Friday, October 14, 2016

How full of shit is Nate Silver?

As long as I'm doing this, I might as well do this.  In my morning post, I was not kind to Nate Silver.  Those who have spoken with me about the topic know that this is not a new thing.  I have just never been impressed with his work.  Yes, he correctly called just about every race last time.  So did everyone who looked at simple polling averages.  Fuck Dylan, but you don't need a weatherman to know which way the wind blows.  (And yes, I know what that line was really about, but I'm still not impressed with Dylan either).

Anyway, I said Nate Silver was full of shit, but let's ask an updated question.  How full of shit is Nate Silver?  As of typing, the 538 "polls only" forecast gives Trump a 15.3% chance.  PredictWise gives Trump a 9% chance.  Let's say PredictWise is right, and 538 is wrong because, on average, the prediction markets beat the polls.  So what?  Both put the odds heavily against Trump.  It is only a 6.3 percentage point difference.

Well, 6.3/9=.7.  If PredictWise is right and 538's "polls only" forecast is wrong, then 538 overestimates Trump's chance of victory by 70%.  If you are betting, sorry, "investing," that costs you big.  You will buy and sell at the wrong prices if you follow Silver's advice.

Do not let yourself get crucified on a cross of Silver.

And to repeat:  if a major terrorist attack occurs, or something like that to give Trump the presidency, that is not a vindication of Silver's calculation method, nor is it a vindication if the polls are all just wrong across the board.  His method is based on assumptions about independence and interdependence between state contests that cannot be justified.  Unforeseen and unforeseeable events would simply be a demonstration of the intrinsically unpredictable nature of this year.

Settle down, monkeys.

No comments:

Post a Comment