Nancy Pelosi probably won't be House Speaker next year (J. Scott Applewhite -- AP)

Political scientists and statisticians have gotten pretty good at modeling presidential and Senate elections. Most presidential models — including the continuously updated ones of Drew Linzer, Nate Silver and Sam Wangshow Obama winning comfortably.

Linzer's prediction — 332 electoral votes for Obama to 206 for Romney — is especially notable given that the underlying model is Alan Abramowitz's "Time for Change" model, which explains an astonishing 97 percent of variation in results of past elections. Silver has also developed a model for the Senate, which has Democrats poised to take 52.4 seats, and a 87.6 percent chance of keeping control overall. Wang predicts 53 seats for Democrats, with the 95 percent confidence interval of 52 to 55 seats — so basically no chance of a Republican takeover.

But House modeling is trickier. As I reported a month ago, Wang's model put Democrats as overwhelming favorites to retake the body, whereas a model by Eric McGhee, John Sides and Ben Highton predicted a Democratic pick-up of only one seat. That's a much bigger gap than has been present between models' predictions for the presidential or Senate races.

Since then, Wang amended his model to predict a tied House, a big change from his initial prediction. McGhee, Sides and Highton updated theirs to account for ad spending in different races, which didn't change their prediction but did reduce the margin of error. Factoring in super PACs didn't change the picture for McGhee, Sides and Highton either.

Now Wang has updated his model again, this time looking at district-by-district polling as well as taking into account Republicans' consistent disadvantage in national House polls. He predicts that the chance of a Democratic takeover is 18-33 percent, and a Republican majority of 227 to 208, give or take four seats. That's a big Democratic gain relative to the current 240 to 190 majority, but not enough to flip control. And his chance of a Democratic takeover estimate is not too much greater than McGhee et. al.'s 10 percent estimate.

So the predictions are converging around the prediction that Democrats will probably make gains, but not retake the House. That comports with the predictions of non-modeling analysts like University of Virginia's Kyle Kondik.

So Democrats probably won't retake the House. That means that the likeliest scenario for the 113th Congress is a repeat of the 112th, with a reelected Obama having to cut deals with Speaker Boehner on the debt ceiling, spending, taxes, etc.