Just as a heads up, we can roughly work backwards from the Iowa caucus and the exit polls (which are much more accurate than prior polls because they are definitely selecting the right sample and only have MoE issues to worry about) to see who was most accurate. We know from the exit poll people who decided in the last day split 46/46 between Clinton and Sanders, so they had no impact in changing the Clinton/Sanders balance. About 10% decided in the final few days of the poll, so they should have shown as undecided to pollsters. Martin O'Malley added about 1.8% to Sander's score and 1.3% to Clinton's after redistribution. Therefore, a perfectly accurate pollster would have shown on the 31st January Clinton 45, Sanders 43, O'Malley 3, Undecided 10.
This made Selzer, quite significantly, the most accurate pollster (within the MoE on all accounts), then NBC/WSJ/Marist in close pursuit. In the middle ground was Quinnipiac, then Monmouth University, then ARG. PPP (D) and CNN/WMUR performed quite poorly, and finally Gravis and Emerson were absolutely terrible and should really just be ignored from now on.
So, no, Quinnipiac is not a terrible pollster. It has a good track record both from prior elections and performed reasonably well in Iowa. It's not perfect - it seems to have a structural bias in favour of Sanders by ~ +4 points in Iowa, although we can't guarantee it will persist at 4 points in demographically different places that require alternative weighting - but it's above the average pollster so far this cycle, certainly. This one is probably an outlier, yes, but it's an outlier in the literal sense of the 5% of the time you get a sample outside the MoE, not an outlier as in "Quinnipiac sucks and can be ignored".
For what it is worth, PPP (D) had about a +5 structural advantage in Clinton's favour, so if you want to do what 538 did in 2012 and actually correct for pollster biases rather than just weighting them differently but adding them up without accounting for anything else like they're doing now, then Quinnipiac's 44-42 is roughly equivalent to 48-38, and PPP (D)'s is roughly equivalent to 48-37, so... rather close and where I imagine the race is at right now. In other words, when you adjust for how they performed in the past, they actually show very similar results. That's reasonable movement for Sanders, and I think more will come as the news percolates, but not really sufficient to be a real challenger; unless New Hampshire shakes up the news cycle further.
Moving on from Quinnipiac to New Hampshire, then: Sanders needs to win at least 54.1% of the vote. New Hampshire awards 12 delegates to each of the two congressional districts, split proportionally according to share of the vote in each. If Sanders won 54-46 exactly in both districts, he and Clinton would still get 12 delegates each because 0.54*12=6.4 which would round down (and conversely round up for Clinton) to 6 each. If he got below 54.1% in both, then it's over right now, we all go home.
This is unlikely, though, and I think a better test is if he can get 62.6% or above in at least one district. If he can take 62.6% to Clinton 37.4% in one of the districts, then he gets 15 delegates to Clinton's 9, and it would generate other big news burst that can hopefully push him further on again; it would be the sign that his momentum is still rolling. If he takes 62.6% to 37.4% in both, then that's big news and indicates he will be approaching competitiveness nationally. If he doesn't breach this in either district, then it'll be a relatively limp 14-10 win (or even 13-11 if he falls under 54.1% in one of them).
tl:dr 16-8 great, 15-9 great, 14-10 okay, 13-11 bad, 12-12 it's all over, 11-13 adam probably explodes.
