Two years ago a very bright guy named Nate Silver pitched an Electoral College shutout by accurately predicting the results in all 50 states in the election between Barack Obama and Mitt Romney on his FiveThirtyEight.com blog.
Since then, a lot of other organizations have been applying statistics, local knowledge, history, and not a little salagadoola menchicka boola to tease out what the odds are of a GOP take over.
None of the players in this game have come up with a formula that has the Democrats keeping control. The Washington Post's Chris Cillizza summarized the predictions earlier this week.
When Cillizza wrote his piece, his paper's Election Lab put the odds at 94 percent. The aforementioned Nate Silver had the GOP winning control but gave it a 58 percent shot. The New York Times (for whom Silver used to write before going off to ESPN) split the difference with a 64 percent probability.
As of last night, the Post's model slipped one percentage point to 93 percent; the Times' model has ticked up to 73 percent, and Nate Silver has narrowed his odds to 60 percent for the GOP.
Polling, as we said above, is a great deal of science, with a healthy serving of art thrown in. As you know, I am not a pollster. I thought a regression analysis was something Freud cooked up.
In spite of this competition, pollsters have said for decades that a poll is not a predictor of what will happen on election day. It is a snapshot of what the people who participated in the poll said at the time they were contacted.
Because each polling firm has its own way of choosing who to contact, what the specific wording of the questions are, and in what order they are asked (among other things) it is very risky to compare a poll about the Senate race in, say Iowa taken by the Des Moines Register with a poll taken by the Associated Press.
They probably won't agree because of the differences in their methodologies, but it is too easy to claim "TREND!" when a neck-and-neck race like that either gets closer or one candidate grabs a lead.
But, almost everyone does it because we can't help ourselves and because, unlike national polling during a Presidential campaign, there aren't enough polls taken by the same outfit to compare apples with apples in a given State.
Having said all that, time is growing short for Democrats to make a late move. The Democratic Senatorial Committee cut funding to the race in Kentucky, admitting that Alison Lundergan Grimes will not defeat the Republican incumbent, Mitch McConnell.
On the other hand, the race in South Dakota - which has been off the table for months - suddenly has generated interest as the polls have tightened in the three-way race there. The Democrats have ordered up about a million dollars in ads to try and knock the leader, Gov. Mike Rounds, off the top of the leaderboard.
I have been on both sides of these elections and, as a general rule, the party that is behind tends to lose all the close elections, and some elections that you thought were safe become nail-biters on election night.
But...there are always election night surprises and which way those surprises break will determine which party controls the Senate.
So, like the statisticians toiling at the Post, the Times, and FiveThirtyEight, your best bet is to put 'em together and what have you got?
On the Secret Decoder Ring page today: Really useful links to the web pages of each of the three prediction units that you can follow day-by-day. Also a mind-numbing license plate Mullfoto.