I noticed that pollster.com are including the truly horrible Zogby Interactive shit polls (most of which greatly favor Obama compared to other polls) in their averages, and this has led me to consider possibly changing my methodology of just simply using the pollster.com average.
I’m generally of the attitude of one shouldn’t pick or choose polls, but these poll’s methodology is so horribly terrible (it says Obama is freakin’ ahead in Arizona for god’s sake) that I think they qualify for an exception to that rule. Yes, I’m against using those interactive polls that much.
If I changed my methodology, I’m not sure what I’d change it to – probably some sort of average where polls are weighted based on number of respondants and how old they are (though how I would do that, I’m not sure).
For the meantime, I’m updating the averages whenever another poll comes for a state, but I’m not going to update states based solely on the Zogby Interactive polls that have been released.
Update
OK, since getting home, I’ve played with a few equations and think I’ve found one that works reasonably well. It seems to be reasonable on the 3 or 4 states I’ve tested it on, but I’ll know more once I actually do it for all states.
Given that I don’t feel like trying to keep 50 spreadsheets to do this, I’ll probably write a program and set up a database to keep track of this myself, so I’ll actually need to write the code for it. Luckily it doesn’t look like it’ll be too hard, so it’d be nice if I could get it working in not too long. Once it’s done, I may run it for a week or two along with how I’m doing it now just to make sure I don’t suddenly see some freakish result, so if all goes well, I may launch my new methodology in somewhere like a week and a half to two weeks or so.
