Nate Silver at 538 said Bernie's chance of winning MI was <1%

[Kelvin Mace]

http://projects.fivethirtyeight.com/election-2016/primary-forecast/michigan-democratic/

Discuss!

[Champion Jack]

1. Yeah, Nate screwed up

[Ford_Prefect]

2. I think Nate is plugged into the wrong set of clues and can't see the ones that work now.

[Cleita]

7. Exactly. What was true in past elections isn't the same today.

This is what happens when you don't leave the bubble to look at what has changed outside of it.

[Ford_Prefect]

9. Imagine that......

[PonyUp]

24. Nate has his Hillary blinders on.

[RobertEarl]

3. 538 is toast

[Cheese Sandwich]

4. 538 is snake oil

It's a total fraud. Any one of us could do what they do just by looking at the internet and making a website about our predictions.

[Cleita]

5. I guess Nate Silver needs some polish to remove the tarnish on

his reputation.

[marions ghost]

6. Those crazy voters

[earthside]

8. I'm sick of Nate Silver.

This little statistical prediction game is boring.

[SMC22307]

12. "MATH!" indeed.

I'll second your "sick of."

[Dem2]

10. Doesn't say much about Nate

Says a lot about how terrible the majority of polling firms are in certain types of open primaries.

[cemaphonic]

14. Agreed

538's statistical models may well have been flawed, and I'm sure they and others will be studying this primary for some time. But ultimately, it was the polls themselves that were consistently off from the final result.

[DirkGently]

11. Silver tweeted something about having

an "intuition" that Sanders might win. From the man who scoffs at gut feelings in the face of inexorable numeric truth.

He's fine, really. Very smart. But the aura of invincibility is gone from 538.

Their model went splat tonight.

[BernieforPres2016]

17. Silver is full of shit

In my experience, people who don't have much or any background in things like statistics, economic modeling or financial modeling tend to be way too impressed with people who grind up a bunch of numbers and spit out something that implies a level of precision that can't possibly exist. With all the unknowns and moving parts, like what percentage of which types of voters will show up, are the polls properly accounting for independents and people without land lines and so many other issues, nobody can aggregate a bunch of polls and data and say there is a 69% chance this will happen and an 82% chance that will happen.

And then he always has the out clauses when he misses by a mile, well we missed this and that surprised us, but now we'll incorporate that into our analysis and next time we'll be right!

And then next time it will be some different issues that he didn't quite account for.

[Dawgs]

21. Been saying that for years.

He took the average of the state polls in 2008, took the results as his, and became famous for it. Way too many Democrats fell for it. It's quite embarrassing.

[jfern]

13. I want more decimal points. Was it less than 0.1%? 0.01%?

[TheDormouse]

15. Less than 1% is NOT 0%

so give them a break

... unless they are consistently predicting Sanders will lose and being proved wrong again and again.

[pa28]

16. One possible avenue of defense for Nate.

The polling data on which he based this conclusion on was mostly taken before Hillary's smear campaign against Bernie's record. Maybe a spectacular rejection of her dishonesty played a factor.

I don't think it was the only factor but I think it backfired on her badly.

[BernieforPres2016]

18. I don't buy that

The spin yesterday and today was about Bernie's "ghetto" comment at the debate on Sunday night. The press was trying to turn that into a big thing and largely ignoring Hillary's smear on Bernie about the auto bailout. Then Bernie wins and they say wait, maybe Hillary's smear really hurt her.

If Silver is such a great analyst and the Hillary smear was a problem, why didn't he pick that up and factor it into his models?

[Dawgs]

22. There is no defense. His only job is to be accurate. If he fails at that it's his fault.

It doesn't matter what other factors exist.

[Flying Squirrel]

19. Garbage in, garbage out.

The polls can not be relied upon to even come close anymore.

[madokie]

20. Silver was once right

that is all

or right once, however you want to put it

[VulgarPoet]

23. Even people in the OTHER group are throwing him under the bus.

[Cobalt Violet]

25. K & R!

[Renew Deal]

26. That's better than the chance to win lotto

[PoliticAverse]

27. ...

https://en.wikipedia.org/wiki/Misuse_of_statistics

[Kip Humphrey]

28. Just like DU, many Hilary supporters do not accept the revolution and don't see it coming...yet.

Nate, I think, is one.

[LWolf]

29. Well, that's just wrong.

Polls don't lie, and neither does Nate Silver. He's the best, and he's way smarter than every voter out there, so voters are just supposed to nod their heads when he speaks and take his pontifications as gospel.

Since Nate Silver predicted a big win for Hillary, the headlines and numbers showing a Sanders win are wrong. Wrong, wrong, wrong.

You'll see. Just wait.

?

[ladjf]

31. Those that don't "see" the Revolution have some serious political "cataracts".

They need to consult with Dr. Sanders. Political "cataracts" can be successfully treated.

[LWolf]

32. Yep. nt

[ladjf]

30. To borrow a line from "The Last of the Mohicans",

"We were wrong to have held him in such esteem."

[Le Taz Hot]

33. They keep polling Democratic voters.

They're missing the Indies and the Republicans and yes, we ARE winning over Republicans.

[Bluenorthwest]

34. Nate's reputation is less than zero.

It's the math. It's the math. He keeps saying that even when he uses words to draw his conclusions 'my opinion counts as an equation, I AM the math!!!'

[LonePirate]

35. Lots of people seem to have no or little understanding of his modeling/predictions.

Yes, he was wrong; but given how he compiles his information, there was almost no way for him to be right. He relies on polling data which was universally wrong through no fault of Nate and his team. It's the old "garbage (data) in, garbage (data) out" effect which blew up his model last night.

[Kelvin Mace]

36. Exactly.

Models are only as accurate as the data used. When your data is faulty, conclusions are faulty. This is not your mother's/grandmother's election.