Gallup Tracking : M 49, O 44

http://www.gallup.com/poll/election2008.aspx

Time to make another contribution.

I predict this is their high water mark in the polling before the election.

Next week: "Oh it's just 15 points".

M had a very large single day sample over the weekend. It will drop off the rolling average in a day or two.

or do you just accept the worst polls. You obviously don't understand that conventions produce bounces and it takes a few days for bounces to fade. (or the Hotline tracking poll showing it tied).

watch what happens tomorrow or wednesday, that'll be key.

this really isn't that bad.

and a good Obama day dropped off.

corrected with critique from nsd

the daily picture for saturday's release looked like this.

Wednesday: Obama +7.8
Thursday: Obama +2.4
Friday: McCain +4.2

To get the Sunday lead of about 3 points we saw, something like this had to happen:

3-(-2.4+4.2)/3 = x/3;

Therefore x would have to have been around 7.

So, assuming that's right (and rounding can throw us off quite a bit, as is the fact we are trying to solve N equations with N+2 unknowns (guesstimate, not solve, can't solve it). we have for today's release:

(4.2 + 7.2 + x)/3 = 5;

so x= 15-7.2-4.2 = -3.6

*However* there's so much rounding going on here, it is possible the lead may have flipped, although unlikely.

So the 2.4 in your Sunday equation should have been negative, giving a Saturday number of 7.2, not 11. With this number, yesterday's number would have been 3.6, not -0.2.

In any case, Silver's estimates have to be taken with a grain of salt, as he himself has so aptly described. Extracting individual day numbers from the three-day average is a hard problem.

the point remains that yesterday's lead is still down from the preceding.

And I'll correct it in my post.

.

This will be its high point for McCain then it will likely drop to a tie or Obama leading.

How are people getting the individual day results?
Everywhere I look I only see the 3 day average.

Anyone got any links. I am curious to see if it is 2 low days + 1 high day.
I would think 2 low + 1 high is better than 3 average days.
Of course 3 days with progressively lower results would be the best. :)

and pick the ones that make the most sense (say, if you minimize daily "swing" in results).

I say this will be their high water mark.

.

So people are trying to estimate them by analyzing how the three-day averages evolved over time and making assumptions about things like how much day-to-day variability in individual numbers is acceptable and the effect of round-off error. This is apparently harder than it sounds, as different models produce significantly different results.

a 6-point McCain convention bounce. Not much difference. No real worry yet unless it's this way in a week from now.

This is a good excuse to do it.

After all, for an election theft to be pulled off successfully, the cooked election results have to be plausibly close to polls, so the polls have to be cooked too.

Really. Does anyone really believe that a corrupt fundy loonie tune VP and a gramps candidate who's being chroma-keyed onto pornos after speaking in front of a green screen is really going to be that popular?

McBane had a similar, and brief, jump in May. It lasted all of a week, followed by some roller-coastering, and then MONTHS of O holding above.

I expect similar things this time, and won't make any big plans until the simmering is done.