For the Record: Election Handicapping

 

Photo via.

Following up on the previous, some thoughts I had with regard to a longstanding with Thornton/Henry, who's right on this problem with election forecasting (not polling, forecasting, claiming to know with a certain probability who's going to win):

It's interesting to me how we sneak up on this as we go from coin tosses to poker to baseball to elections. It's really the jump from poker to baseball where the fundamental error happens.

— Henry Porter (@HenryPorters) November 22, 2022

Right, exactly. The error isn’t in what is happening but in 538’s ability to discern what’s happening.

— The Fig Economy (@figgityfigs) November 22, 2022

Not error in the sense of wrongness, error in the sense of margin of error. When a poll has a margin of error it isn’t because hey anything could happen, it’s inherent in the measurement. That’s all I mean.

— The Fig Economy (@figgityfigs) November 22, 2022

A poll--a "snapshot" of the present, as opposed to a forecast of the future--does mean something, for sure, and I'm convinced the margin of error is a valid concept. It's the forecasts that are really uninformative.

— Fully Weaponized Monster (@Yastreblyansky) November 22, 2022

say the one in the Daily Racing Form. Bettors put their money down according to that basic information and their own "hunches", some informed and some random, and the odds at post time are the summary of that crowd wisdom.

— Fully Weaponized Monster (@Yastreblyansky) November 23, 2022

What strikes me for the first time is that the value of odds on a horse (or a more complex bet like exacta or triple) is a form of information value--it summarizes what we know about the horse in the context of this particular (one-time-only) field.

— Fully Weaponized Monster (@Yastreblyansky) November 23, 2022

I think that could be a basis for understanding what the forecast "means", and evaluating it, though I don't know how one would go about doing that. (It would value Sam Wang's 2016 forecast over 538's because the former accurately predicted how Trump victory would shock us).

— Fully Weaponized Monster (@Yastreblyansky) November 23, 2022

(That is to say, because Wang's prediction kicked against the consensus--Rachel Bitecofer's "true" prediction of House races in 2018 was valuable in the same way, because that result shouldn't have been surprising at all.)

— Fully Weaponized Monster (@Yastreblyansky) November 23, 2022