2016 Postmortem
Related: About this forumNate Silver Obama 92.2% Chance of winning
315.3 EV
Onward and upward!
http://fivethirtyeight.blogs.nytimes.com/
MSMITH33156
(879 posts)a bit too.
Doctor Jack
(3,072 posts)Just when I thought everything was as good as it would get!
BlueDemKev
(3,003 posts)Almost seems too good to be true. The biggest question I have with Nate Silver's numbers are how can he be so confident that a state (any state) will vote a certain way when the poll averages show a 1 or 2-point difference?
Doctor Jack
(3,072 posts)Just like New York and Illinois
CthulhusEvilCousin
(209 posts)If you have all the polls saying 1 to 2 digit lead, then that 1 to 2 digit becomes a higher probability of being real. Though there are probably many factors that get plugged into the simulation.
BlueDemKev
(3,003 posts)I mean an 80+% probability if a candidate is ahead by, say, 50% to 48%?
CthulhusEvilCousin
(209 posts)But the polls in OH, for example, aren't all just 1 or 2 points, but some are as high as 5 points. Nate's model does take into account other factors, like the lean of certain polls (whether R or D) and the margin of error, but when you have so many polls saying the same thing, I think the probability makes sense. You should check out Nate's blog, as I've read on it a few times how he works out his probabilities.
MSMITH33156
(879 posts)but essentially, he uses consistency of polling. It's not just average.
For example, if one state has 4 polls all with Obama+2, then he is comfortable that Obama has a 2-point lead and the percentage of win will be much higher.
However, if he has 4 polls, with 2 of them +6 for Obama and 2 of them +2 for Romney, the average will be +2 again, but in this case, his percentage his much lower.
And this is a scientifically sound statistical method. Essentially, 1 poll can have a large margin for error, but when you examine the polls in totality, the margin of error all but disappears. I believe he posted something the other day that shows that something like the candidate won 30 of 33 races where he was up by 0.1% to 5%. This is pretty robust. The only way Romney wins is if the polls as a whole, are collectively, severely biased across multiple states.
On Sam Wang's web site, you can even do the Romney +2 number, and Obama still wins in that case. The polls need to be more than 2 points off, in multiple states.
aletier_v
(1,773 posts)Each passing hour narrows the probability window.
Just a guess.
MSMITH33156
(879 posts)that drops he will run another simulation.
budkin
(6,699 posts)timlot
(456 posts)If he is right. He will supplant Rove and co as the new Oracle of election predictions.
zebe83
(143 posts)yellowcanine
(35,693 posts)Particularly since he turned Florida light blue. He should be using prophecy instead of polls to predict who is going to win.
CthulhusEvilCousin
(209 posts)wish that was so, but they love their political party moreso than they love God. If they did love God, they wouldn't discard their principles for short term political gain, since the Kingdom of God is specifically "not of this world."