General Discussion
Related: Editorials & Other Articles, Issue Forums, Alliance Forums, Region ForumsQuestion about Nate Silver...
The guy predicting Obama's win.
Has he ever been dead wrong?
MADem
(135,425 posts)He's a libertarian who loves his work. He's a numbers geek. It's his "thing."
He also predicted victory for the Republicans in 2010.
He predicted Obama would win in the 2008 election within 1 percentage point.
The Right says Nate is a cheerleader for the Democrats but they forget he predicted the Republicans would pull it off in 2010.
forthemiddle
(1,381 posts)At that time, people on DU chose not to believe him, but he was of course, correct.
Maybe I am remembering that incorrectly?
Jennicut
(25,415 posts)Tnliberallee
(59 posts)Silver and Sam Wang of Princeton have both been dead on...Indiana was the only state Nate missed in 2008...Dr. Wang only missed the single EV vote that came out of Omaha.......Dr Wang has Obama winning also....
Shivering Jemmy
(900 posts)2010 (2009?) United Kingdom elections. His model did not work well for their parliamentary system.
pointsoflight
(1,372 posts)So his forecasts have been solid in both directions, and he's put them out there in the face of criticism on both sides.
cemaphonic
(4,138 posts)His model gives Romney about 1/7 chances of winning, largely on the possibility that the polling might be less accurate this cycle than before. So if Romney wins, it doesn't make him wrong, just says that the more improbable event happened.
That said, a Romney win would almost certainly end his 'polling guru' reputation. And yeah, he has been very accurate in the past.
pokerfan
(27,677 posts)Imagine it's NLHE and all the money's in the middle before the flop. Obama has AhAd vs. Romney's KsQs and we're only going to run it once. That's basically all Silver is saying.
Those pocket aces will lose 17.5% of the time.
pintobean
(18,101 posts)Reid won by 5.5 %.
Brickbat
(19,339 posts)He caught the trend in MN08 late two years ago. He adjusted his expectations late and ended up wrong on election day, but he was moving toward the outcome that happened.
http://elections.nytimes.com/2010/forecasts/house/minnesota/8
Cravaack won.
Viking12
(6,012 posts)Less data = less accurate.
Recursion
(56,582 posts)And he had McCain carrying I think Indiana in 2008.
I think he acknowledges his own personal leanings and hedges some closes races the other way because of that.
Archae
(46,335 posts)"Dewey Beats Truman" and whatnot.
Logical
(22,457 posts)Archae
(46,335 posts)And Truman held up a newspaper that said "Dewey Defeats Truman."
Logical
(22,457 posts)RomneyLies
(3,333 posts)Nate, as any statistician, deals in PROBABILITY.
Right now, the PROBABILITY of an Obama win on November 6 is 85.1%.
There is no way to make an accurate data driven prediction, you can only work in probability. Until probability reaches precisely 100%, there is always a chance for the opposite outcome.
Quixote1818
(28,946 posts)cleduc
(653 posts)The accuracy of his November 2008 presidential election predictionshe correctly predicted the winner of 49 of the 50 stateswon Silver further attention and commendation. The only state he missed was Indiana, which went for Barack Obama by 1%. He also correctly predicted the winner of all 35 Senate races that year.
He's also done very well in the 2008 primaries, the 2010 election, the 2012 primaries, senate races, and the Walker recall.
In very tight races/toss ups - almost coin tosses, Nate's been on the side of right 70+% of the time.
When Nate has given an election a 90% chance or better for the favorite, the favorite has never lost. And one could make a case that that makes Nate wrong because something less than 10% of those should have lost. I'm not making that case.
But that's a key point many seem to miss about Nate and his approach. On any given prediction where Nate says something is less than 100%, that also means there's some sort of a smaller chance it could go the other way. And what I have concluded is that on the high end, Nate may be a bit conservative with his numbers. Maybe Nate's 90% chance works out closer to 99% for example.
In the 80-90% category, I think he's been right slightly less than 80% of the time and therefore, not as conservative.
As Nate provides % of chance, his work cannot be evaluated over one contest. It must be evaluated over a series of contests. To date, over many contests, his work stands up as arguably the best in the business of what he does.
grantcart
(53,061 posts)The fewer polls that he can use the less viable the model.
That is why he is somewhat desperate for poll data.
nadinbrzezinski
(154,021 posts)he uses statistical analysis.
gollygee
(22,336 posts)FreeState
(10,572 posts)Ruby the Liberal
(26,219 posts)but would be voting for Gary Johnson if he was. It was in an interview the other day, but I don't remember which publication.