|
Vorik posted:There shouldn't be such a large discrepancy. You have double digit national polls while a bunch of states are in toss up range or in the 3~5 range and trending down. IIRC Nate talked about how the difference could be explained by already blue states turning out even higher number of voters as well as solid red states, which although had a large shift towards biden, they will still remain solid red. Those are two scenarios which could be skewing the national polls and which won't actually mean anything for the electoral college. I'm not sure it makes sense to say "a bunch of states are in toss up range" on the basis of one poll where Biden was winning Pennsylvania by 4 points, but if you want to do that: https://twitter.com/PpollingNumbers/status/1318008453340811265
|
#
¿
Oct 20, 2020 04:39
|
|
|
|
| # ¿ Apr 28, 2024 11:18 |
|
|
WWWWWWWWWWWWWWWWWW posted:https://www.toptradeguru.com/news/dems-in-absolute-panic-as-early-data-shatters-narrative-in-key-states/ The author probably used inspect element to modify the plots. You can use the same source and they don't look like the version in the article.
|
|
#
¿
Oct 20, 2020 04:55
|
|
|
Vorik posted:Yeah I think I'll go by what the actual polls from high quality pollsters say rather than whatever voodoo magic graph the owners of that website come up with. You realize that this is saying you trust the results of one poll over the collective results of multiple polls.
|
|
#
¿
Oct 20, 2020 14:42
|
|
|
TyrantWD posted:AZ partisan lean relative to PA is basically the same as it was in 2016. Biden being up 3 in AZ is a lot like him being down 2 in Texas. It is more a reflection of Biden's national lead, than some state specific transformation like we saw in VA, where Northern VA became an extension of DC/MD. If you say this, it seems to me you should also take the Arizona polls showing the race not tightening as evidence the race is not tightening in Pennsylvania.
|
|
#
¿
Oct 20, 2020 15:48
|
|
|
TwoQuestions posted:I've also got no idea how they see Biden winning while 3 points behind where the polls say he is, being that he's hanging by a thread in FL and NC, and not doing too much better in PA and AZ. I don't see how his polling is strong enough to withstand a 2016 style error, especially when the SCOTUS inevitably weighs in. That NYT link is taking the polling errors from 2016 and 2012 and applying them to the state poll averages, so if 2020 polls are exactly as wrong as 2016 polls Biden would win Florida by less than 1%. It's not especially important that Florida and Wisconsin are close in that scenario, since Biden gets over 270 electoral votes with PA and MN which he wins by 2 points. It's just a rhetorical device though, since 2020 polls won't be off by exactly the same margins as 2016. The point is that Trump needs a larger error than 2016 to win (and absent any evidence, it's most likely that 2020 polls aren't overestimating Biden as much as 2016 polls overestimated Clinton).
|
|
#
¿
Oct 21, 2020 01:13
|
|
|
Grouchio posted:What's Ohio and Georgia flipping then? It's almost nothing, they went from barely over 50% to barely under 50%. The 538 forecast just doesn't show tossups.
|
|
#
¿
Oct 21, 2020 04:40
|
|
|
Grouchio posted:Why would they hurt Trump if those fake emails were Proud Boys threatening voters to vote republican? Unless Iran was expecting to get caught, wouldn't their operation have gotten Trump more votes? If you got a threatening email from the Proud Boys telling you to vote republican, would it make you more likely to vote republican?
|
|
#
¿
Oct 22, 2020 01:53
|
|
|
JazzFlight posted:Doesn't it make people's tap water flammable and also contribute to earthquakes? Like there's some weird chemicals involved in the process that fucks poo poo up? I don't think it's entirely clear whether it's related to flammable tap water. Fracking fluid is mostly water and sand, if tap water is catching fire that's methane and not some weird oil industry chemical (but fracking should be regulated better). The hydraulic fracturing itself does not cause earthquakes, but induced earthquakes happen in places like Oklahoma because of deep wastewater injection. The wastewater injection can be related to fracked wells or conventional natural gas. edit: in general I don't think there's a big difference in safety or environmental impact between fracking and conventional natural gas, it's an issue of regulation.
|
|
#
¿
Oct 23, 2020 15:49
|
|
|
CAT INTERCEPTOR posted:How the hell is there a 50/50 chance for any of that poo poo, he's geniunely just making it up If the polls move they're equally likely to move toward Trump as Biden, a polling error is equally likely to favor Trump as Biden, if both of those happen the election is a toss up. It's an order of magnitude estimation that comes up with the same probability of a Trump win as the 538 model. It's just a way of explaining the 1/8 chance of Trump winning. You could disagree with the 50% (like if polls move 1% toward Trump and there's a 1% polling error favoring Trump, Biden probably wins) but that defeats the purpose. brugroffil posted:But there's not a 50/50 chance of the polls sliding to Trump. That's like saying I have a 50/50 shot of winning the lottery because I either win or I don't. A priori there's a 50/50 chance that polling on election day will be better for Trump than polling now. If you don't know what direction the polls will move, the safest assumption is they're equally likely to go in either direction. You could argue that polls are more likely to move in one direction (in the 538 model they're more likely to move toward the mean), but if you don't support that with data it's punditry which Nate sucks at and if you do support it with data it overcomplicates your order of magnitude estimation.
|
|
#
¿
Oct 24, 2020 01:47
|
|
|
Spiritus Nox posted:No, there isn't necessarily If you don't have any other knowledge, you have to assume that polls are equally likely to move in either direction. In reality it probably isn't equally likely. The 538 model assumes that polls are more likely to move toward Trump. But the point of the 0.5^3 tweet is to contextualize the model's 1/8 chance of Trump winning, not to make a prediction. The tweet is just Nate stating the obvious in a kind of dumb way.
|
|
#
¿
Oct 24, 2020 02:48
|
|
|
The uncertainty in the 538 forecast because their win probability is based on 40,000 simulations is about +/- 0.4%. A 1% change in the displayed win probability isn't meaningful: not only does it have no practical meaning, it's not even worth discussing which polls explain Biden going from 88% to 87%.
|
|
#
¿
Oct 24, 2020 16:29
|
|
|
TyrantWD posted:Polls well off in Florida by 4 points in 2018 in favor of the GOP, even after pollsters learned the lessons of 2016, and in the middle of a blue wave that somehow hit the entire country except Florida. Democrats haven't won state wide in FL since 2012, and are lagging in the early vote in Miami-Dade. All of those factors in a state where Biden is only up 2, doesn't sound like a recipe for success. Polls having been off before is not a reason to believe that polls now will be off in the same direction. For example Obama beat his polls in Florida in 2012.
|
|
#
¿
Oct 25, 2020 18:41
|
|
|
TyrantWD posted:https://www.wsbtv.com/news/local/at...HGNWHJFBVHKUCY/ Polls are random samples that have some intrinsic variability, even for the highest quality pollsters. This is why election forecasters use an average instead of the results of one poll. Of course Republicans look better if you cherry pick the best poll for Republicans out of the entire past month. edit: also that poll result is not inconsistent with Democrats winning James Garfield fucked around with this message at 19:05 on Oct 25, 2020 |
|
#
¿
Oct 25, 2020 19:00
|
|
|
That entire article about Biden running behind Clinton in Philadelphia is based on a crosstab with n=54, because of course it is.
|
|
#
¿
Oct 27, 2020 00:43
|
|
|
Space Gopher posted:And my point is that we had that exact scenario in Seattle, about a year ago. It also happened in the 2018 Senate election in Arizona with a fully Republican state government and Trump tweeting about fraud, and they didn't manage to throw it to McSally. Even if the Republican state officials there are different, it's not a sign in favor of throwing the election to Trump.
|
|
#
¿
Oct 27, 2020 04:14
|
|
|
the next act of Congress after adding 16 seats to the Supreme Court and admitting Washington D.C. as 70 states should require Salena Zito to write two years worth of Biden voter profiles
|
|
#
¿
Oct 27, 2020 22:21
|
|
|
Grouchio posted:Why are we taking Michigan and Wisconsin for granted again? I don't know who's taking it for granted, but in the 538 polling average that includes republicans making GBS threads out methodologically questionable polls that are 10 points more republican than anything else Biden leads by 8 points in Michigan and 7 points in Wisconsin. There are multiple safe red states polling closer than Michigan.
|
|
#
¿
Oct 28, 2020 03:56
|
|
|
Grayly Squirrel posted:I was speaking more to the 24% chance of him picking up at least one Clinton state-- especially since he cross links the state results. Three Clinton states with a ~10% chance = 1-(.90)^3= about a 24% chance, but that assumes each event is independent. Which they aren't. Supposedly. On the 538 map Trump has 2% in Oregon, 3% in Colorado and New Mexico, 6% in Minnesota, 10% in Maine and Nevada, 12% in New Hampshire. If those are independent it's 38% for Trump to win a 2016 Clinton state, so there's some room for the outcomes to be correlated and still get 24%. From the map tool MN, NV, and NH don't seem strongly correlated with one another. I think you could safely argue that 538 has too high of probabilities for Trump winning individual 2016 Clinton states (the Economist has all of those states except for NV and barely NH as safe blue, and 538 has things like a ~1% chance that Trump wins Oregon and Biden still wins the election) but given their state probabilities I don't think a 24% chance that Trump wins a 2016 Clinton state is obviously wrong. James Garfield fucked around with this message at 22:36 on Oct 28, 2020 |
|
#
¿
Oct 28, 2020 22:34
|
|
|
nachos posted:Is the la times tracking poll still a thing? I have some, uh, fond memories of that in 2016 The USC tracking poll that kept getting coverage for showing Trump leading? It's still around with the flaw that resulted in Trump winning the popular vote fixed, and it agrees with all the other polls.
|
|
#
¿
Oct 29, 2020 01:06
|
|
|
Grouchio posted:What about that Siena poll from yesterday that had Biden +5? You know the pollster that had it at +9 in Sept? It does not exist
|
|
#
¿
Oct 30, 2020 03:37
|
|
|
Fojar38 posted:A little known factoid, the polls for 2018 were among the most accurate on record. It really is weird how thoroughly 2018 got memory holed.
|
|
#
¿
Oct 30, 2020 04:40
|
|
|
Trafalgar posted more crosstabs https://twitter.com/JMilesColeman/status/1321993837875789824
|
|
#
¿
Oct 30, 2020 14:31
|
|
|
McCloud posted:Has this been posted yet? Biden and Clinton both had campaigns that try to get people to vote for them, so they will both lose Trump 2020 does not so he will win Not sure why Trump won in 2016 though
|
|
#
¿
Oct 31, 2020 00:17
|
|
|
Majorian posted:I'd like to see the tweet if you're going to make a claim like this, please. Very easy to google https://twitter.com/EoinHiggins_/status/1306260608585695234 https://twitter.com/EoinHiggins_/status/1307445786620440576 https://twitter.com/EoinHiggins_/status/1307110334898016257
|
|
#
¿
Oct 31, 2020 00:41
|
|
|
Maybe the tweets saying Trump was the clear favorite and Biden can't win because he doesn't knock doors weren't saying Biden was losing, but instead the prelude to a 1000 word nate silver thread about how Biden might win even though Trump is ahead.
|
|
#
¿
Oct 31, 2020 00:58
|
|
|
The story where Republicans run a quiet fascist in 2024 and win seems to have a serious weakness in that it assumes the quiet fascist will beat a loud fascist in the Republican primary.TyrantWD posted:It does make me wonder what happened to the Hispanic voters who flocked to Bernie. Is this 2016 all over again and they voted Bernie because they don’t like Biden? Young Hispanics are Trump’s best minority demo, and were somehow also one of the cornerstones of Bernie’s 2020 run. In the 2016 primary a meaningful part of Bernie's coalition was a chunk of disproportionately conservative white male voters who supported him because he wasn't Hillary Clinton. You could maybe take it as foreshadowing of the general election but I don't know how valid it would be. I don't think anything like that happened this year, and Bernie not doing as well at least doesn't contradict it.
|
|
#
¿
Oct 31, 2020 03:50
|
|
|
eke out posted:Also he literally says trump needs to win PA by 4-5 points because voter fraud. Surely there are QAnons saying COVID is fake and all the deaths were actually shipped to Pennsylvania to steal the election.
|
|
#
¿
Oct 31, 2020 04:56
|
|
|
Why are we discussing the Biden campaign failing with <cherry picked demographic group> when Texas is close? Anyway I hope this  polling hasn't been posted yethttps://mobile.twitter.com/FrankLuntz/status/1322553981340561409 https://mobile.twitter.com/BenjySarlin/status/1322545158467047424
|
|
#
¿
Oct 31, 2020 19:28
|
|
|
Pobrecito posted:Weird they seemed to respond to candidate that actually made it a priority to reach out and engage them while offering them policies that would actually help improve their lives. The people that voted Bernie in the primary are overwhelmingly voting for dems in the general.
|
|
#
¿
Oct 31, 2020 20:04
|
|
|
No amount of pollster quality will stop polls from being random samples. One A+ poll is still less useful than an average.TyrantWD posted:A lot of Trump voters returning home, and who were always going to. Why would one poll in Iowa senate mean Republicans keep the Senate? NC has favored the dems more than Iowa for the whole election cycle.
|
|
#
¿
Nov 1, 2020 00:49
|
|
|
Grouchio posted:There are comments saying the polling group was only 200. Where'd they get that from? The poll sampled 800 people or about 200 in each district. The results by congressional district aren't as reliable as the overall result.
|
|
#
¿
Nov 1, 2020 01:27
|
|
|
Phlegmish posted:Washington state up to 98% of the 2016 total. Do they get to continue voting tomorrow? Washington is vote by mail and mail in ballots only need to be postmarked by election day.
|
|
#
¿
Nov 1, 2020 03:09
|
|
|
Nessus posted:https://twitter.com/Politics_Polls/status/1322708443959099393 Clinton lost Utah by 18 points in 2016, not counting the McMuffin votes. 538 has Trump+10 but their average is mostly survey monkey.
|
|
#
¿
Nov 1, 2020 03:49
|
|
|
He's a campaign manager too https://mobile.twitter.com/GregRubini/status/1322863870583734275
|
|
#
¿
Nov 1, 2020 18:30
|
|
|
Willo567 posted:This is probably a stupid question, but is the GOP for the most part likely going to play along with Trump's claims that the ballots should only be counted on Tuesday, and not beyond that? He tried the same thing in the 2018 Arizona senate election (and some California house elections) and the Republicans running the election didn't. The 2018 Arizona senate election was almost certainly closer than the presidential election will be. edit: Trump also can't get to 270 votes without at least one of MI, WI, PA, AZ, and NC, all of which favor Biden and have a Democratic governor and/or secretary of state who has to sign off on the results. James Garfield fucked around with this message at 22:17 on Nov 1, 2020 |
|
#
¿
Nov 1, 2020 22:13
|
|
|
TyrantWD posted:Isn't that because the election models correlate Nevada with Arizona, and WI/MI/AZ gets you to 269, and then all you need is the 1 vote from NE-2. Giving Nevada to Biden halves Trump's win chance because half of the Trump win scenarios involve polls being wrong enough across the board that Trump wins Nevada. If Biden wins Nevada there's at least one close state where polls weren't completely meaningless the way they'd have to be for Trump to win.
|
|
#
¿
Nov 3, 2020 02:42
|
|
|
Grondoth posted:Republicans are asking where ballots are stored, who will be transporting them, and when. It is extremely difficult for me to look at this as anything other than exactly the kind of thing you'd ask for if you wanted to physically steal them. Look up the projected margin in Pennsylvania on 538 and the number of votes in 2016, and consider that any ballots the Trump campaign stole wouldn't all be for Biden. It's very hard to make the arithmetic work, even if you think they can steal ballots. Also the county involved has voted heavily Republican for decades and refused.
|
|
#
¿
Nov 3, 2020 03:19
|
|
|
Grouchio posted:I'm surprised we aren't discussing Trump nakedly telling his operatives to go steal ballots in Pennsylvania. Whether he can steal many of them or not, logistics, if the postal police will counter fast and hard enough, stuff like that. Trump is not going to steal Pennsylvania against a projected margin of 300k votes by stealing ballots from a county with 240k people in it that voted R+18 in both 2012 and 2016.
|
|
#
¿
Nov 3, 2020 04:00
|
|
|
TyrantWD posted:If 3 swing state polls are off, and all 3 undercounted Trump support, it is far more likely we are seeing a systematic polling error again. If the 2018 polling error accurately reflects the 2020 polling error, Biden wins Texas.
|
|
#
¿
Nov 3, 2020 04:03
|
|
|
|
| # ¿ Apr 28, 2024 11:18 |
|
|
sexpig by night posted:But guys, have we considered the venerated Helmut Norpoth model I'm Trump winning Hawaii but losing Colorado and New Hampshire
|
|
#
¿
Nov 3, 2020 04:51
|
|