|
NYT is giving Trump 95% chance of winning North Carolina.
Betfair is giving 79% odds to trump for reelection.
Chinese yuan is tanking heavily.
looks like we have something on the cards, here
|
On November 04 2020 12:10 Stratos_speAr wrote:Show nested quote +On November 04 2020 12:09 LegalLord wrote:On November 04 2020 11:58 KwarK wrote:On November 04 2020 11:54 LegalLord wrote:On November 04 2020 11:49 KwarK wrote:On November 04 2020 11:45 LegalLord wrote:On November 04 2020 11:42 KwarK wrote:On November 04 2020 11:41 PhoenixVoid wrote:
I don't know if pollsters and election modellers will recover from this. This is looking like a 5-6 point miss despite all their pleas that they adjusted their models weighing for education and asserted that the shy Trump supporter and social desirability effect didn't exist, or performed well in the midterms. It appears they didn't catch the significant jump in support for Trump or voting depression for Biden in black and hispanic support, and the overwhelming turnout for Trump. Democrats aren't capturing that midterm energy at all.
This is looking like a very good map for Trump right now. Biden's slipping in Ohio and that's one of his few chances at keeping his election odds from not completely collapsing. If he loses Ohio, Pennsylvania is favouring Trump. Models assign probability, they don’t pick winners. You can’t say that the statistical model was wrong just because a coin cane up heads twice in a row. More like a die landing on 6 twice in a row, really. Polls were off by too much to simply write it off as "in the margin of error" two elections in a row. Which is why models like the 538 models gave Trump a serious chance to win, despite the polls, because they know that you can't trust polls blindly. 30-35% chance Trump in 2016? Reasonable chance he wins, that can be in the margin of error. That, followed by the 10% chance Trump has according to 538 in 2020? Less credible that the underlying data is reliable. Are you sure you know what a margin of error is? There isn't a margin of error with something saying 10% Trump, 90% Biden, just as there isn't with them saying 25% HH, 25% TT, 50% 1 heads 1 tails in 4 coinflips. A specific result doesn't imply an erroneous probability. Eh, that's a pretty weak argument on semantics. While it's possible for it to be off by that much purely "by chance," getting two results in a row that seem out-of-family is a perfectly good reason to wonder if your underlying assumptions were off. If I had a die that I could only roll twice and both times the result was a 6, that'd be a very good reason to suspect it might not be a fair die. Could just be a 3% random chance, but the odds are against that. Two 6's in a row is a 1 in 36 chance. As a person who plays several hobbies that roll large amounts of D6's, you would be amazed how many times I see utterly ridiculous rolls, e.g. rolling snake eyes twice in a row, or rolling three 20's on 3D20, or rolling five 1's out of five D6's. These things do happen.
They do. I've rolled 8 heads in a row as well on what was probably a "fair" coin. Happens with large sample sizes.
Doesn't mean it wouldn't be suspect in a small sample size to see something like that.
|
On November 04 2020 12:12 LegalLord wrote:Show nested quote +On November 04 2020 12:10 Stratos_speAr wrote:On November 04 2020 12:09 LegalLord wrote:On November 04 2020 11:58 KwarK wrote:On November 04 2020 11:54 LegalLord wrote:On November 04 2020 11:49 KwarK wrote:On November 04 2020 11:45 LegalLord wrote:On November 04 2020 11:42 KwarK wrote:On November 04 2020 11:41 PhoenixVoid wrote:
I don't know if pollsters and election modellers will recover from this. This is looking like a 5-6 point miss despite all their pleas that they adjusted their models weighing for education and asserted that the shy Trump supporter and social desirability effect didn't exist, or performed well in the midterms. It appears they didn't catch the significant jump in support for Trump or voting depression for Biden in black and hispanic support, and the overwhelming turnout for Trump. Democrats aren't capturing that midterm energy at all.
This is looking like a very good map for Trump right now. Biden's slipping in Ohio and that's one of his few chances at keeping his election odds from not completely collapsing. If he loses Ohio, Pennsylvania is favouring Trump. Models assign probability, they don’t pick winners. You can’t say that the statistical model was wrong just because a coin cane up heads twice in a row. More like a die landing on 6 twice in a row, really. Polls were off by too much to simply write it off as "in the margin of error" two elections in a row. Which is why models like the 538 models gave Trump a serious chance to win, despite the polls, because they know that you can't trust polls blindly. 30-35% chance Trump in 2016? Reasonable chance he wins, that can be in the margin of error. That, followed by the 10% chance Trump has according to 538 in 2020? Less credible that the underlying data is reliable. Are you sure you know what a margin of error is? There isn't a margin of error with something saying 10% Trump, 90% Biden, just as there isn't with them saying 25% HH, 25% TT, 50% 1 heads 1 tails in 4 coinflips. A specific result doesn't imply an erroneous probability. Eh, that's a pretty weak argument on semantics. While it's possible for it to be off by that much purely "by chance," getting two results in a row that seem out-of-family is a perfectly good reason to wonder if your underlying assumptions were off. If I had a die that I could only roll twice and both times the result was a 6, that'd be a very good reason to suspect it might not be a fair die. Could just be a 3% random chance, but the odds are against that. Two 6's in a row is a 1 in 36 chance. As a person who plays several hobbies that roll large amounts of D6's, you would be amazed how many times I see utterly ridiculous rolls, e.g. rolling snake eyes twice in a row, or rolling three 20's on 3D20, or rolling five 1's out of five D6's. These things do happen. They do. I've rolled 8 heads in a row as well on what was probably a "fair" coin. Happens with large sample sizes. Doesn't mean it wouldn't be suspect in a small sample size to see something like that.
Small sample sizes are actually more susceptible to variance and low odds.
Larger sample sizes are precisely what prove odds correct.
|
Fox's needle is giving NC 89% Biden, conversely. Weird situation.
|
On November 04 2020 12:05 shawster wrote:Show nested quote +On November 04 2020 11:58 KwarK wrote:On November 04 2020 11:54 LegalLord wrote:On November 04 2020 11:49 KwarK wrote:On November 04 2020 11:45 LegalLord wrote:On November 04 2020 11:42 KwarK wrote:On November 04 2020 11:41 PhoenixVoid wrote:
I don't know if pollsters and election modellers will recover from this. This is looking like a 5-6 point miss despite all their pleas that they adjusted their models weighing for education and asserted that the shy Trump supporter and social desirability effect didn't exist, or performed well in the midterms. It appears they didn't catch the significant jump in support for Trump or voting depression for Biden in black and hispanic support, and the overwhelming turnout for Trump. Democrats aren't capturing that midterm energy at all.
This is looking like a very good map for Trump right now. Biden's slipping in Ohio and that's one of his few chances at keeping his election odds from not completely collapsing. If he loses Ohio, Pennsylvania is favouring Trump. Models assign probability, they don’t pick winners. You can’t say that the statistical model was wrong just because a coin cane up heads twice in a row. More like a die landing on 6 twice in a row, really. Polls were off by too much to simply write it off as "in the margin of error" two elections in a row. Which is why models like the 538 models gave Trump a serious chance to win, despite the polls, because they know that you can't trust polls blindly. 30-35% chance Trump in 2016? Reasonable chance he wins, that can be in the margin of error. That, followed by the 10% chance Trump has according to 538 in 2020? Less credible that the underlying data is reliable. Are you sure you know what a margin of error is? There isn't a margin of error with something saying 10% Trump, 90% Biden, just as there isn't with them saying 25% HH, 25% TT, 50% 1 heads 1 tails in 4 coinflips. A specific result doesn't imply an erroneous probability. A poll has a margin of error, they say it'll be 52 +-4. That means any result in the 48-56 range is acceptable and anything outside that indicates a fuckup. A probability does not have a margin of error. If they say there's a 10% chance of it happening and it happens then that doesn't indicate a fuckup. I think this boils down to your interpretation of MoE. One usually defines a confidence interval using some cut-off, i.e. we believe with 95% certainty between 52 +- 4. At the end of the day, any event is possible in a statistical model it just may be unlikely. Furthermore, 52 +-4 does NOT imply that any value between 48 and 56 is equally likely. The point people are trying to make is that the results either lie outside or consistently to one side of the model's prediction. This is possible, sure, but it is unlikely and it is more likely that the polling methodology has not accounted for some bias. At the end of the day, polling is super hard. The 538 guys talk about the lack of survey response which leads to extreme selection bias which needs to be corrected.
Yeah, pretty much this.
|
Don't trust people on betting sites when it comes to politics. They're panicky morons.
|
On November 04 2020 12:14 Nevuk wrote:
Fox's needle is giving NC 89% Biden, conversely. Weird situation.
Nate Cohn just said sonething along the lines that NYT needle is way more accurate than Fox. Fox is really behaving weird today. They called virginia early today and after that trump took the big lead there
|
Dems just flipped a house seat in Missouri with a notably progressive candidate.
|
On November 04 2020 12:04 KwarK wrote:Show nested quote +On November 04 2020 12:01 Monochromatic wrote:On November 04 2020 11:57 DarkPlasmaBall wrote:On November 04 2020 11:52 Monochromatic wrote:On November 04 2020 11:50 Starlightsun wrote:On November 04 2020 11:43 FueledUpAndReadyToGo wrote:
Trump is way too competitive...
Four years of this awful administration, Biden who is not 'uniquely hated' like Clinton, a white house that infected itself with a disease just weeks ago, because of their pure cognitive ignorance, yet still we have to calculate pathways to win and Trump is over performing in several places.
It truly is outrageous. On the flipside, in Florida, 40% of people said they were better off than 4 years ago. Only 20% said they were worse off. This is despite a massive pandemic. At the end of the day, people care if their life is better. Trump has delivered on that front. Except 40% of Floridians are literally wrong. I'm glad you know their life situation better then they do. At the end of the day, the economy is what wins elections. It's been amazing under Trump. You often hear about how China grows GDP 6% a year, yet the US stock market has matched them in total return the past decade, a significant part of which was the past four years. Europe, during this same period? Less than 1/4th the return. I don't think you know what you're talking about. Trump promised growth of https://www.realclearpolitics.com/video/2017/12/16/trump_were_going_to_see_economy_growth_of_4_5_and_maybe_6_percent.htmlbut per his own government's figures delivered growth of about 2.5%/year. https://apps.bea.gov/iTable/iTable.cfm?reqid=19&step=2#reqid=19&step=2&isuri=1&1921=surveyWhat you've done here is confused the economy and the stock market.
I'm saying that despite the economy growing slower, his corporate tax cuts and pro-business policies allowed US corporations to match the valuation as in China. That's great, especially compared to the rest of the world.
|
United States10402 Posts
This is some weird shit right now. Arizona has biden up 9 points with most of that early vote coming in. Will trump have enough left to bring it back?
|
On November 04 2020 12:14 Stratos_speAr wrote:Show nested quote +On November 04 2020 12:12 LegalLord wrote:On November 04 2020 12:10 Stratos_speAr wrote:On November 04 2020 12:09 LegalLord wrote:On November 04 2020 11:58 KwarK wrote:On November 04 2020 11:54 LegalLord wrote:On November 04 2020 11:49 KwarK wrote:On November 04 2020 11:45 LegalLord wrote:On November 04 2020 11:42 KwarK wrote:On November 04 2020 11:41 PhoenixVoid wrote:
I don't know if pollsters and election modellers will recover from this. This is looking like a 5-6 point miss despite all their pleas that they adjusted their models weighing for education and asserted that the shy Trump supporter and social desirability effect didn't exist, or performed well in the midterms. It appears they didn't catch the significant jump in support for Trump or voting depression for Biden in black and hispanic support, and the overwhelming turnout for Trump. Democrats aren't capturing that midterm energy at all.
This is looking like a very good map for Trump right now. Biden's slipping in Ohio and that's one of his few chances at keeping his election odds from not completely collapsing. If he loses Ohio, Pennsylvania is favouring Trump. Models assign probability, they don’t pick winners. You can’t say that the statistical model was wrong just because a coin cane up heads twice in a row. More like a die landing on 6 twice in a row, really. Polls were off by too much to simply write it off as "in the margin of error" two elections in a row. Which is why models like the 538 models gave Trump a serious chance to win, despite the polls, because they know that you can't trust polls blindly. 30-35% chance Trump in 2016? Reasonable chance he wins, that can be in the margin of error. That, followed by the 10% chance Trump has according to 538 in 2020? Less credible that the underlying data is reliable. Are you sure you know what a margin of error is? There isn't a margin of error with something saying 10% Trump, 90% Biden, just as there isn't with them saying 25% HH, 25% TT, 50% 1 heads 1 tails in 4 coinflips. A specific result doesn't imply an erroneous probability. Eh, that's a pretty weak argument on semantics. While it's possible for it to be off by that much purely "by chance," getting two results in a row that seem out-of-family is a perfectly good reason to wonder if your underlying assumptions were off. If I had a die that I could only roll twice and both times the result was a 6, that'd be a very good reason to suspect it might not be a fair die. Could just be a 3% random chance, but the odds are against that. Two 6's in a row is a 1 in 36 chance. As a person who plays several hobbies that roll large amounts of D6's, you would be amazed how many times I see utterly ridiculous rolls, e.g. rolling snake eyes twice in a row, or rolling three 20's on 3D20, or rolling five 1's out of five D6's. These things do happen. They do. I've rolled 8 heads in a row as well on what was probably a "fair" coin. Happens with large sample sizes. Doesn't mean it wouldn't be suspect in a small sample size to see something like that. Small sample sizes are actually more susceptible to variance and low odds. Larger sample sizes are precisely what prove odds correct.
Sadly we have but one presidential election every four years, so we have to make do with inferences based on small sample sizes of "prediction was correct" vs "prediction was incorrect."
It's hard to justify assuming that polling was just off by a standard margin of error rather than systematically failing to capture the true story under the current circumstances. Possible, but not likely.
|
Most of Trump's economic accomplishments can be matched by just pointing to Obama and saying "but Obama was better on all of these metrics". Ie, he inherited a pretty roaring economy.
The economy was good under Trump, but has better growth under Obama and covid has wrecked the economy of tons of places.
|
On November 04 2020 12:12 Batmankills wrote:
Chinese yuan is tanking heavily.
I'm seeing a 1% drop, after a 1% gain when Biden led.
|
Arizona continues to look very good for Biden.
|
On November 04 2020 12:16 Batmankills wrote:Show nested quote +On November 04 2020 12:14 Nevuk wrote:
Fox's needle is giving NC 89% Biden, conversely. Weird situation. Nate Cohn just said sonething along the lines that NYT needle is way more accurate than Fox. Fox is really behaving weird today. They called virginia early today and after that trump took the big lead there
Nate Cohn also works for the NYT and designed the needle. Of course he's going to say it is better. Generally I would trust him more, as he has a solid stats background, but I think we are safest just waiting a few extra hours on NC
|
Hm, that's a pretty nice margin Biden has in Arizona from "early voter only" results. Thinking he might just manage to win Arizona after all.
|
United States43989 Posts
On November 04 2020 12:17 Monochromatic wrote:Show nested quote +On November 04 2020 12:04 KwarK wrote:On November 04 2020 12:01 Monochromatic wrote:On November 04 2020 11:57 DarkPlasmaBall wrote:On November 04 2020 11:52 Monochromatic wrote:On November 04 2020 11:50 Starlightsun wrote:On November 04 2020 11:43 FueledUpAndReadyToGo wrote:
Trump is way too competitive...
Four years of this awful administration, Biden who is not 'uniquely hated' like Clinton, a white house that infected itself with a disease just weeks ago, because of their pure cognitive ignorance, yet still we have to calculate pathways to win and Trump is over performing in several places.
It truly is outrageous. On the flipside, in Florida, 40% of people said they were better off than 4 years ago. Only 20% said they were worse off. This is despite a massive pandemic. At the end of the day, people care if their life is better. Trump has delivered on that front. Except 40% of Floridians are literally wrong. I'm glad you know their life situation better then they do. At the end of the day, the economy is what wins elections. It's been amazing under Trump. You often hear about how China grows GDP 6% a year, yet the US stock market has matched them in total return the past decade, a significant part of which was the past four years. Europe, during this same period? Less than 1/4th the return. I don't think you know what you're talking about. Trump promised growth of 4%, 5%, or maybe even 6% https://www.realclearpolitics.com/video/2017/12/16/trump_were_going_to_see_economy_growth_of_4_5_and_maybe_6_percent.htmlbut per his own government's figures delivered growth of about 2.5%/year. https://apps.bea.gov/iTable/iTable.cfm?reqid=19&step=2#reqid=19&step=2&isuri=1&1921=surveyWhat you've done here is confused the economy and the stock market. I'm saying that despite the economy growing slower, his corporate tax cuts and pro-business policies allowed US corporations to match the valuation as in China. That's great, especially compared to the rest of the world.
I think you're confused.
Let's say you work on a farm. In year 1 the farm produces 100 units of grain. In year 2 the farm produces 102 grain. That's 2% growth in the output of the farm. What you're trying to argue is that how much grain the farm produces isn't what matters, what actually matters is how much a speculator would pay to buy the farm. But there's two issues with that. Firstly, you don't own the farm, you just work there, it doesn't make any difference to you how much the speculator thinks it's worth on a given day. And secondly, the only thing that matters at the farm is how much food it produces. An idea of how much it's worth will go up and down daily but will never feed a single person.
Trump promised to deliver 5% increases in farm yield. He did not deliver that. He delivered increased opinions in the value of the farm, but that doesn't mean there's any more food for people to eat.
|
Trump is ahead by 5.5% in ohio now with 72% votes in
|
The NC numbers right now are such that I can how a model could see them as 95% for Trump or 80% for Biden, depending on what inputs you gave it. Most of the 95% plus reporting districts are red, but some of the margins in blue districts are pretty bad for Biden (conversely, some of the raw turnout numbers in the cities that are still sub-90% are good for him as well).
Personally I do not think NC will be called tonight, especially given that they tabulate mail-ins received post election. But I am no election scientist.
|
On November 04 2020 12:20 Nevuk wrote:Show nested quote +On November 04 2020 12:16 Batmankills wrote:On November 04 2020 12:14 Nevuk wrote:
Fox's needle is giving NC 89% Biden, conversely. Weird situation. Nate Cohn just said sonething along the lines that NYT needle is way more accurate than Fox. Fox is really behaving weird today. They called virginia early today and after that trump took the big lead there Nate Cohn also works for the NYT and designed the needle. Of course he's going to say it is better. Generally I would trust him more, as he has a solid stats background, but I think we are safest just waiting a few extra hours on NC
not arguing on this point at all.
|
|
|
|
|
|
EPT
