So, I delay it a day and then only one poll comes out on Saturday, figures. The last two saturdays had TONS of polls come out, but i switch it up to account for that and I shift Texas by 0.2. Bleh. Anyway. Here's this week's update.
Senate
All in all, some mild change, mostly in favor of the democrats. Nebraska is dead heat with no clear leader. Kansas, FREAKING KANSAS is only R+0.2. Iowa remains a tossup, and Maine remains one.
Reflecting more on Maine, and having watched part of the debate, I think I figured out what's going on there. Susan Collins is a normie moderate republican who can successfully distance herself from the Trump administration, and Troy Jackson has the charisma of a wet blanket. The debate was SO BORING. I tried to watch it but it just couldnt keep my attention. Both these guys are like watching grass grow. For Susan Collins it feels like a welcome change of pace given the flair and pizazz of the Trump administration, but with Jackson it just...doesn't work.
Anyway, those are the tossups. Alaska is lean D with a relatively weak margin. Minnesota is...idk wtf is going on there, should be D+6, not D+2. Texas, Ohio, and Michigan are all D+3ish. And yeah, the senate is looking pretty good for democrats. On the republican side South Carolina is R+2.4. That's kinda crazy. Florida is D+7, and then Montana is D+14, with Bodnar being held back by Bankhead refusing to drop out and allow the candidate with a chance a chance.
House
So I added Alaska's district to the list of districts I auto flip. Unlike the others, it's not gerrymandering that does it, it's polling. The fact is, according to my model, the outcome is supposed to be D+9, so basically soft safe D. According to polling, it's supposed to be R+9 so soft safe R. Given my model and polling are 100% at odds with each other, I flipped this district, recognizing the flaws in my own model. Honestly, there are other districts where my model and polling disagree, but not like that. Most other disagreements are within the margin of error, or have polling within the margin of error. Like WA-03, my model predicts D+14, whereas the polling predicts R+1. A massive difference, sure, but R+1 still means 40% chance of democratic win, so I figured to just let that one ride. Another one is VA-01. My model predicts R+2, polling predicts D+4. These two numbers are within the margin of error of each other so I won't auto flip them.
There isn't really enough polling to really justify making a full polling only model. Like seriously, we only have like 5 districts of 435 polled right now. And only around 75 or so are in play so to speak. So yeah, do the math, not worth my time. Polling that goes against the cook ratings also often ends up being wrong. Again, the Alaska polling is the exception. Not only are the differences massive, but cook still has it as likely R, the polling tells me likely-safe R depending on how lenient you are with ratings (5 point MOE and it's "likely" instead of safe). And only my model has it has like, safe D. So my model is very obviously wrong there.
Anyway, this leads me to giving a R+3 correction to my model, where dems have 234 seats and republicans have 201, with the democrats still having a 99% chance of taking the house. Idk how reliable that is given my other system, but to be fair I dont trust that other system either. I'll go with this one for now as I seem apt at detecting errors and flipping outcomes that are out of line.
Governors
Conclusion
Anyway, that's this week's update. The senate is largely holding steady, we lost a tiny amount of ground in the house that is of little statistical significance, things are gonna fluctuate I guess, and the governors are where they are. I'll update weekly through election day, I'll probably make a couple predictions before election day itself like on Monday with me constantly editing it, and you'll have the final one by Tuesday afternoon before polls closed.
Btw, I just voted today so....yeah. This crap is already going on. Love mail in ballots.
No comments:
Post a Comment