Nate SilverVerified account

Wouldn't this be the same as saying - in almost all cases pollsters were wrong?

A narrative forms. Polls are herded according to the narrative. The narrative turns out to be inaccurate.

That's kinda psychology. We look for confirmation of what we expect, in our numbers and would tend to minor errors in our favor.

And as a biostatistician - in all the experiments with an outcome expected for theoretical reasons.

Can add Ireland's Same Sex Marriage ref to list. Initial polls 75-80% in favour. Tightened to 65%. Final much closer than expected - 62%.

But how do you assess conventional wisdom except by using a poll? Isn't a poll just an aggregation of conventional wisdom?

Yes but how do you quantify before the fact, as part of a prediction? Otherwise, you're operating on the same plane as astrologers

Evidence that we're living in a simulation.

Also holds for Trump/Russia Collusion freak-outs, fake dossiers from suspended British spies paid for by politicians . . .

The election of Justin Trudeau in Canada late 2015, same thing.

If it were in the "right" direction, it would already be captured in the analysis?

You think it's pollsters over-correcting? We were under on Trump so we should give LePen every benefit, etc....

Didn't those people expect LePen to do better BECAUSE of what they saw with Trump and Brexit?

Maybe better way to say this is (anecdotally) polls are lying between conventional wisdom and reality -- polls overfit some, we do more