Again, that's a complete misunderstanding of 538's models. They are emphatically about predicting the November 8 result, not about predicting one week ahead or something.
[RD] US 2016 election: Poll watching thread
- Thread starter dutchfire
- Start date
Again, that's a complete misunderstanding of 538's models. They are emphatically about predicting the November 8 result, not about predicting one week ahead or something.
onejayhawk
Afflicted with reason
No. They are about what will happen on 8 November if nothing changes. If what you said were true, there would be a wide difference between the Polls-only and Vote-now. There is not. One is a tweak of the other, mostly aggressively applying recent information. Regardless, both are snapshots of now not projections of the most likely future.
J
They're about what will happen on 8 November based on currently available data and what can be inferred from that data. Thus, it's not valid to say that 'things will change in a week', because the models take the likelihood of change into account. That's why the polls-plus model shows almost double the chance for Trump to win by 10+ than the now-cast shows.
If the polls-only model were saying 'this is what will happen on 8 November if nothing changes', then it'd be literally identical to the now-cast model.
It is possible to say, for instance, "I think there's a 100% chance that there'll be some cataclysmic event in October which will swing the election 15 points in Clinton's favour, and thus there's a 100% chance that she will win by 10+ points", but that's pure unsupported supposition that has absolutely no place in this discussion of data-based chance.
If the polls-only model were saying 'this is what will happen on 8 November if nothing changes', then it'd be literally identical to the now-cast model.
It is possible to say, for instance, "I think there's a 100% chance that there'll be some cataclysmic event in October which will swing the election 15 points in Clinton's favour, and thus there's a 100% chance that she will win by 10+ points", but that's pure unsupported supposition that has absolutely no place in this discussion of data-based chance.
metalhead
Angry Bartender
- Joined
- Apr 15, 2002
- Messages
- 8,031
Why do you insist on making totally irrelevant comparisons? A pollster would have to be a total idiot to use a midterm electorate to try to predict a general one.
Not as inexplicable as using primary exit polls instead of general election polling to try to predict demographics, but still. Come on, man.
Timsup2nothin
Deity
- Joined
- Apr 2, 2013
- Messages
- 46,737
Using the midterm electorate produces the outcome he wants. That's so obvious I'm surprised you even asked.
onejayhawk
Afflicted with reason
The problem is that the Now-cast is almost identical to the Polls-plus. The projections are very cautious and low confidence. It is more realistic to treat them as snapshots of the now.
On this you simply wrong. No one said the pollsters used the midterm electorate to try to predict the general. I also do not recall using exit polls to predict demographics. You are mixing apples and orangutangs to make fruit salad. How's that going for you?
J
metalhead
Angry Bartender
- Joined
- Apr 15, 2002
- Messages
- 8,031
That's exactly what you said. "Pollsters are reacting to being wrong in 2014." I don't know what you could have possibly meant by that, other than the midterm electorate is affecting their projections.
Seriously, you act enough like Donald Trump that you should drop your own pretense about why you're supporting him. You like him, because he's a lot like you. Full of crap and doesn't care who knows it.
onejayhawk
Afflicted with reason
That is what I said. It is not what you said. It's almost mutually exclusive of what you said, so you were almost exactly wrong. Good show.
That's exactly what you said. "Pollsters are reacting to being wrong in 2014."
If you did not understand, why not ask before making an error. Pollsters with egg on their face might reconsider their methods to avoid similar missteps.
J
Timsup2nothin
Deity
- Joined
- Apr 2, 2013
- Messages
- 46,737
Another rock solid observation by Metalhead as J tries to deny what he said in a cloud of ridiculous obfuscations. Positively Trumpian.
Why do we have to be J's chosen village?
onejayhawk
Afflicted with reason
Nate Silver writes better than myself. He sums things up nicely.
http://fivethirtyeight.com/features...-if-the-polls-still-look-like-this-in-a-week/
J
http://fivethirtyeight.com/features...-if-the-polls-still-look-like-this-in-a-week/
Polls-only assumes that there’s still a lot of uncertainty about the outcome. But it also mostly assumes1 that the current condition of the race — Clinton ahead by around 2 points — is a statistically unbiased prediction of the Nov. 8 outcome. In other words, it assumes that Clinton is as likely to continue losing ground as opposed to regaining ground from this point forward.
J
metalhead
Angry Bartender
- Joined
- Apr 15, 2002
- Messages
- 8,031
I just don't know how you can be this breathtakingly stupid. You keep arguing you didn't say something where THIS IS EXACTLY WHAT I SAID YOU SAID. You're saying they should use 2014 data to tweak their projections. THAT IS THE ONLY THING YOU COULD BE TALKING ABOUT BECAUSE OF HOW POLLING WORKS.
onejayhawk
Afflicted with reason
My only point is that 2014, plus several European elections, have pollsters reevaluating their methods.
Back on point, the margin is under 1% pretty much everywhere. Will the color flip from blue to red in the next couple of days?
J
Moderator Action: and here this specific discussion ends and everyone goes back to actually discussing current polling.
onejayhawk
Afflicted with reason
It did not happen in the next couple of days. Hillary had a small bounce back which could indicate Trump's momentum is slowing.
It is getting time to start looking at the Senate and the House races. For the moment, RCP has the Democrats picking up four Senate seats and the Republicans one, so the net is three blue. The only major changes from six months ago are two races--Florida and Colorado--which were thought competitive are no longer, one for each party. The house is largely unchanged, though there are many more Republican seats in the toss-up group.
http://www.realclearpolitics.com/epolls/2016/senate/2016_elections_senate_map_no_toss_ups.html
http://www.realclearpolitics.com/epolls/2016/house/2016_elections_house_map.html
J
It is getting time to start looking at the Senate and the House races. For the moment, RCP has the Democrats picking up four Senate seats and the Republicans one, so the net is three blue. The only major changes from six months ago are two races--Florida and Colorado--which were thought competitive are no longer, one for each party. The house is largely unchanged, though there are many more Republican seats in the toss-up group.
http://www.realclearpolitics.com/epolls/2016/senate/2016_elections_senate_map_no_toss_ups.html
http://www.realclearpolitics.com/epolls/2016/house/2016_elections_house_map.html
J
Clinton seems to be rising in Florida. After a bunch of polls showing Trump winning the NYT and a small Monmouth poll are giving it to her.
Of course if Florida turns blue on election day Trump has no chance.
http://www.realclearpolitics.com/ep...rump_vs_clinton_vs_johnson_vs_stein-5963.html
Of course if Florida turns blue on election day Trump has no chance.
http://www.realclearpolitics.com/ep...rump_vs_clinton_vs_johnson_vs_stein-5963.html
Yeah this seems to be correct... I just made a quick stop on 270towin.com to verify, and even if you give Trump NV, CO, NC, VA, ME, NH, OH, IA, and IN plus all the "Red States" he still loses without FL.
Spoiler :
onejayhawk
Afflicted with reason
Timsup2nothin
Deity
- Joined
- Apr 2, 2013
- Messages
- 46,737
Okay...so, both Michigan and PA are likely to remain blue. We all knew that. Is there some point here that I'm missing?
onejayhawk
Afflicted with reason
I stated that Michigan (currently blue) is likely to change colors before Pennsylvania (also blue). The parallel was that Missouri (red) is likely to change colors before Florida (also red). The statement has to do with the order in which the party in the lead changes if there is a change, not whether such change is likely or unlikely.
J
