Notice the Bush trend starts later than the Kerry trend
Note to poinq the whole point of checking a trend is to watch where it is GOING, not hyperventalate about where it has been
Yes, but if you begin the Bush trend at the same time as sKerry's, the Bush trend line is still positive, though its slope decreases. That may be due to earlier voter disinterest, poll sampling numbers, and so forth. The president's trendline never slopes below sKerry's.
But if you were to take a trend from September 15th onward the trend would be very skerry.
True, but the whole point of the analysis is that September's numbers were so out of whack that the numbers likely were manipulated - even if only unconsciously - to engineer an sKerry October Momentum, which may be this year's October Surprise. The very short trendline you suggest discards important data and therefore adopts consistency (and so, credibility) issues.
I've been trying to find it again, but a paper published at a university somewhere in 1998 studied poll predictions for the 1996 elections. The polls in 1996 were so wrong in the same way, predicting sizable Democrat gains in both number of votes and seats in Congress, that the authors of the study concluded that some sort of prejudice against conservatives and Republicans had to be part of the problem. The conclusion included a comparison of probability: there was, if I recall correctly, greater chance of a coin being flipped 100 times successively "heads up" than to have all the polls agreeing so substantially and consistently, erroneously.
Your shortened trendline of data that is questionable for many reasons is likely erroneous as well.