There was a time when this survey/poll was good for observing trends; you just had to realize it was about 5-6 points too friendly to Trump. But lately it doesn't even seem to be good for looking at trends. It's basically remained steady since mid-September despite a clear shift in the race since then.
Well it's 3000 of the same people. It's pretty clear at this point (1) that they've all basically made their minds up, and (2) they screwed up when they picked their sample of people.