We've all seen the Hays Research poll showing support for Miller crashing, and McAdams poised to win if Murkowski can't get all those people to write in her name, but I would like to offer some data that is probably outside most people's paradigm, new polling data from Alaska that is probably more realistic in the sense of where McAdams and Miller are relative to one another (although the Murkowski numbers are probably garbage).
Here is the link to the PDF, called the "ALASKA PUBLIC OPINION RESEARCH SURVEY"
Bottom line: 45.7% Murkowski (they said her name during the interviews, so its probably a garbage result), Miller 30.4%, and McAdams at 23.9%.
More analysis below
The breakdown of party affiliation in the poll, FWIW, is: None, 46.2%, GoP 37.9%, and Dems at 15.3%. I have no idea how realistic those numbers are for Alaska. Also, less than 20% of the respondents were under 45 years old, meaning that it was a pretty senior sample.
Nate Silver has an updated page for the Alaska Senate election:
Nate doesn't use the Hayes Research poll, so all of his source data shows McAdams in the 20s. The trends shown on the page show a small drop by Miller and a small gain by Murkowski. There aren't that many polls on this race, so the data is sparse. Error bands on Nate's page are pretty high.
So ......... I am not saying that any of these polls are accurate. My only point here is that you shouldn't look at one poll - in this case, the Hayes poll - and ignore the other data. If you don't take into account all of the data, your view of the election will necessarily be biased. Having said that, the unknown factor of the write-in candidacy makes 2nd place in the polls a pretty good place to be, and McAdams is not far from there.