US 2020 Presidential Election prediction thread

[alfred russel]

This may be a matter of semantics, but I think everyone would agree that on election day 2016 Trump didn't have a 30% chance of winning--he had a ~100% chance.

That's wrong.  I know I definitely wouldn't agree, so no, not everyone would agree.  What you're saying is that the probability of pulling an ace out of a shuffled deck is not 1/13, but either 0 or 1.  The deck is already shuffled, so the top card is predetermined, so it's either an ace or not, right?

What lay people often fail to grasp, because it is pretty tricky and unintuitive to grasp, is that uncertainty due to imperfect information is for all purposes exactly the same as the uncertainty due to a truly random effect.  The probability of pulling an ace is the same before shuffling and after shuffling, it's still 1/13.  We had no way of knowing how the people who were going to vote would vote, even if it was predetermined, like a deck is after shuffling.  Unknowably predetermined is no different from unknown.

It isn't the same as a shuffled deck.

The respective campaigns (and motivated supporters) spent years trying to convince the public to support their side. By election day they either succeeded or failed--and the state of the electorate is discoverable with additional resources. A shuffled deck--it is part of the ethics of the game that you don't try to discover the identity of the card on the top deck and it is randomly determined. Whether a particular person will vote for Hillary or Trump is not random: both sides seek to influence the choice, and the success is freely open to discovery to anyone who wishes to investigate.

Put it this way: If you bet on the chance the top card is an ace on a randomly shuffled deck, with straight up odds, I would say that was foolish because your expected outcome was less than what you bet--regardless of outcome. But if you taught a political science course and said that Hillary ran the best campaign because she had a 70% chance of victory on election day, and she just got caught by 30% of bad luck, I would not agree with that either.

[Berkut]

Indeed.

This is like arguing that the chances of a 1 coming up on a six sided die is not 1/6, rather it is zero or 100%, depending on the nearly infinite number of variables that go into rolling a die, and if you could predict all of them, you could predict the outcome of the roll. It *might* be technically true (depending on your views on predetermination in the universe), but the distinction is functionally meaningless.

[DGuller]

It isn't the same as a shuffled deck.

The respective campaigns (and motivated supporters) spent years trying to convince the public to support their side. By election day they either succeeded or failed--and the state of the electorate is discoverable with additional resources. A shuffled deck--it is part of the ethics of the game that you don't try to discover the identity of the card on the top deck and it is randomly determined. Whether a particular person will vote for Hillary or Trump is not random: both sides seek to influence the choice, and the success is freely open to discovery to anyone who wishes to investigate.

It's true that there is a middle ground in my analogy.  If the unknown information can theoretically be knowable, but just isn't, then you can say that by failing to know what is knowable, your model is not as good as it can be.  Theoretically you can develop x-ray vision, so what is a 1/13 probability to an unsophisticated bettor can be a 1 or 0 probability to you.  Most of the time it's just pointless semantics and mental masturbation, though, because even something theoretically knowable is usually not practically knowable.  There is no reason to think that Nate Silver didn't act on the information he had or could reasonably get, so there is no good reason to say he made a bad call.

Put it this way: If you bet on the chance the top card is an ace on a randomly shuffled deck, with straight up odds, I would say that was foolish because your expected outcome was less than what you bet--regardless of outcome. But if you taught a political science course and said that Hillary ran the best campaign because she had a 70% chance of victory on election day, and she just got caught by 30% of bad luck, I would not agree with that either.

Most people suffer from results-oriented thinking.  It's a very common bias to think that if something happened, then it must've been likely to happen.  I wouldn't expect political science instructors to break the trend.

[The Brain]

The best way to predict the outcome is to control the outcome.

It's a bit like the old problem solving scenario: your mission is to catalogue with great precision all species and numbers of vertebrates on a small tropical island. You nuke the island. Results done, with little uncertainty.

[alfred russel]

Most people suffer from results-oriented thinking.  It's a very common bias to think that if something happened, then it must've been likely to happen.  I wouldn't expect political science instructors to break the trend.

I can sit in my room and flip coins and the probability of a heads is going to be about 50% over a long enough time horizon. If election day 2016 repeated a thousand times like in the movie groundhog day, I would posit that Trump would win all of those elections.

No one really understands what is going through the minds of 140 million american voters, but my understanding is that the election results are interpreted to show their preference, and not the probability weights before the election takes place.

[The Brain]

No one really understands what is going through the minds of 140 million american voters

[DGuller]

I can sit in my room and flip coins and the probability of a heads is going to be about 50% over a long enough time horizon. If election day 2016 repeated a thousand times like in the movie groundhog day, I would posit that Trump would win all of those elections.

You're comparing apples and oranges.  To make these two things analogous, your coin flips are repeated after the coin left your fingertips (I assume you let it fall to the ground), and with all the environmental factors the same each time.  In that case you will get either tails or heads 1000 times as well (ignoring chaos theory and quantum mechanics).

[PDH]

There is a zero chance something will have had happened until the time it does.

[crazy canuck]

There is a zero chance something will have had happened until the time it does.

There is no chance that nothing will happen.

[alfred russel]

I can sit in my room and flip coins and the probability of a heads is going to be about 50% over a long enough time horizon. If election day 2016 repeated a thousand times like in the movie groundhog day, I would posit that Trump would win all of those elections.

You're comparing apples and oranges.  To make these two things analogous, your coin flips are repeated after the coin left your fingertips (I assume you let it fall to the ground), and with all the environmental factors the same each time.  In that case you will get either tails or heads 1000 times as well (ignoring chaos theory and quantum mechanics).

This is really getting philosophical! Are votes not yet (all) cast analogous to a coin that has left my fingertips or one that has not yet been completely tossed? Is there really uncertainty in anything, or are we just collections of matter / energy playing out the inevitable course set during the big bang and defined by the laws of physics (assuming a solution to the uncertainty principle)?

But...to bring this back to a more tangible problem with Nate's model...the best polling isn't publicly available and he isn't using it -- the best polling is probably done by campaigns and parties trying to make resource allocation decisions. So when the candidates have recently come to Georgia, the questions start to be raised--"maybe their internals have it more likely that Georgia will be the tipping point state" than the free polls.

If we had gone camping this weekend, and were just now returning to civilization, you might tell me that based on your analysis there was a 75% chance that Michigan would beat Michigan State. I would argue your analysis is dumb and that the event is already decided. You would argue that we don't know the outcome, and your analysis takes into account all available information that we have at hand.

In your political science course, you might argue that in 2016, no one would be certain how the population would respond to different campaign tactics. Hillary listened to the experts and ran a campaign emphasizing health care and experience, while Trump made racist comments and was caught on tape bragging about sexual assault. You would teach that listening to the experts generally results in better election strategies, and so Hillary ran a better campaign. I would teach that the public apparently responded better to the racism and sexual assault boasts, and so contrary to the conventional wisdom during the period of the campaign, he is the one that won the election.

[Rex Francorum]

So here is my a bit spiced up prediction (to make things interesting). It might be bold but not entirely unimaginable.

[The Brain]

Reminds me that I never played Fortress America.

[Eddie Teach]

So here is my a bit spiced up prediction (to make things interesting). It might be bold but not entirely unimaginable.

Interesting that you rate both GA and TX blue, but PA red.

[PJL]

Well here's my rather pessimistic prediction.

https://projects.fivethirtyeight.com/trump-biden-election-map/#AZ:1,CO:0,FL:1,GA:1,MI:0,MN:0,N2:0,NH:0,NV:0,NC:1,OH:1,PA:0,TX:1,WI:0,M2:1,N1:1,N3:1,AK:1,HI:0,CA:0,OR:0,WA:0,NM:0,IL:0,NY:0,MA:0,VT:0,ME:0,RI:0,CT:0,NJ:0,MD:0,VA:0,DE:0,MT:1,WY:1,UT:1,ID:1,NE:1,SD:1,ND:1,KS:1,OK:1,AR:1,MO:1,IA:1,LA:1,MS:1,AL:1,TN:1,KY:1,IN:1,WV:1,SC:1,M1:0,DC:0

[FunkMonk]

Browsing right-wing Twitter is fascinating

https://socialbubble.so/