The Biden v Trump Thread

[Ebo]

This was updated 3 hours ago. When I last looked earlier this week, Biden had an 18 in 20 chance of winning the Electoral College vote with Trump 2 in 20. It is now Biden 19 in 20 and Trump 1 in 20.

https://projects.economist.com/us-2020-forecast/president

 

[Professor]

This was updated 3 hours ago. When I last looked earlier this week, Biden had an 18 in 20 chance of winning the Electoral College vote with Trump 2 in 20. It is now Biden 19 in 20 and Trump 1 in 20.

https://projects.economist.com/us-2020-forecast/president

It's been fluctuating between 94-97% for a few days. For context it was in the 70% region in early September. I think the margin is too high, though it's worth bearing in mind this is very much a Bayesian model, but the extent it relies on current polls, versus past voting and other factors is less clear.

The Nate Silver model is a bit more nuanced and has more iterations (40K versus 20K-essentially predicting using every data combination and determining the frequency of outcome like Dr Strange in Avengers).

https://projects.fivethirtyeight.com/2020-election-forecast/?cid=rrpromo

This is better for Trump, though like the Economist model has moved more towards Biden. The static nature of polls and betting odds is marked.


Betting is essentially 1/2 for Biden (or 1.52 in decimal odds) to 2/1 for Trump (2.9). This is an implied probability of 65% for Biden, 35% for Trump.

The issue is the poll variation to the vote. At the moment variation would need to be greater than that seen in 2016 for Trump to win, barring a major late swing. But entirely possible. As is a landslide for Biden. But Trump has a clear chance still. The decision does not rely on Florida but more likely the midwest.

 

[Professor]

This was updated 3 hours ago. When I last looked earlier this week, Biden had an 18 in 20 chance of winning the Electoral College vote with Trump 2 in 20. It is now Biden 19 in 20 and Trump 1 in 20.

https://projects.economist.com/us-2020-forecast/president

It's been fluctuating between 94-97% for a few days. For context it was in the 70% region in early September. I think the margin is too high, though it's worth bearing in mind this is very much a Bayesian model, but the extent it relies on current polls, versus past voting and other factors is less clear.

The Nate Silver model is a bit more nuanced and has more iterations (40K versus 20K-essentially predicting using every data combination and determining the frequency of outcome like Dr Strange in Avengers).

https://projects.fivethirtyeight.com/2020-election-forecast/?cid=rrpromo

This is better for Trump, though like the Economist model has moved more towards Biden. The static nature of polls and betting odds is marked.


Betting is essentially 1/2 for Biden (or 1.52 in decimal odds) to 2/1 for Trump (2.9). This is an implied probability of 65% for Biden, 35% for Trump.

The issue is the poll variation to the vote. At the moment variation would need to be greater than that seen in 2016 for Trump to win, barring a major late swing. But entirely possible. As is a landslide for Biden. But Trump has a clear chance still. The decision does not rely on Florida but more likely the midwest.

Two other points

1. As a modelling colleague always tells me the model is as good as the data. There is a lot of partisan polling so those showing huge leads for Biden or large Trump leads in Florida are more questionable.
2. "All models are wrong, some are useful". is very true. A probabilistic model tells you the likelihood or frequency of an event. They are rarely 'spot on' but often correct in a general outcome. The probability of Trump winning in 2016 was around 28%-Between 1 in 3 and 1 in 4, so the model had shown this as unlikely, but certainly possible. It seems less likely this time, but again not a crazy idea.

 

[Best_loser]

https://m.youtube.com/watch?v=AtUwq_rdo38

Just a few of the " best people " trump gave jobs to

 

[Monty]

https://twitter.com/ossoff/status/1321603398752415746

 

[Eddy the Beagle]

DOJ official confirms there is an ongoing FBI criminal investigation into Hunter Biden and his associates.

FBI are a disgrace. Been "ongoing" since 2019. Obviously just sitting on it and hoping fellow swamp creature Joe Biden wins and they can release details after the election.

Trump needs to win and impeach Joe. Much more, and much more credible, evidence, for corruption against Joe than the joke of impeach 45.

 

[mart6]

DOJ official confirms there is an ongoing FBI criminal investigation into Hunter Biden and his associates.

FBI are a disgrace. Been "ongoing" since 2019. Obviously just sitting on it and hoping fellow swamp creature Joe Biden wins and they can release details after the election.

Trump needs to win and impeach Joe. Much more, and much more credible, evidence, for corruption against Joe than the joke of impeach 45.

:lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol: :lol:

 

[Eddy the Beagle]

This is what happens when liberals venture outside their cosy little bubble.

Shocked and surprised that ordinary folk are not going to vote the way they should do and have been told to by their superiors like her.

If you don't vote for Joe you're not an autoworker.

https://www.youtube.com/watch?v=K6CMzLo9i1M

 

[realswan]

This reminds me of second Brexit vote, I think the silent majority will come out in force to vote Trump.

As a neutral, I find the odds on Trump winning at 2/1 to be super value.

Trump's rallies look impressive compared to Biden's. He has 80 million plus twitter followers compared to Biden's 11 million. If you put all the scandals aside Trump looks like he's want the country back to normal and Biden looks like he want to carry on the covid cautious line. So it depends on how many voters are in each camp and I certainly would not pay any attention to polls after their bad record of late.

 

[J_B]

The polls 4 years ago were nowhere near as bad as people are making out. The problem with American pollsters is that very many of them are aligned to the GOP or the Dems and they should be ignored. Taking the rest, the average Clinton lead on the day before the election was 2% in 2016 and that was very close to the national vote. The non aligned polls were incredibly accurate for the 2018 mid term elections.

The individual state polls were the least accurate in 2016. Most of them were reflecting the late Trump surge but by definition polls are 48/72 hours behind the curve by the time they are published And if there is a very late swing they won’t pick it up in time.

80 million votes already cast which is probably over 50% of what will be an unusually high turnout.

It will be very interesting.

 

[Tummer]

https://twitter.com/ossoff/status/1321603398752415746

Ah yes, Jon Ossoff is quite the authority on lying, according to his own campaign staffer.

https://www.youtube.com/watch?v=NrXHu-nzmxw

https://www.youtube.com/watch?v=yr7Y_9OYtWs

 

[DwightYorkeSuperstar]

I do not think polls collected in person accurately reflect people's voting intentions. If you're with a group of friends who you know will react negatively toward you if you were to respond with Trump/Tories for example, you're going to tell the pollster you're voting for somebody else.

This is common in workplaces in the UK where people are afraid to admit to voting Tory because there is usually someone who will jump at you, claiming you wish to starve children, sell the NHS to Donald Trump, are a racist or something else equally outrageous.

 

[Professor]

I do not think polls collected in person accurately reflect people's voting intentions. If you're with a group of friends who you know will react negatively toward you if you were to respond with Trump/Tories for example, you're going to tell the pollster you're voting for somebody else.

This is common in workplaces in the UK where people are afraid to admit to voting Tory because there is usually someone who will jump at you, claiming you wish to starve children, sell the NHS to Donald Trump, are a racist or something else equally outrageous.

Very few done that way now DYSS. Most are phone or online. Looking at poll data alone may be misleading and in reality the polling error could be at 5%, though the US polls are far more uniform than one may expect.

 

[DwightYorkeSuperstar]

I do not think polls collected in person accurately reflect people's voting intentions. If you're with a group of friends who you know will react negatively toward you if you were to respond with Trump/Tories for example, you're going to tell the pollster you're voting for somebody else.

This is common in workplaces in the UK where people are afraid to admit to voting Tory because there is usually someone who will jump at you, claiming you wish to starve children, sell the NHS to Donald Trump, are a racist or something else equally outrageous.

Very few done that way now DYSS. Most are phone or online. Looking at poll data alone may be misleading and in reality the polling error could be at 5%, though the US polls are far more uniform than one may expect.

That's good to know. Hopefully America remains peaceful after the result. Unfortunately I think it is inevitable we will see further riots from both sides. Let's hope no more innocent police officers and business owners are injured or killed when defending themselves and their property from the looters.

 

[Professor]

Very few done that way now DYSS. Most are phone or online. Looking at poll data alone may be misleading and in reality the polling error could be at 5%, though the US polls are far more uniform than one may expect.

That's good to know. Hopefully America remains peaceful after the result. Unfortunately I think it is inevitable we will see further riots from both sides. Let's hope no more innocent police officers and business owners are injured or killed when defending themselves and their property from the looters.

I get that feeling. Some very entrenched views. A bit like Israel where it’s hard for moderates to achieve compromise. Let’s hope not