It is nice to point out that pollsters get graded by their ability to accurately make predictions with their LAST poll. From the moment that they start polling (2 years ago) until this point, they can pretty much get away with predicting anything. They can be chastised for being an outlier, but generally the numbers that they publish can not be proven to be inaccurate.
The pollsters themselves are most interested in themselves of course. The election is the only time where they themselves have incentive to be as accurate as possible.
Nobody is going to compare the election results to Rasmussen's or PPP's or Gallup's numbers published in July.
An unscrupulous pollster could conceivably make up numbers all season long, with the intention of driving the narrative. However, the last set of numbers will be used to measure their own accuracy so all pollsters have an interest in being accurate at this point. It will be interesting to notice whether any polling outfits will show dramatic shifts near the end of the campaign.
I was just reminding my friends that the incentive for the pollsters changes in an interesting way at this point in the campaign.