Ed FiveThirtyEight GE Forecast

I'm going to start paying more attention to the Senate forecasts. Polls and Polls-plus showing a near-even chance of a Democratic Senate, nowcast still showing R's up 54-46. Lots of close races for Republican incumbants; the only likely R pickup appears to be Harry Reid's seat in Nevada. The Dems should have lost that in 2010, but the R's picked a complete loon for a candidate.
 
MikeG said:
How come the odds have moved so much towards the Republicans in the last few months on that poll tracker graph?
Click to expand...

Trebuchet said:
Dunno, I was wondering that myself.
Click to expand...

Isn't there some popular wisdom that significant numbers of unaffiliated/swing voters want to try balancing out the power by voting opposite for the Presidency and House?

Or did I imagine that?
 
Eighteen of the seats Republicans are predicted to win are incumbents. Incumbents tend to trend better the closer to the election. Louisiana is predicted for Republicans, and is currently held by a Republican. As said before, Nevada is the only expected pick up for Republicans.

Democrats are predicted to pick up Illinois, Indiana, New Hampshire, Pennsylvania, and Wisconsin. A number of those are even against incumbents. That is a pretty good swing.

Who knows though, as we get closer there can be shake up as fallout from Trump trickles down.
 
[IMGw=640]http://i.imgur.com/BQT83jA.jpg[/IMGw]

And this is without pussygate fully factored in.
 
Last edited:
Tony Stark said:
[IMGw=640]http://i.imgur.com/BQT83jA.jpg[/IMGw]

And this is without pussgate fully factored in.
Click to expand...

It says, up from about 48 hours ago....

NY Times - 87% +4
538 - 83% +1
Daily Kos - 95% +4
Huffington Post - 88% +4
PredictWise - 88% +2
PEC - 97% +4

Again, all of these are aggregates and they use many of the same polls, but they weight the results differently.

In terms of the RCP average, because they just added the Atlantic poll, Hillary's now on an average there of 6.5+. This is still without the full impact of the weekend's developments as some of those polls are before/after the Friday **** storm. The gap far exceeds anything during Obama/Romney. It's about the same as at this stage Obama/McCain.
 
bonzombiekitty said:
538 Polls only now at 84% and Polls Plus at 80.9% (highest yet).
Click to expand...

I wish 538 would change their algorithm for showing EC votes. If you take Now-Cast vs Polls Only vs Polls Plus there are only two different one-state swings. Now-cast gives Hillary both AZ and IA and 339.7 EC votes. Why? By my count that would be 365 EC votes.

And then, starting with their reduced total,... Polls Only takes away AZ but gives her IA and 334.5 EC votes. Why? There are 11 votes in AZ. Polls Plus gives her neither of those, but 319.7 EC votes. Again, two states with a total of 17 EC votes, but they reduce by 20. Makes no sense. It should be relatively simple - they have the EC votes and know that with the exception of NE(2) and ME(2) every state is winner take all.
 
Foolmewunz said:
I wish 538 would change their algorithm for showing EC votes. If you take Now-Cast vs Polls Only vs Polls Plus there are only two different one-state swings. Now-cast gives Hillary both AZ and IA and 339.7 EC votes. Why? By my count that would be 365 EC votes.

And then, starting with their reduced total,... Polls Only takes away AZ but gives her IA and 334.5 EC votes. Why? There are 11 votes in AZ. Polls Plus gives her neither of those, but 319.7 EC votes. Again, two states with a total of 17 EC votes, but they reduce by 20. Makes no sense. It should be relatively simple - they have the EC votes and know that with the exception of NE(2) and ME(2) every state is winner take all.
Click to expand...

I think it's just how they run the simulations. If you win a given state a certain number of times in the simulation, then you are given part of its electoral vote. It's intended to be a reflection of the distribution of EVs from all the simulations, rather than a whole number - i.e. when you average it all out after the simulations Hillary has a mean of 334.5 EVs.

It's actually a useful way of doing it. It's a high level display of the certainty of the results. The closer they are to the real, whole number, possibility the stronger their certainty that their predictions are correct. And a wider spread indicates higher certainty that it won't get closed.
 
Last edited:
bonzombiekitty said:
I think it's just how they run the simulations. If you win a given state a certain number of times in the simulation, then you are given part of its electoral vote. It's intended to be a reflection of the distribution of EVs from all the simulations, rather than a whole number - i.e. when you average it all out after the simulations Hillary has a mean of 334.5 EVs.

It's actually a useful way of doing it. It's a high level display of the certainty of the results.
Click to expand...

Yeah, I get it but it's like averaging fifty judges in a boxing match and coming out with scores like 127.5 to 84... they're impossible with the points you are required to give out and the number of judges giving them. You know the Electoral votes for each state. If a state moves over to one or the other, move the finite number of votes.
 
bonzombiekitty said:
I think it's just how they run the simulations. If you win a given state a certain number of times in the simulation, then you are given part of its electoral vote. It's intended to be a reflection of the distribution of EVs from all the simulations, rather than a whole number - i.e. when you average it all out after the simulations Hillary has a mean of 334.5 EVs.

It's actually a useful way of doing it. It's a high level display of the certainty of the results. The closer they are to the real, whole number, possibility the stronger their certainty that their predictions are correct. And a wider spread indicates higher certainty that it won't get closed.
Click to expand...
I suspect that's the case. It's simple to check: hover over each state, multiply the percentage you get with the number of EVs of that state, and add them up. :)
 
Foolmewunz said:
Yeah, I get it but it's like averaging fifty judges in a boxing match and coming out with scores like 127.5 to 84... they're impossible with the points you are required to give out and the number of judges giving them. You know the Electoral votes for each state. If a state moves over to one or the other, move the finite number of votes.
Click to expand...

The point of those EV numbers in the model isn't to say "If Clinton wins all the states we predict she will win, this is the number of EVs she will have" it's more a display of the certainty of their predictions and who will ultimately win.
 
That would be more intuitive, but less informative.

It would effectively discretize the total EC estimate to whatever the EC value of the swingiest state is at the time. If they did that then a shift of 6 EC votes, say, could be just noise as a swing state flops over or could represent a substantive trend, and you'd have no way of knowing which until you dug deeper and looked at the individual states. Every time.
 
Foolmewunz said:
Yeah, I get it but it's like averaging fifty judges in a boxing match and coming out with scores like 127.5 to 84... they're impossible with the points you are required to give out and the number of judges giving them. You know the Electoral votes for each state. If a state moves over to one or the other, move the finite number of votes.
Click to expand...
No, because that's not mathematically accurate. The EC is the outcome of 54 separate races (*). The expected number of EVs is not necessarily simply the sum of all those races you're likely to win with 50+%.

And it's unfortunately also not as easy as I wrote in my previous post; I retract that. That would be the case if all those 54 races were independent, but they're not. If there's another scandal hitting Trump, it will decrease his chances in all those races, but not to the same extent. Or if Hillary says that ex-coalminers can screw it, it will hit her chances in WV, but not or hardly not in CA.

(*) number of races:
48 winner-take-all states
1 for DC
2 Maine electoral districts
3 Nebraska electoral districts
 
Foolmewunz said:
Yeah, I get it but it's like averaging fifty judges in a boxing match and coming out with scores like 127.5 to 84... they're impossible with the points you are required to give out and the number of judges giving them. You know the Electoral votes for each state. If a state moves over to one or the other, move the finite number of votes.
Click to expand...


Which is why they have the actual distribution from the simulations on the same page as the forecast: (This is the Polls-only version)



So, the most likely outcome is about 360 EV for Clinton, which is probably the most meaningful number.
 
You must log in or register to reply here.

Back
Top Bottom