The problem with Rasmussen, which they are not acknowledging here, is that they apply a likely voter screen to every poll they do and "weight" outcomes according to their own, rather odd, assessment of the "correct" ratio of Democrats to Indies to Republicans.

A "likely voter screen" is not just asking people whether they intend to vote in the next election, although that's part of the suite. Rather, it is a suite of questions dictated by a regression analysis --certain answers to certain questions that are mathematically found, over time, to correlate strongly with a likelihood of voting in the next election. Party identification, education, income, employement, profession, issues deemed most important.

well designed likely voter screens do a great job of identifying people who are likely to vote three or four weeks before the election. And the closer it gets to election day, the better they do. However, likely voter screens are highly unreliable for identifying likely voters months out and, more importantly, they skew the poll result rightward.

Let me put it another way. The group of people whose answers to the likely voter suite cause the model to deem them "likely to vote" is very different from the group that gives the "right" answers a year before the election. And, for reasons too complex to go into here--but the short version is "because they are generally more intense and angry about damn near everything all the time"--people who give the answers necessary to be deemed "likely voters" are much more likely to be conservative and Republican the further away the next election is.

So here's how the Rasmussen scam works. By applying a likely voter screen to their raw data far in advance of the next election, Rasmussen generates results that are pleasing to their customer base (Republican politicians), and thus make it more likely they'll hire them for election polling, and yet perserve their reputation for "accuracy" because their results get more accurate the closer to election day the polls are taken. And, as an added bonus, they get to use their skewed, pre-election, results to try to distort public opinion in a way that conforms to the owners' own ideology.