My 2022 Senate Model.

[ProgressiveModerate]

Making another model for this cycle.

Last cycle (2020), my senate model had an ok performance. The model correctly predicted partisanship would win out, and had races like AZ a lot closer than CO, and while early on it gave Democrats good chances in races like SC, KS, those races had moved to lean/likely R by election day. It also never bought into races like AL or KY being competative. The race where the model was the most off was Maine, giving Gideon about a 3/4 chance of flipping the seat in the end. Good polling for Gideon, partisanship, and fundraising had pushed the race into Gideon column, and while Collins did get a sizeable incumbency advantage, it wasn't enough to put her ahead in my model.

Lessons:

- Don't weight fundraising as high
- Incumbency can be a good or bad thing depending upon the circumstances
- Fundementals should be weighted alongside the polls instead of on top of the polls
- Calculate 2 party vote rather than margin
- Senate races have higher uncertainty than Presidential races
- Try to make and use as many objective metrics as possible
- Use T distributions for certain aspects of uncertainty rather than a standard distribution

2022 model changes thus far:

-Negative incumbency advantages, I think we see that sometimes being an incumbent isn't always a positive
- Calculation of average 2 party vote share, as well as the 90th percentile for either side
- New 2020 election data used for things like PVIs
- I'm better at math now than I was a few months ago, so better formulas for certain things
- New uncertainty clock

Current 2022 ratings:

Safe R:

North Dakota: >99.9% R
Idaho: >99.9% R
Oklahoma: >99.9% R
Utah: >99.9% R
South Dakota: >99.9%
Alabama: >99.9% R
Arkansas: 99.9% R
Kentucky: 99.8% R
Louisiana: 99.4% R (Still have to code the runoff, should bring Rs chances down a tad)
Kansas: 99.1% R
Iowa: 98.3% R (Chance will go down to ~94% if this ends up being an open seat)
South Carolina: 97.8% R
Indiana: 96.9% R
Missouri: 95.9% R
Ohio: 95.6% R

Safe D:

Hawaii: >99.9% D
Vermont: >99.9% D (Will go lower if it's open or if Scott runs)
Maryland: 99.7% D
California: 98.9% D
Connecticut: 97.7% D
Washington: 96.7% D
New York: 96.5% D (Right now, my model thinks Biden only won NY by 15%)

Likely R:

Alaska: 89.1% R
Florida: 85.8% R

Likely D:

Oregon: 94.3% D
Illinois: 87.7% D
Colorado: 84.3% D

Lean R:

North Carolina: 73.5% R
Georgia: 68.6% R (Will change depending upon who wins runoff and by how much)
Wisconsin: 66.9% R
Pennsylvania: 64.6% R
Arizona: 63.9% R (flip)
Nevada: 61.6% R (flip)

Lean D:

None

Tossup:

New Hampshire: 58.0% D


Remember, this far out, there is a lot of uncertainty, and overall, these are just initial probabilities; a 51% chance someone wins doesn't mean they are bound to win by any means.

I'll try to do a House model as well but it's way too early for that right now.

Any thoughts?

[Astatine]

Thanks for your contribution and the enormous effort you put into this!

Seems about right in general. From a "handicapper" perspective, NV and AZ stick out like a sore thumb considering the NH numbers, is there any way to "implement" the factor that Sununu might run (if it's possible for VT in case Scott runs)?

[ProgressiveModerate]

Thanks for your contribution and the enormous effort you put into this!

Seems about right in general. From a "handicapper" perspective, NV and AZ stick out like a sore thumb considering the NH numbers, is there any way to "implement" the factor that Sununu might run (if it's possible for VT in case Scott runs)?

Thanks you!

Unfortunately, until we know who the canidates are, there's no fair or objective way to resolve this problem; I can make a score for how often a state splits ballot, but that would jut increase uncertainty in places like VT and NH, it wouldn't actually flip NH to the GOP. Furthermore, the fundementals in NH work more in favor of Democrats than NV or AZ; Biden won NH by a lot more than NH or AZ, Jeanne Shahen easily won re-election, and Democrats relatively easily won both House seats, so assuming Kelly, Hassan, and Cortez-Masto all faced generic Rs, one at face value would assume Hassan would be in a better place than Kelly or Cortez-Masto.

These things are definately imperfect, and obviously this is nowhere near complete, though, once we know who the canidates actually are, my model will give a different output for many races than  it currently does.

[MATTROSE94]

Seems about right, though Maryland, Illinois, Vermont, and Connecticut are also at play if we take Rick Scott for his word.

[ProgressiveModerate]

Trying to make "potential" scores, basically based on how many statewide offices the opposing party hold, and how many congressmen there are for >5 terms. In open seats, both sides are given a potential score.



Basically my attempt to account for if Hogan/Scott/Sununu run. These scores will go away once canidates are selected.

Have done NH, MD, VT, CA, NY, FL and NC thus far since those are the cases where it's likely to have the biggest impact.

[Frenchrepublican]

Yeah you results seem plausible, if you were to run a generic republican vs a generic democrat in every
state it could look like this, obvioulsy NH will depend heavily of who will challenge Hassan.

[Amenhotep Bakari-Sellers]

Your map is wrong Cook already put out his rankings, you already was wrong about NC, ME, and FL in 2020 and we should still listen to your predictios, no


Wbrooks87, Landslide Lyndon and you said Collins and Tillis were DoA, wrong

[Roll Roons]

https://twitter.com/realDonaldTrump/status/1333556458575818754

Watch Ducey run for Senate and then get beat in the primary by Paul Gosar or Kelli Ward, ensuring Kelly's reelection.

[Amenhotep Bakari-Sellers]

No one likes Ducey, even Trump, he hasn't established a natl profile outside of AZ and he is dull and boring

[AGA]

Nevada isn't Lean R.

[Progressive Pessimist]

I guess a net loss of nothing is going to be the best case scenario for Democrats.

[Amenhotep Bakari-Sellers]

Nevada isn't Lean R.

Don't listen to Progressive Moderate he to us alonf wit wbrooks87 and Landslide Lyndon that ME and NC were slam dunks for D's now he is going the other way around and projecting a net loss of seats, tiyical

[Amenhotep Bakari-Sellers]

Here are my ratings

Narrow advantage to INCUMBENT party
AZ, NH and NV D, FL, IA, and NC R

Pure Tossups

WI and PA and GA pending Runoffs Tom Nelson with Evers winning reelection will blanch Johnson whom should of lost in 2016

Everything else is safe


It's gonna be a base Election, the D's can succeed in the Midterm due to all the battleground states are in the 306 freiwall, including the House

[Saruku]

Those probabilities added together (which does assume that there is zero correlation between contests, which obviously isn't true, but still) amount to a Republican gain of 1.4 seats.

[ProgressiveModerate]

Those probabilities added together (which does assume that there is zero correlation between contests, which obviously isn't true, but still) amount to a Republican gain of 1.4 seats.

When I correlate it the average gain is just under R+2.

[Amenhotep Bakari-Sellers]

Won't happen, D's will succeed in a midterm next