The most thorough poll average you can find

Freedom Window ^ | 10/12/24 | The Ref

[TheRef]

These poll averages look at every poll's sample and cross-tabs. Nothing goes into the average without determining exactly where bias exists. Bias is corrected before including.

*LATEST HEADLINES 24/7

Current Electoral Projection:

Trump 296

TOSSUP 16

Harris 226

ALL BATTLEGROUND STATE LINKS:

AZ - GA - NC - MI - PA - WI - NV

Click Here to See My Turnout Assumptions

Battleground States

POLITICAL REF'S AVERAGES (48/50)

POLLSTER AVERAGES

-

LIKELY DEM

LEAN DEM

TOSSUP

LEAN GOP

LIKELY GOP

ARIZONA

Trump +2.8

Trump +2.8

GEORGIA

Trump +1.5

Trump +1.6

MICHIGAN

Trump +1.7

Trump +1.6

NEVADA

TIED

Harris +0.3

N. CAR.

Trump +1.8

Trump +1.2

PENN.

Trump +1.1

Trump +0.7

WISCONSIN

Trump +0.4

Harris +0.2

POLLS ARE ONLY ONE FACTOR, additional factors include my assessment of turnout, 2020 results & exit polls, absence of polling, polling difficulty, primary voting, registrations, in-state operations, media influence, demographic makeup, future likely developments such as planned spending.

[null and void]

Will my fellow Nevadans please help get us get on the right side of a deadlock?

Nevada Ping!

To add your name to the growing Nevada ping list, FReepmail me...

[Political Junkie Too]

These numbers mostly align with my poll model, except that I age older polls to have less relevance the older they are.

Source	Arizona	Georgia	Michigan	Nevada	N. Carolina	Pennsylvania	Wisconsin
The Ref's Averages	Trump 49.5 (+2.8) - Harris 46.7	Trump 48.6 (+1.5) - Harris 47.1	Trump 48.9 (+1.7) - Harris 47.2	Trump 48.0 (0) - Harris 48.0	Trump 49.0 (+1.8) - Harris 47.2	Trump 48.4 (+1.1) - Harris 47.3	Trump 48.6 (+0.4) - Harris 48.2
PJ's Averages	Trump 49.3 (+2.3) - Harris 47.0	Trump 48.9 (+1.8) - Harris 47.1	Trump 47.9 (+1.7) - Harris 47.2	Trump 47.7 - Harris 48.2 (+0.5)	Trump 48.9 (+0.9) - Harris 48.0	Trump 48.4 (+0.3) - Harris 48.1	Trump 48.4 (+0.3) - Harris 48.1
PJ's Probabilities	Trump 74.7%	Trump 68.9%	Trump 59.3%	Harris 55.7%	Trump 62.3%	Trump 54.8%	Trump 55.4%

My Monte Carlo model produces this Electoral College result:

Trump 287 - Harris 251, with a 65% probability of Trump winning.

I expect Trump's ECV to grow as new polls keep coming in.

NOTE: The above result is using the New Jersey Polls from end of June that favored Trump. If I give NJ to Harris, the ECV result changes to this:

Trump 277 - Harris 261, with a 60% probability of Trump winning.

If I give NJ to Harris but assume a 1% bias against Trump (adjust Trump +1 and Harris -1), the ECV result changes to this:

Trump 297 - Harris 241, with an 81% probability of Trump winning.

If I give NJ to Harris but assume a 2% bias against Trump (adjust Trump +2 and Harris -2), the ECV result changes to this:

Trump 313 - Harris 225, with an 93% probability of Trump winning.

-PJ

[EasySt]

Sorry, but I can’t / won’t respond to polls.

It feels much too much like divulging troop deployments to the enemy.

[They] will know how many of us have awakened to all [their] gaslighting,
shortly after the only poll that truly matters.

[TheRef]

Yes, I agree that older polls should matter less. I actually take them out of my average altogether. My approach is use the data the pollsters give you in the cross-tabs, shape that data with the right voter sample assumptions, then you’ve got something. The problem is dishonest pollsters like NYT/Sienna coming out with that nonsense PA poll today, and the Arizona one for that matter. They actually have Harris trailing by only five with men in PA, and by on three with white ID in PA. Give me a break! Can’t do much with polls like that. All you can do is adjust using party ID in a case like that because the bad party ID impacts all other categories. I like your model. do you put it on a website or just forums like this?

[SharpRightTurn]

Nice work, TheRef.

[SharpRightTurn]

Interesting observations, PJ2.

[Political Junkie Too]

statespoll.com used to do what you're doing. They popped up during the 2016 election and reweighted polling data with their own turnout history analysis. Unfortunately, it looks like they got beaten into submission during the 2022 mid-term cycle and gave up. Their history is still there, so you can look at what they did for inspiration.

My model is a simple (for me) Excel spreadsheet. I developed a formula in 2010 that converts a poll and margin of error into a probability of the leader actually winning. Nate Silver figured out the same formula in 2012 and built his fivethirtyeight website around it.

I collect up to 15 distinct polls per state, convert each to a probability, date-weight the probabilities into an overall state average, and then run a Monte Carlo simulation of 5,000 iterations to generate an outcome probability distribution.

New polls replace older polls by the same pollster. At this point in the election, I no longer use Registered Voter polls. I used to put out a weekly report here on Free Republic showing the changes in polls for the week, but I stopped putting that much effort into it due to the impacts of systemic voter fraud and push-poll bias. Now I just post snippets like today when an appropriate thread pops up.

I also do a Senate poll. As of now, Republicans have an 82.1% probability of winning the Senate, with an expected value of 51.79 seats.

Regarding adjusting for bias, I leave that to the "experts." I figure that extreme polling error will get washed out in the averages, especially if an outlier poll becomes stale against newer polls. I chose to make my report be on the polls as they are, since that tells a story, too. Hopefully, it ends up being the right one.

-PJ

[TheRef]

I’ll keep an eye out for your posts. How do you adjust for toplines that are impacted by biased sampling? It could come in the margin of error calculation you do perhaps. My problem with Nate Silver is the garbage in garbage out problem. If you input the toplines of 80% of polls that are oversampling Dem leaners just a little bit in every ID, like slightly too many women, slightly too many non-white ID, slightly too little white ID, slightly too many college degrees/post-grads, etc. and they don’t balance those out with comparable oversamples of GOP-leaning bias in other categories, won’t your model end up leaning too Dem? Maybe you take a cross section of polls where you have a even number of GOP-leaning and Dem-leaning polls.

[Political Junkie Too]

Yes, you are correct that my model will be imbalanced, too. As I said, I'm not trying to correct the balance, I'm telling what the polls are saying but hopefully better than just the RCP average. I add the probability of winning that the MOE brings to the polls. Just because the poll is "within the margin of error" doesn't mean it's a "statistical tie" as the LAAP-dog media likes to say.

My hope was that, over the entire breadth of the polls it would balance out. In 2020, Emerson's polls were the most balanced while Quinnipiac's were the most tilted to Democrats. If I see an egregious error in the sample, I will exclude that poll from the average, but that's as far as I go to try to "correct" the polls.

I had hoped that peer pressure and social stigma from wildly incorrect polls would self-police the polling firms, based on their wanting to remain credible from a business model perspective.

-PJ