Specifically, this is about predicting which states end up voting for which presidential candidate.
(Default 538 model.)
@vitamind clearly you haven't seen the strange 538 model this year
@nikki And what would that be, for those of us who don't know (including myself somewhat)?
Additionally, I wouldn't forecast Manifold to have a 2/3 chance of beating any other model either as a heuristic. I think the market is good in the ~40-55% range
@StarkLN https://stat.columbia.edu/~gelman/research/published/Harvard_Data_Science_Review.pdf
Bayesianism is not a good way to model tail events
@nikki I think I asked the wrong question. It would be better to ask: in what ways do Elliot's thinking majorly differ from Nate Silver's in a way that you think hurts the model. I didn't keep up with the Nate Silver Elliot Morris twitter stuff + it looks like most of it has been deleted by now.
@LarsDoucet I would advocate for "whatever the default model is displayed to a new visitor to the site", but would be nice to get that confirmed @IsaacKing.
@IsaacKing What probability would you say 538 had given for a state it predicted to be ">99%" likely to have a particular outcome?
Interesting to consider how manipulable this market is. With 538-betters being profitable this last cycle, I might expect prices to be very close to 538 the day before the election. A coalition could attempt to match the manifold predictions to the 538 predictions closely and then make a single big bet on a very likely state which would then dominate the Brier score.