2016 Postmortem

ashtonelijah

(340 posts) Tue Nov 6, 2012, 01:38 PM Nov 2012

Shouldn't Nate Silver Have Obama With 332 EVs, not 313?

I'm looking at Nate Silver's map. He has Obama either winning or leaning towards winning every swing state (including FL) except North Carolina. If that's the case, why does his model predict 313 EVs instead of 332?

7 replies

= new reply since forum marked as read

Highlight:

Shouldn't Nate Silver Have Obama With 332 EVs, not 313? (Original Post) ashtonelijah Nov 2012 OP

He runs thousands of simulations, so he takes the average result. nt geek tragedy Nov 2012 #1

because the ev prediction is an average of all the simulations dsc Nov 2012 #2

Doesn't work that way. HERVEPA Nov 2012 #3

So when Nate Silver says he accurately predicted 49 of 50 states in 2008... ashtonelijah Nov 2012 #4

Check out Dr Sam Wang amuse bouche Nov 2012 #5

Correct HERVEPA Nov 2012 #6

313 is the Median Number Hokie Nov 2012 #7

geek tragedy

(68,868 posts)

1. He runs thousands of simulations, so he takes the average result. nt

Reply to ashtonelijah (Original post)

Tue Nov 6, 2012, 01:39 PM

Nov 2012

dsc

(52,162 posts)

2. because the ev prediction is an average of all the simulations

Reply to ashtonelijah (Original post)

Tue Nov 6, 2012, 01:40 PM

Nov 2012

and not based on the map. The lean in Florida is minimal meaning he wins it in barely more simulations than he loses it in. Florida is only adding a couple of votes to the EV model due to that.

HERVEPA

(6,107 posts)

3. Doesn't work that way.

Reply to ashtonelijah (Original post)

Tue Nov 6, 2012, 01:40 PM

Nov 2012

The total is based on the percentage chance of winning each state, not the total of the electoral votes where he says he's winning.

ashtonelijah

(340 posts)

4. So when Nate Silver says he accurately predicted 49 of 50 states in 2008...

Reply to ashtonelijah (Original post)

Tue Nov 6, 2012, 01:41 PM

Nov 2012

He's talking about the map, not the averages, right?

amuse bouche

(3,657 posts)

5. Check out Dr Sam Wang

Reply to ashtonelijah (Reply #4)

Tue Nov 6, 2012, 01:47 PM

Nov 2012

I think his past predictions have been perfect. He gets on TV but for some reason, not as much hype as Nate

http://election.princeton.edu/

he is calculating at the moment but his last predictions were

SNAPSHOT as of November 6, 12:00PM EST:
Obama: 312
Romney: 226
Meta-margin: Obama +2.46%

Probability of Obama re-election: Random Drift 99.2%, Bayesian Prediction 100.0%

HERVEPA

(6,107 posts)

6. Correct

Reply to ashtonelijah (Reply #4)

Tue Nov 6, 2012, 01:47 PM

Nov 2012

Hokie

(4,288 posts)

7. 313 is the Median Number

Reply to ashtonelijah (Original post)

Tue Nov 6, 2012, 01:47 PM

Nov 2012

Nate plugs in all the polling data and economic factors into his model and runs many simulations. The output is the distribution of EV totals. The outcome with the highest probability is Obama 332 at 20%. However the median is at 313. At 313 the odds of the total being higher or lower are equal. Put another way if you took the sum of the length of all the bars above 313 and the sum of all the bars below 313 they would be equal.

Reply to this discussion