The Washington Post's Election Lab paints a GOP takeover as a near certainty at 94 percent. LEO, the New York Times' model, pegs it at a 64 percent probability while FiveThirtyEight is even more guarded at 58 percent. (I talked to the three men behind our Election Lab model to understand what explained the differences. Read it here.)
As you might expect, the gaps in the certainty of the overall outcome are also reflected race to race. The two most prominent examples are in Colorado and Iowa. Let's look at each individually.
* Colorado: Election Lab (80 percent chance GOP win), FiveThirtyEight (57 percent GOP), LEO (57 percent GOP)
* Iowa: Election Lab (84 percent GOP win), FiveThirtyEight (62 percent GOP), LEO (57 percent GOP)
In both races, the Republican candidate -- Rep. Cory Gardner in Colorado, state Sen. Joni Ernst in Iowa -- does lead in composite polling averages albeit vary narrowly. Ernst holds a .8 percentage point edge over Rep. Bruce Braley while Gardner leads Sen. Mark Udall (D) by 1.3 points. Both parties are spending heavily in both states and the outcome still seems very much up in the air.
And, it's not just Republican opportunities where Election Lab is more certain of outcomes. In North Carolina, where all three models agree Sen. Kay Hagan (D) is a favorite, Election Lab is by far the most certain pegging a Democratic win at a 97 percent probability. LEO puts Hagan's chances at 80 percent, FiveThirtyEight at 79 percent. Hagan carries a 2.4 percentage point lead over state House Speaker Thom Tillis (R) in Real Clear Politics polling average in the race.
While the probabilities differ, the predictions from all three models of which side is favored to win each race really don't at this point in the cycle. Only in Kansas is there a discrepancy. Election Lab is confident of a victory by Sen. Pat Roberts (R), marking it at a 96 percent probability. Both FiveThirtyEight (59 percent) and LEO (68 percent) see independent businessman Greg Orman as the favorite.
If every race broke the way the majority of the models believe they will, Republicans today would win eight Democratic seats (Alaska, Arkansas, Colorado, Iowa, Louisiana, Montana, South Dakota and West Virginia) while losing one of their own (Kansas) for a net seven seat gain -- one more than they need to retake the majority.