The higher the number of people polled, the more the percentages stabilize and stay the same. That's how polls work.
Internet polls with small numbers have the control problem of being limited to the regular readership for the polling site and who that readership notifies. But it's obvious that a lot more people were notified of this poll than any other Net poll I've seen.
It's an interesting one, for sure. I think the RNC poll dude should know about it. Most political people use the Net.
IMO, George Bush will win by a lot but only if we keep getting the vote out and vote ourselves.
I realize how polls work. One does expect "convergence" and as more people vote, each succeeding vote has less weight. Thus, as a poll gathers more voters, the percentages will typically change much more slowly, if at all.
However, the phonomenon here is quite different. The percentages seemed to be fixed while the polling numbers were quite low. Once 20,000 votes were cast, they didn't move AT ALL, even though we are well above 300,000 votes cast.
One would expect at least a little drift in the percentages. With a web poll the community that votes at work is purported to be different than the population which casts votes at home in the evening. Even the time shift should have provided a difference of some type.
This result is just weird.
Another conjecture: The poll is being driven by automated engines; i.e., the human participation is at such a low rate that it is essentially "noise". The Republican engine(s) are running slightly faster than the Democrat engine(s).
The poll numbers did seem to be going up incredibly fast. I should have taken time samples and determined the slope of the numbers. Of course, the numbers being posted could be constant because the poll engine itself is running at top rate due to normal voting. [This is one possibility which caused me to not perform any time samples.]