re: Landrieu outperformed the rcp poling averages by 8 points

An election like this is very difficult to predict with precision. The big problem is turnout. In a general election we have a lot of previous experience that can be used to predict who the likely voters are and how the likely voters are likely to vote. In a runoff election like this, it is more difficult to predict who is going to vote. I suspect that the polls showing Cassidy up 24 were based on likely-voter models that were flawed. They were based on a general election instead of a unique situation like a runoff, and hence they overpredicted turnout among Cassidy voters.

One thing that strikes me about this result is that ML did not improve appreciably on her vote percentage in the first election. In a multi-candidate field she got 42% of the vote, and in the runoff she got 43% of the vote. In order to win she needed to get a higher turnout rate from her supporters than from the supporters of the other candidates, and this did not happen. I don't remember seeing a poll throughout the campaign in which she broke 45% of the vote. Barring a major mistake by Cassidy, she has been a goner for some time.