[subego]

The morning of the 8th, the Silverometer will be 70% Hillary, 30% Trump.

She'll squeak by.

[subego]

This is assuming Hill doesn't have the mother of all oppo burning a hole in the chamber.

[besson3c]

Oddly enough the chances of the Democrats gaining control of the Senate seems to be going up. Silver has it at 72.8% now.

[besson3c]

Originally Posted by subego 

This is assuming Hill doesn't have the mother of all oppo burning a hole in the chamber.

Or there is a weekend announcement that the Carlos Danger emails had nothing of significance in them? Would that be a positive or a negative at this point, as far as horse race politics go?

[besson3c]

I'm also interested in the accuracy of the NYTimes and Huffingtonpost models that are about running tens of thousands of simulations, and how this model differs from Silver's? The simulation models still have Clinton's chances at 90%+. I know that Silver's models also provide a historical accuracy skew to offset the results of various pollsters.

There are some pollsters that I've seen reported on Realclearpolitics with parent companies run by Republicans and Democrats, one would think that these sorts of pollsters would need some sort of obvious skew/offset, so I'm assuming that Silver's models are better, but I'm kind of curious whether this is right, wrong, over-simplified, etc. Maybe I'll research this when I have the time.

[subego]

Originally Posted by besson3c 

Or there is a weekend announcement that the Carlos Danger emails had nothing of significance in them? Would that be a positive or a negative at this point, as far as horse race politics go?

I don't see how this is even a question.

[besson3c]

Originally Posted by subego 

I don't see how this is even a question.

Well, my thinking is this...

If they come back really quickly with a "oooppss, false alarm" sort of message, who does this impact the most? The desperation of the right (for those that want to believe they were pushing the FBI a certain way), something that bolsters those that want to do the false equivalency thing relating Clinton to Trump (i.e. "they're both evil"), something that bolsters the right-wing base/conspiracy theory crowd, or just a taint on the FBI? How long will people remember this? How long before this reflects on polling numbers?

I honestly don't know how this shakes out.

[subego]

Originally Posted by besson3c 

I'm also interested in the accuracy of the NYTimes and Huffingtonpost models that are about running tens of thousands of simulations, and how this model differs from Silver's? The simulation models still have Clinton's chances at 90%+. I know that Silver's models also provide a historical accuracy skew to offset the results of various pollsters.

There are some pollsters that I've seen reported on Realclearpolitics with parent companies run by Republicans and Democrats, one would think that these sorts of pollsters would need some sort of obvious skew/offset, so I'm assuming that Silver's models are better, but I'm kind of curious whether this is right, wrong, over-simplified, etc. Maybe I'll research this when I have the time.

I'm thinking most of the magic is in the skew, the pollster ratings, and then weighing those ratings.

There's some submagic in how a poll gets weighted as it ages. Along with weighing methodology... likely voters, includes cellphones, etc.

[subego]

Originally Posted by besson3c 

Well, my thinking is this...

If they come back really quickly with a "oooppss, false alarm" sort of message, who does this impact the most? The desperation of the right (for those that want to believe they were pushing the FBI a certain way), something that bolsters those that want to do the false equivalency thing relating Clinton to Trump (i.e. "they're both evil"), something that bolsters the right-wing base/conspiracy theory crowd, or just a taint on the FBI? How long will people remember this? How long before this reflects on polling numbers?

I honestly don't know how this shakes out.

As luck would have it, this is a prediction thread.

The FBI will drag its feet. The longer it takes, the more plausible the argument Comey needed to make the announcement.

What follows from there is the Clinton machine running the counter-media campaign of a lifetime. It'll get interesting because the FBI itself will be off-limits in the attack. She's going to have to go for Comey's jugular.

Latching on to how Comey fumbled his delivery of the Weiner bomb is the most obvious strategy. Should have been more a missile. Watch for that getting spun into Comey trying to railroad her. Lots of pearl-clutching in the media about how the very foundation of democracy is under threat while...

She's taking a nasty hit. Probably not fatal, but shit just got real. If the oppo cavalry aren't headed in to save her, the scenario where Trump gets, like, 271 EV isn't an outlier anymore.

[subego]

Originally Posted by subego 

Lots of pearl-clutching in the media about how the very foundation of democracy is under threat

Ahem...

Washington Post - James Comey is damaging our democracy

[Waragainstsleep]

How long will Comey still have a job if Clinton wins?

Will Trump make him Supreme King of Law Enforcement?

[...deleted...]

[Chongo]

Originally Posted by Waragainstsleep 

How long will Comey still have a job if Clinton wins?

Will Trump make him Supreme King of Law Enforcement?

Clinton has already said that Lynch will remain AG if she wins.

[Chongo]

Originally Posted by subego 

Ahem...

Washington Post - James Comey is damaging our democracy

Ahem, ahem.
This happened the weekend before the 1992 election.
Flashback: Bill Clinton cheered 11th hour indictment that doomed Bush re-election | Washington Examiner

[Waragainstsleep]

Originally Posted by Chongo 

Clinton has already said that Lynch will remain AG if she wins.

Isn't Lynch a family friend? What does that have to do with Comey?

[P]

Originally Posted by besson3c 

I'm also interested in the accuracy of the NYTimes and Huffingtonpost models that are about running tens of thousands of simulations, and how this model differs from Silver's? The simulation models still have Clinton's chances at 90%+. I know that Silver's models also provide a historical accuracy skew to offset the results of various pollsters.

They wrote an article on that:

Election Update: Why Our Model Is More Bullish Than Others On Trump | FiveThirtyEight

The TL;DR is that insecurity is high this year because a) there are a lot of undecideds still; b) the model uses data from as far back as 1972, which is slightly more insecure, instead of relying on the highly secure data from 2000 and on, where polls had a better track record; c) 538 uses a t-distribution instead of a normal distribution, which essentially means that they are insecure about the standard deviation itself and therefore get "fatter tails"; d) 538 is certain that the polls are highly correlated, so if there is a miss in Ohio there is going to be a miss in Pennsylvania, which means that a big margin in the EC in the polls doesn't automatically translate into a big probability of an outcome.

[besson3c]

Whatever you think about Clinton and Trump, it is pretty troubling that this election might literally come down to Anthony Weiner's penis.

[subego]

Hillary has a massive dick problem.

[subego]

The proof is that joke almost works on more levels than there are words in the sentence.

[Chongo]

You mean the Weiner Probe?

[andi*pandi]

Prediction: Costco is going to do very good business this weekend. Stock up folks!

[subego]

Civil Wargate.

[The Final Dakar]

Predictions:

It's found Russia or Russian backed parties helped fund Trumps campaign

The GOP tries to block any SCOTUS hearings/votes

The GOP runs on impeaching Hillary in 2018

[andi*pandi]

So, how many cases of bottled water should I buy? I can't find the beef jerky, but have a decent stockpile of batteries.

[Chongo]

Originally Posted by andi*pandi 

So, how many cases of bottled water should I buy? I can't find the beef jerky, but have a decent stockpile of batteries.

Buy an Alexapure™ or other water filtration system. A good supply of Wise™ products is a good idea.

[subego]

Originally Posted by andi*pandi 

So, how many cases of bottled water should I buy? I can't find the beef jerky, but have a decent stockpile of batteries.

This is why guns are so important. With those you can just take other people's jerky.

[andi*pandi]

Good point. I don't think Costco sells them though. I might have to <blech> go to Walmart!

[The Final Dakar]

I forgot to add, Pence doesn't even sniff the nom In 2020. Scott Walker 2.0 at best

[subego]

Originally Posted by The Final Dakar 

I forgot to add, Pence doesn't even sniff the nom In 2020. Scott Walker 2.0 at best

Perhaps this indicates my true feelings are stronger than I let on, but I want Hillary to win enough, I'm not going to shake the karmic balance by making predictions which assume her victory.

Maybe two weeks ago, but not right now.

[The Final Dakar]

I mean all my predictions are based on the caveat that the stats experts know their game. Granted, two major presidential elections isn't a big sample size, but the state by state accuracy is encouraging.

If Hillary loses it likely won't be because the polls are wrong but some seismic events occurs in the remaining week. That I have no control over.

[subego]

Originally Posted by The Final Dakar 

I mean all my predictions are based on the caveat that the stats experts know their game. Granted, two major presidential elections isn't a big sample size, but the state by state accuracy is encouraging.

If Hillary loses it likely won't be because the polls are wrong but some seismic events occurs in the remaining week. That I have no control over.

How even does it have to be for answers the polls give us to be labeled "unclear".

If the trends continue, Trump will have about a 1 in 3 chance of winning. The platonic ideal of unclear would be 1 in 2.

I'd personally say 1 in 4 is where one can start speaking with confidence.

[The Final Dakar]

You've lost me. What are you replying to?

[subego]

I'm responding to the part where the "stats experts know their game".

If the Silverometer says 50% Clinton to 50% Trump, that means the polls are right no matter who wins.

If the Silverometer says 90% Clinton to 10% Trump, a Trump victory means the polls are wrong.

Somewhere between 90% Clinton and 50% Clinton on the Silverometer there's a switch between the polls calling a candidate and saying it's too uncertain.

I personally say that switch is near 66% Clinton

Does that make sense?

[OreoCookie]

Originally Posted by subego 

Does that make sense?

That's not how statistical modeling works. If Fivethirtyeight's model give 90:10 odds and Trump wins, then this just means a less likely event took place, not that the polls were necessarily wrong. Polls always come with various margins of errors, and Nate Silver's team tries to quantify those margins of errors and run all variations within those margins. Rinse and repeat that for 50 states and you see why things are so complicated and the probabilities shift much more rapidly than the rather small variations in the popular vote.

[subego]

You are correct. I am conflating the two ways there could be a Trump victory.

The first is as stated above. The roll of the dice lands on Trump.

The other is the polls are genuinely wrong. Like, Trump wins California or something.

What I'm saying is the roll of the dice is approaching 33% Trump. I feel that's near the point where predictions assuming a Clinton victory are misplaced.

It's like planning a vacation before playing Russian roulette.

With two bullets.

[The Final Dakar]

Originally Posted by subego 

You are correct. I am conflating the two ways there could be a Trump victory.

The first is as stated above. The roll of the dice lands on Trump.

The other is the polls are genuinely wrong. Like, Trump wins California or something.

What I'm saying is the roll of the dice is approaching 33% Trump. I feel that's near the point where predictions assuming a Clinton victory are misplaced.

It's like planning a vacation before playing Russian roulette.

With two bullets.

Sure, I buy that.

What I don't buy is Silver's model volatility. PEC on the other hand has been too certain for my taste. I've been using Nate's successor at the upshot, as my guide until election day sorts out the winners and the losers.

[OreoCookie]

Originally Posted by The Final Dakar 

Sure, I buy that.

What I don't buy is Silver's model volatility. PEC on the other hand has been too certain for my taste. I've been using Nate's successor at the upshot, as my guide until election day sorts out the winners and the losers.

Well, making polls and extracting quantitative predictions from them is in part a black art, because you have to add some things empirically (e. g. there are in some cases systematic and detectable differences between how people claim they vote in a poll and who they actually cast their ballot for), and if you look at the Polls-Only data, you see they are already much more stable than Polls Now and less stable than Polls Plus.

[The Final Dakar]

Polls now is garbage not be referenced.

[OreoCookie]

Originally Posted by The Final Dakar 

Polls now is garbage not be referenced.

True, but I used that just as an example where you need to make a choice between stability and reflecting the current mood. I like that it's there, because it gives you an indication of the slope of the trend line.

[The Final Dakar]

Time to post electoral maps!

I've decided to pull a Nate Silver, and have "polls only" and "polls-plus" variants.

Polls Only


Pretty Straight-forward, no surprises. Hillary doesn't do as well Obama in either '08 or '12, but pulls in an electoral margin that is irrefutable.

BUT
The thing to me is, there's a lot of factors that as not a data scientist, I'm not sure are reflected in the polls. These are:
• GOTV: This is the biggest one. If Hillary has tons of infrastructure and Trump has next to none, this has to contribute extra votes, right?

• Early-voting: I have no idea how early voters get reflected in polls. Are they less likely to answer polling calls because they already voted? Or do they show up as "very enthusiastic" voters?

• Shy voters: This one helps both Trump and Hillary. People too embarrassed to admit they hate Hillary so much they'll vote for Trump, and voters to embarrassed to admit they like Hillary while voting for her (Husband/Wife ticket splitting is the one example we've seen). Of course, perhaps these cancel each other out?

• Enthusiasm fluctuation: One of the stories of polls this year is how each candidate would take a negative hit after bad news would come out, then start to return to the mean after some time had passed. Do conservatives that left Trump after the Hollywood Access tape return when Election Day finally hits? Do swing voters disgusted by the Comey release last Friday bite the bullet on Tuesday?

• Third Party and Undecideds: This is one of bigger factors in Nate's outlier status, and not very unjustified. We have seen a greater amount of each due to each candidates favorable, but as a matter of course, the numbers have dropped as predicted after Labor Day. BUT given we started out with so many, we still have a sizable chunk of undecideds still remaining, enough to matter if they break one way. Polling has indicated undecideds leaning towards Clinton most of the season, but that only matters if they vote. As for Johnson voters, do they commit or have second thoughts in the booth? (I should add, I don't McMuffin winning Utah. His moment has passed. Trump voters have returned to the mean)

• Demographics: The secret sauce of all polling. Each firm creates what they think the electorate will look like come election day. If I understood how they determine this, I'd see it as less of voodoo. It's one of the reasons why 2014 was so far off (demographic predictions don't work if voter participation swings widely, I think).

This year we're expecting a few things: A. Apathetic and marginalized (see: racist) white voters reemerging to vote Trump. B. Hispanics turning out to vote against Trump. C. AAs turning out to vote against Trump. As far as I can tell, no one has talked about how the surge in Hispanic registrations has impacted polls. This to me is the biggest ? on election day. They have the power to swing Florida heavily, Arizona, and to make Texas almost purple. Coupled with the early voting numbers, this is where I'm adjusting my Polls-plus map the most. I don't think Florida will be close. Nevada has been decided. If black turnout is up, I see Ohio going light blue and Georgia and South Carolina as toss-ups leaning R.



Polls-Plus


This Hillary's best case scenario if GOTV, early voting, and demographics end up being things not accurately represented in the polls. In reality, I think she'll lose Iowa, Georgia, and SC by increasing margins, but that map looked so similar to the other one (two states difference: Arizona & Ohio) that I didn't post it (EV Total: 352). Consider that my 'hedge'. In either scenario, however, Clinton outperforms Obama's 2012 margin.

Oh yeah, in either map Utah should be hilariously closer than it should be, but never in place where Hillary had a serious chance.

---

As for senate predictions, polls only is 50, polls-plus best case 54 (Hedge: 52). Florida fascinates me, as polling indicates Rubio with a comfortable margin based on ticket-splitting, yet EV results indicate hispanics leaning D. This could end up an election night surprise. Kander is Missouri also has upset potential.

Now to see if anyone reads/cares.

[The Final Dakar]

I should add, this was mostly done before the Comey announcement. That adds a level of complexity to voter enthusiasm I decided to just ignore.

[The Final Dakar]

I should be flogged – I forgot one of the biggest hurdles I see for Democrats outside of the polls: Voter Suppression!
In states like North Carolina, Ohio, Wisconsin, this could have an impact on margins, electoral votes, or senate candidates. It has the power to negate, dampen or reverse gains from GOTV, EV, and demographics. Already we've seen the long lines and reduced sites and hours. It probably won't be enough to change the presidential outcome, but even in its court-reduced form it will likely effect down ballot races.

[The Final Dakar]

No one else wants to do electoral maps?

[subego]

I'm about to win with my first prediction. Why would I **** up my record?

[subego]

Originally Posted by subego 

The morning of the 8th, the Silverometer will be 70% Hillary, 30% Trump.

[The Final Dakar]

Originally Posted by subego 

I'm about to win with my first prediction. Why would I **** up my record?

*applause*

Of course, why the **** did it spike 6 points in a single day? It reeks of those shaky pollsters that get serious down the stretch.

Meanwhile the upshot sat pat at 84% the past three days.

[subego]

I must admit, I was assuming the trend from 10 days ago was going to go straight to 70, and not get there on a bounce.

For a good stretch I was cursing myself for not going 1 in 3.

[The Final Dakar]

Also:

Johnson <5%
Stein <1%

[subego]

It's sad Johnson probably won't break 5%.

I must admit the "don't reward the Libertarians for nominating a dork" argument is having more pull with me than I'd expect.

[The Final Dakar]

Given the libertarian philosophy, it was either that or a sociopath.