Quote:

The Electoral Vote website actually updates itself constantly. The numbers I posted above included the following new polls in the following states, conducted today (Oct 5th): AL FL MI MN NH NJ NV OH OK PA



Updated today and "Conducted today" are two seperate things. You can hover your mouse over the states and look at what the most recent polling data is from.

For example, Alabama is listed as a "new poll" but it was conducted September 27th - 30th.

The newest polls are from October 1-3rd, which are as new as you're going to get. But many of those are not in the battlegrround states, which is all anyone really cares about.

Quote:

Not only does the site update daily with new polls, it also only shows polls from the previous 7 days. It is probably the most accurate polling site on the net currently.



I don't know where you read that, but it's certainly not true. It may be what they want, but some states just don't get polling data every few days. Currently the most current polls for some states are from September 8th, 9th, 10th, 13th, 17th, 20th, 24th, 25th...

In fact, out of all the states, less than 10 have poll results on electoral-college from the last 7 days. And most of those are just eeking by with data that's 6 days old as many polls were conducted just prior to the debates.

It is a very good site, there's no doubt. But it is by no means the be-all, end-all and isn't significantly more useful than the average national poll because of the various assumptions that are made in regards to their regression analysis.

EDIT: I see now where you're getting the "7 days" comment from. The webmaster is referring to the look-back, which while I understand where he's coming from in terms of volatility, is also the reason why his regression looks so very weird. A 7-day look-back compares other polls from the previous 7 days to determine what the outcome might be. It's much better than a 30-day look back in regards to volatility -- however we're still not at a point where we're getting daily-polls so the prediction is being based on only one poll because there's nothing within the past 7 days to compare it to.

What was happening is that, with a 30-day averaging you don't see a significant change in the nunmbers when something as unsettling as the last debate happens because the impact is spread over a large time. This isn't exactly realistic as an election can change in a second. And since the webmaster is a big Democrat this is what he would want to see of course. But with 7-days you need more data to be inputted to get an accurate picture of what's currently happening. Until that happens the data is going to be off quite a bit.

2003 Mazda6s 3.0L MTX Webpage 2004 Mazda3s 2.3L ATX