10/31 Election update: taking out the trash

 EDIT: Heavily edited due to changes in polls over the course of the day

Okay, so I was strongly considering not making another update until Monday, but given Trafalgar made a new poll that downgraded Biden's averages in Arizona to the point Trump is a clear favorite to win the election there, I decided enough is enough. I've been talking about Trafalgar Group a lot recently. They're a polling company that comes out with these extreme outlier polls in Trump's favor and seems to believe there's a hidden Trump vote. Considering how their polls are clearly not made in good faith, with good data, but merely to offer reassurance to the MAGA crowd that their fuhrer will win reelection, I'm removing them. I'm also removing InsiderAdvantage and Susquahanna which seem to be associated with the "Center for American Greatness", meaning more extreme outlier polls skewing averages while likely just making the Trump vote look good. Don't get high on your own internal polling guys, you'll be disappointed come election day like Clinton in 2016 and Romney in 2012. Always with the losers believing their own hype. Not wanting to make that mistake, those troublesome pollsters will be removed from the averages and I will recalculate the averages from RCP's data without them. This should provide a more realistic picture of what happens. And before anyone asks, no I'm not removing democratic leaning polls. There aren't that many, and they aren't messing with the averages. I'm also not removing a lot of right leaning pollsters like Rasmussen. I'm not really interested in playing favorites with polls. I'm really only trying to remove a handful of pollsters with blatant biases that don't even try to produce good data and are screwing up my forecast. I'm assuming most other pollsters actually act in good faith. I may re-add these in future election cycle so this is only for the rest of 2020, I'm assuming after this cycle that these guys will see how delusional they are and change their methods. After all these guys weren't necessarily particularly wrong in past election cycles. 

That said, let's get to it.

2 way: Biden +7.8%

4 way: Biden +7.6%

Not much has changed. It is the weekend after all. And I don't expect opinions to shift much before election day. 

States	Winner/Margin	Z score	% D Win	% R Win	EV D win	EV R win
Michigan	Biden +8.8%	-2.200	98.6%	1.4%	232	322
Wisconsin	Biden +7.5%	-1.875	97.0%	3.0%	242	306
Minnesota	Biden +5.3%	-1.325	90.7%	9.3%	252	296
Pennsylvania	Biden +5.1%	-1.275	89.9%	10.1%	272	286
Nebraska CD2	Biden +5.0%	-1.250	89.4%	10.6%	273	266
Nevada	Biden +4.7%	-1.175	88.0%	12.0%	279	265
Maine CD2	Biden +2.3%	-0.575	71.7%	28.3%	280	259
Florida	Biden +2.2%	-0.550	70.8%	29.2%	309	258
North Carolina	Biden +2.1%	-0.525	70.0%	30.0%	324	229
Arizona	Biden +1.3%	-0.325	62.7%	37.3%	335	214
Ohio	Biden +0.8%	-0.200	57.9%	42.1%	353	203
Georgia	Biden +0.8%	-0.200	57.9%	42.1%	369	185
Iowa	Trump +0.6%	+0.150	44.0%	56.0%	175	169
Texas	Trump +2.3%	+0.575	28.3%	71.7%	413	163
Alaska	Trump +6.0%	+1.500	6.7%	93.3%	416	125
South Carolina	Trump +7.5%	+1.875	3.0%	97.0%	425	122
Missouri	Trump +7.5%	+1.875	3.0%	97.0%	435	113
Montana	Trump +8.0%	+2.000	2.3%	97.7%	438	103

Chance of a Biden win: 89.9%

Chance of a Trump win: 10.1%

Deciding state: Pennsylvania at +5.1% Biden

Most likely electoral college map: Biden 369-169

Yeah, another change. I'm reverting the color scheme. I only changed it because polls were bouncing around so much, but part of that was due to the Trafalgar effect, so now we have a better idea of what's going. 

But yeah, removing those bad polls shifted the map significantly. Michigan and Wisconsin are now solid blue states. Minnesota and Pennsylvania are much bluer, which raises Biden's chances to almost 90%, meaning there's only a 1 out of 10 chance for Trump to win according to the trend model. This is not 2016. This isn't even 2012. This is probably similar to 2008 when the last failing republican president was blown out and we got Obama. 

I've been skeptical of calling it like this for a while. After writing "Is Trump Screwed?" in 2016, with the obvious answer of "yes, yes he is" for most of the election cycle, only for Hillary's odds to buckle in the last week, and for Trump to win, I kind of was humbled after that. Even though I seemed to predict it more than anyone, it was shocking to see happen.

Between looking at these odds and the simulations yesterday based on the old data, I'm confident in a Biden win. I mean, even in that simulation model, even though it wasn't perfect, we saw a blatant weakness Trump could exploit in taking the rust belt, most notably Michigan and Pennsylvania. This isn't going to happen this time around. It's too statistically unlikely. While there is still a slight chance of a Trump win, he would need to overpoll significantly to pull it off. I don't think he has the energy this time. In 2016, he was up against Hillary. And everyone hated Hillary, and Trump got the last minute support and crushed at the ballot box, overperforming. However, as I saw looking into 2012 last night to run the simulation, the opposite effect happened with Obama there. Obama was only 2-3 points ahead mostly. I had a simulation that went Romney there. But because the trend model is ultimately more accurate at predicting elections for the most part, Obama ended up getting all of the states he was projected to and Florida. This time, if I had to guess which direction the energy and enthusiasm are running in, it's Biden's. Meaning I would actually expect better results than projected here. I don't think he will take Texas and get 413, but it's very possible. Clearing out the Trafalgar effect, I see Ohio and Arizona going blue. Georgia also may go blue although I know there's more voter suppression there and I think that with margins that close that Trump could win there. In a fair election, Biden for sure, but I'm not sure it will be particularly fair there. 

Senate Forecast

States	Winner/Margin	Z score	% D Win	% R Win	Seats D win	Seats R win
Michigan	Peters 8.3%	-2.075	98.1%	1.9%	46	55
Arizona	Kelly +6.0%	-1.500	93.4%	6.7%	47	54
Minnesota	Smith +5.0%	-1.250	89.4%	10.6%	48	53
Maine	Gideon +4.0%	-1.000	84.1%	15.9%	49	52
North Carolina	Cunningham +3.8%	-0.950	82.9%	17.1%	50	51
Georgia 1	Ossoff +0.7%	-0.175	56.9%	43.1%	51	50
Iowa	Tie +0.0%	+0.000	50.0%	50.0%	52	49
Montana	Daines +2.4%	+0.600	27.4%	72.6%	53	48
Georgia 2	GOP +2.4%*	+0.600	27.4%	72.6%	54	47
South Carolina	Graham +2.5%	+0.625	26.6%	73.4%	55	46
Kansas	Marshall +4.0%	+1.000	15.9%	84.1%	56	45
Texas	Cornyn +7.5%	+1.875	3.0%	97.0%	57	44

Chance of democratic control: 56.9%

Chance of republican control: 17.1%

Chance of a tie: 26.0%

Most likely outcome: 51-48 DEM

Okay, a lot to unpack here. First of all, Arizona is more solid blue now because of removing the Trafalgar effect. Second, North Carolina is more solid because of that. On the flip side, Iowa became less certain, because ironically, an Insider Advantage poll was +5 for Greenfield, so removing that brought their chances DOWN, which is the only case I've seen this happen in. It is what it is. On the flip side other changes are not due to recalculation. Georgia is really leaning democratic now due to NEW polls. Montana also became slightly more blue friendly. All in all, while democrats' chances went down, republicans' chances went WAY down, and there's a much higher chance of a tie. All in all though, the shift of Georgia shifting more blue does bump the most likely outcome up to a 52-48 split in the senate in favor of the democrats. 

House forecast (abridged)

I always wanted to do a house forecast but lacked the data. There are too many races to keep track of, and many don't have polls. However, I thought of a workaround. I only need FOUR democratic leaning seats to calculate the probability of republican/democratic control of the house, because of the 39 swing districts, the democrats only need to win 4 of them to win the house. So what can I find?

IA1: Dem 10.0%

IA2: Dem 9.0%

IA3: Dem 11.0%

NY22: Dem +9.0% 

Yeah, that didn't take long. 4 "swing" districts all above 8 points. Weakest one being 9 points. That's a 2.25 Z score, which means:

Chance of democratic control: 98.8%

Chance of republican control: 1.2%

Yeah, democrats are winning the house, it's pretty much statistically significant.