11-02-2020, 03:20 PM
I see that the thread is mostly given over to how much certain people hate Trump and his supporters. Whatever.
Modifications to my model:
I'm going to keep the 1.059 multiplier over 2016 (modeled on the 2004 election), even though it doesn't seem to be the conventional wisdom ?:
538:https://projects.fivethirtyeight.com/2020-election-forecast/
Sabato 2020:https://centerforpolitics.org/crystalball/2020-president/
Cook 2020:https://cookpolitical.com/sites/default/files/2020-10/EC%20Ratings.102820.pdf
Realclearpolitics: https://www.realclearpolitics.com/epolls/2020/president/2020_elections_electoral_college_map_no_toss_ups.html
With the 3rd party/write-in/"none of the above"/etc. getting 1.60% of the vote, a 1.059 modifier shakes out to Trump getting 48.63%(vs. 45.93%) and Biden getting 49.77%.
On a state by state (or district by district) basis, the minimum deviation (i.e., the worst result for Trump out of 45) for each state of district would be adivisorof 1.049 vote for one state, a maximum multiplier of 1.211 for Trump's best performing state, and a median multiplier of 1.045 (i.e., 50-50). Example: In 2016, Trump got 48.18% of the vote in PA. If the national vote improves x 1.059, then there is a 1/45 chance that he only gets 45.93% of the vote in PA, a 1/45 chance that he gets as much as 58.35%, and the median gain would be 50.35%. In all cases, if the miscellaneous vote were larger than 1.6%, then Trumps chances would improve a little (since the percentage needed to win PA by a plurality would decline), and if it were smaller, then the percentage needed to win a plurality would increase, and Trump's chances would decline.
I'd like to integrate the idea of elasticity into my model:
https://fivethirtyeight.com/features/ele...onal-mood/
...but I haven't done it yet. Maybe another day. It's largely covered within the standard deviation anyway, with teh more elastic entities having the wider swings and the less elastic ones being closer to the median.
My guess is that the chance I give Trump of winning UT is a bit low (36/45 i.e., 80%). This is an artifact of McMullen's 2020 run, where he got 21.54% of the vote, and Trump got 45.54%. To some extent, my model takes that into account (in my model, a candidate can win UT with a 46.41% plurality), but I suspect it underestimates his chances. Oh well. Que Sera Sera for now. It's only six EVs anyway.
The modifications to the model: I assume (partly based on his performance in 2008 as a VP candidate) that Biden puts at least a 1.030 divisor to Trump's chance of reelection in DE, knocking Trump's chances of winning DE from 2/45 to 0/45. Without modifications, my model gives Trump a 0/45 chance of winning CA. With Harris as Biden's running mate, well, I assume it's still 0/45. I really don't know about how to handle Trump's shift of home state to FL. I gave him a 1.010 modifier for FL, which increases his chance of winning FL from 42/45 to 43/45. Pence is still Trump's running mate, but taking the precedence of Cheney and Biden, VP's don't have the same effect on their home states the second time around as they do the first time, but Trump still probably has a 45/45 chance of winning IN. Result: Electoral Vote count: "R" about 293; "D" about 245.
Extrapolating, I calculate that Trump needs to improve his 2016 popular vote by a 1.045 modifier (from 45.93% to at least 48.00% of the national vote) for a likely win (about 278 EVs). This would assume that Biden gets 50.40% of the popular vote, and that the miscellaneous vote is still 1.6%.
If Trump does no better than he did in 2016 (45.93%), then Biden gets 52.47% of the popular vote, with the miscellaneous vote at 1.6%. Result: About 217 EVs for Trump.
I guess we'll see how everyone's predictions went in less than 36 hours.
Modifications to my model:
I'm going to keep the 1.059 multiplier over 2016 (modeled on the 2004 election), even though it doesn't seem to be the conventional wisdom ?:
538:https://projects.fivethirtyeight.com/2020-election-forecast/
Sabato 2020:https://centerforpolitics.org/crystalball/2020-president/
Cook 2020:https://cookpolitical.com/sites/default/files/2020-10/EC%20Ratings.102820.pdf
Realclearpolitics: https://www.realclearpolitics.com/epolls/2020/president/2020_elections_electoral_college_map_no_toss_ups.html
With the 3rd party/write-in/"none of the above"/etc. getting 1.60% of the vote, a 1.059 modifier shakes out to Trump getting 48.63%(vs. 45.93%) and Biden getting 49.77%.
On a state by state (or district by district) basis, the minimum deviation (i.e., the worst result for Trump out of 45) for each state of district would be adivisorof 1.049 vote for one state, a maximum multiplier of 1.211 for Trump's best performing state, and a median multiplier of 1.045 (i.e., 50-50). Example: In 2016, Trump got 48.18% of the vote in PA. If the national vote improves x 1.059, then there is a 1/45 chance that he only gets 45.93% of the vote in PA, a 1/45 chance that he gets as much as 58.35%, and the median gain would be 50.35%. In all cases, if the miscellaneous vote were larger than 1.6%, then Trumps chances would improve a little (since the percentage needed to win PA by a plurality would decline), and if it were smaller, then the percentage needed to win a plurality would increase, and Trump's chances would decline.
I'd like to integrate the idea of elasticity into my model:
https://fivethirtyeight.com/features/ele...onal-mood/
...but I haven't done it yet. Maybe another day. It's largely covered within the standard deviation anyway, with teh more elastic entities having the wider swings and the less elastic ones being closer to the median.
My guess is that the chance I give Trump of winning UT is a bit low (36/45 i.e., 80%). This is an artifact of McMullen's 2020 run, where he got 21.54% of the vote, and Trump got 45.54%. To some extent, my model takes that into account (in my model, a candidate can win UT with a 46.41% plurality), but I suspect it underestimates his chances. Oh well. Que Sera Sera for now. It's only six EVs anyway.
The modifications to the model: I assume (partly based on his performance in 2008 as a VP candidate) that Biden puts at least a 1.030 divisor to Trump's chance of reelection in DE, knocking Trump's chances of winning DE from 2/45 to 0/45. Without modifications, my model gives Trump a 0/45 chance of winning CA. With Harris as Biden's running mate, well, I assume it's still 0/45. I really don't know about how to handle Trump's shift of home state to FL. I gave him a 1.010 modifier for FL, which increases his chance of winning FL from 42/45 to 43/45. Pence is still Trump's running mate, but taking the precedence of Cheney and Biden, VP's don't have the same effect on their home states the second time around as they do the first time, but Trump still probably has a 45/45 chance of winning IN. Result: Electoral Vote count: "R" about 293; "D" about 245.
Extrapolating, I calculate that Trump needs to improve his 2016 popular vote by a 1.045 modifier (from 45.93% to at least 48.00% of the national vote) for a likely win (about 278 EVs). This would assume that Biden gets 50.40% of the popular vote, and that the miscellaneous vote is still 1.6%.
If Trump does no better than he did in 2016 (45.93%), then Biden gets 52.47% of the popular vote, with the miscellaneous vote at 1.6%. Result: About 217 EVs for Trump.
I guess we'll see how everyone's predictions went in less than 36 hours.