The trend lines in both the LEO and FiveThirtyEight models highlight movement over the past week toward Republicans -- with a slight tick back toward Democrats in the last few days.
Here's the LEO trend:
And here's FiveThirtyEight:
There are only four Senate races where all three models agree there remains real doubt about the outcome. (Worth noting: The Election Lab model, as I have noted before, is far more certain in all of its probabilities -- Democratic and Republican -- than either of the other two.) They are:
1. Kansas: Election Lab (95 percent GOP win), FiveThirtyEight (56 percent independent win), LEO (59 percent independent win)
2. Iowa: Election Lab (87 percent GOP win), FiveThirtyEight (65 percent GOP win), LEO (67 percent GOP win)
3. Georgia: Election Lab (63 percent GOP win), FiveThirtyEight (65 percent GOP win), LEO (63 percent GOP win)
4. Colorado: Election Lab (83 percent GOP win), FiveThirtyEight (69 percent GOP win), LEO (68 percent GOP win)
The most interesting of that quartet is Georgia where Democrat Michelle Nunn's chances against businessman David Perdue (R) have improved markedly over recent weeks -- thanks in large part to revelations about Perdue's comments on outsourcing and polling that shows the open seat race very, very close. The Democratic Senatorial Campaign Committee diverted several million dollars from Kentucky to Georgia last week, another sign of Nunn's increased competitiveness. The problem for Nunn (and Democrats more broadly) is that if she can't get to 50 percent on Nov. 4 -- which looks unlikely given the presence of a third party candidate -- there will be a runoff on Jan. 6. That runoff electorate will likely favor Perdue.
Iowa and Colorado continue to edge closer and closer to the Republican column. Last Monday both FiveThirtyEight and LEO gave Rep. Cory Gardner (R) a 57 percent chance of knocking off Sen. Mark Udall (D). Now FiveThirtyEight puts Gardner at a 69 percent probability of winning while LEO has him at 68 percent. Election Lab had Gardner at 80 percent last week and 83 percent today. In Iowa, LEO now puts Republican Joni Ernst's probability of winning at 67 percent, up from 57 percent a week ago. She has become slightly more likely to win in the FiveThirtyEight model (62 percent last week, 65 percent today) and Election Lab (84 percent to 87 percent).
Kansas remains the lone anomaly among the models -- as LEO and FiveThirtyEight see independent candidate Greg Orman as a slight favorite while Election Lab feels confident the seat will remain in Republican hands. (A note on that: Election Lab's prediction is based on Orman's stated intent to caucus with whichever party hold the majority. Given that Republicans are heavily favored to do so in the EL model, the probability of Kansas staying Republican is based on both a Roberts or Orman win.)
As of today, all three models give Republicans a better than 75 percent chance of winning six Democratic seats: Alaska, Arkansas, Louisiana, Montana, South Dakota and West Virginia. Win only those -- and hold Georgia -- and Republicans have the majority even without Iowa or Colorado.