That shouldn't be hard at all.
I want to stress that this is a very simple calculation that simply averages polls, but weights them based on date. Then, of course, I ran different sets of numbers for different breaks on the undecideds. One very interesting thing I found:
I tried using a wide variety of coefficients for weighting the days. Almost all of them had the same exact result (there was one particular run that had Alaska switch to the Dems, but that was a virtual tossup, anyway, and you know the problems with the Alaska polls). That tells me that, at least through October, the polls have been pretty consistent, and most of the trends we are seeing in the polls are either A) our imagination or B) not yet robust enough to change any outcomes.
Once I had the newer polls, I'll copy to another spreadsheet, and do a separate run starting only with Oct. 20.
Thanks for the kind words.
EDIT: I haven't added any new polls, but I did run the same math starting on Oct. 20. the results are exactly the same for every state in terms of result, although the numbers shift here and there a bit. If the undecideds break Dem, they can hold the Senate at 50-50, but only if Orman causcuses with them AND they win the Georgia runoff.
An example of one of the shifts. By starting at Oct. 20th, Perdue's best case scenario goes from 49.68% to 49.97%. But, even if he tops 50%, at that point, the GOP has probably won enough other states to make it moot. On the other hand, the odds that Cassidy tops 50% actually goes down.
NB: I've recently noticed that my Louisiana data is entirely populated with head-to-head poll results. I'll need to input the three-way numbers and see how that changes things. I still think that ends up a safe GOP seat in the long run.
goat
This post was edited on 11/2 2:33 AM by TheOriginalHappyGoat