Why would you care what the published polls say? McCain has his internals and this is what matters.
- "Likely voter" methodology - use of "expansive" model vs. standard model (that accounts for past voting behavior)
- Overweight Democrat samples based on a common assumption about expected '08 turnout that is at odds with recent electoral history.
- Known higher propensity for Democrats to answer polls.
- "Bradley Effect" - anomalous answers due to perceived social pressure.
My point is that if we accept the idea that McCain has a chance to win, then we must also believe that something very unusual is happening this year, such that the validity of all current polling methodology needs to be called into question. On the face of it, that would seem to be a stretch. But this year, it might just be true.