Democrats have regrettably been winning the narrative battle about the broader meaning of the current NY-20 tie, arguing that a district that with a significant Republican registration advantage and where Republican Jim Tedisco led by double digits only a few weeks ago should have meant the GOP was a lock for a pickup.
They should be reading Nate Silver on this count. He reminds us that NY-20 is a district Obama won by a slim 3 point margin, roughly four points less than his national victory. And it's a place where Republican Sandy Treadwell was not competitive at all against Kristen Gillibrand last November, despite spending $5 million of his own money -- another reason why we called out blind recruitment of self-funders by the Hill committees in the Rebuild plan.
At the end of the day, NY-20 is a Cook PVI R+3 seat -- Silver suggests it's R+2 in the current Congress. With the current Democratic lean in the House, this is essentially a swing seat, and quite possibly the definition of a pure tossup district in the 111th Congress. He posts this chart of how seats between PVI R+1 and R+4 voted in November:


Paradoxically, R+3 seats like NY-20 elected 8 Democrats and 5 Republicans last November. So much for being a Republican lock, especially with the blue undercurrents in the region.
What about the registration advantage? There are a number of regions throughout the country that are ancestrally Republican or Democratic that sport huge one-party registration leads and where the leading party has a lock on all the local offices. Many -- if not most -- areas in the Deep South still have many more Democrats on the rolls than Republicans. The most lopsided Bush/McCain margins in Florida came in the rural north Florida counties with the greatest Democratic registration advantage. On the flip side, Republicans appear much stronger places like upstate New York and rural New England than they actually are because these are traditionally Republican areas that have only recently started voting Democratic for federal offices with a shift driven largely by independents and moderate Republicans.
What these areas of Republican-in-Name-Only and Democrat-in-Name-Only strength all have in common is that they are generally rural areas with little turnover in population. If you're an older voter who has used been used to voting one way for 30 or 40 years, it's easier to rationalize your change in parties as "I didn't leave the party, it left me" rather than change your party registration outright. So there are a lot of Republicans on the rolls who may no longer vote that way. This may not be the case in fast growing metropolitan/exurban areas where political shifts are fueled by demographics and migration (e.g. Fairfax, Loudoun, and Prince William in Virginia).
As for the polls, Jim Tedisco began the race with vastly more name ID than Scott Murphy, who had the money to be competitive. And when a race gets nationalized with tons of earned media, with two relatively strong candidates running, it's hard to avoid a regression toward the mean, in this case, a close race to match NY-20's tossup status nationally.
Note: I've been doing some work for Tedisco.