.

Just made a thread on this but it's probably to page 3 now

There was a lib who made a not so subtle snarky thread about how many people you know who have landlines to discredit the poll prior to the election

Nate has been weighed, he has been measured, and he has been found wanting...

Nate Dumbass has been destroyed. He has no credibility anymore.

Can't wait for the heckling and mocking he'll get when he publishes anything.

Dude should look for another line of work like weather forecasting.

quote:

---

Nate Silver isnt the wiz kid you made him out to be.

---

If I recall correctly, he tweeted or announced some other way that Trump had won the election around 8:30 CST.

I think he went dark for a while after that.

I haven't slept much since Tuesday night, so it may all be running together.

ETA: Let me be clear, I'm not defending this clown. He is certainly a DNC surrogate.
I'm saying he was the first clown to admit Trump won and he got it wrong. It was almost like Josh Chamberlains last post in 2012.

Trafalgar also missed Trump by 9 points in Utah - +10 compared to +19 actual. They had Trump +5 in Nevada and he'll lose by 2.4 there, so off by 7.4. Missed Trump by 3 points in Colorado.

quote:

---

Nate Silver isnt the wiz kid you made him out to be. no need to defend and apologize. He's totally discredited.

---

That may be right. Except 2012 and 2016 were two highly unusual elections from a predictability standpoint.

Sandy + the devastating Project ORCA GOTV collapse depressed Romney's national result 2-5 pts. Perhaps Sandy could have been better modeled. Orca collapse was a one-off, and could never have been predicted.

This time, silent Trump support was huge. Skews were predictable based on participants not wanting to be pegged as part of the deplorable, irredeemable, racist, misogynist hoards of uneducated, unwashed Trumpkins. As I posted many times here, examination of NC predictions vs past performance could have been used to re-model results. But such models would likely vary state-to-state and would involve significant artificial manipulation/skews.

IBD actually chose that approach, and turned out (as in 2012) to be quite accurate. So remodeling was doable, but required reworking standard techniques. Given different candidates, standard models would have much more closely tracked results.