "1 wrong in 5 years"?
- tomp 10y agoNo, it's "1 wrong out of 50 events", but elections are only every 4 years.
- dllthomas 10y agoAh, that's what I was missing. That said, regardless of how often these events are occurring, we shouldn't be all that surprised by a 1/5 event.
- dllthomas 10y agoThinking further about this, I think framing it in years is highly misleading. If we imagine an event that occurs once every ten-thousand years, with even probability of two outcomes, then if I pick one of the two as my guess and I correctly assign a 50% probability, you can say "you're saying your guess will only be wrong, on average, once every twenty-thousand years!" That's technically true, but I'm clearly not making too bold a claim!
- tomp 10y agoYeah I sometimes question the sensibility of even trying to assign probability to what are essentially one-time events. I prefer to think of it, the probability is either 0% or 100%, we just don't know it yet!
- dllthomas 10y agoIn a deterministic universe, there's a sense in which that's always the case. But really, probability is always "given the information available". "I know everything about a deterministic universe" is just a special case. Related to the broader topic: https://en.wikipedia.org/wiki/Scoring_rule#Proper_scoring_rules https://en.wikipedia.org/wiki/Scoring_rule#Proper_scoring_ru...