Jay Cost has devised a new system for analyzing this year's midterm elections, in which he looks at the quality of opponents to incumbents, the popularity of the president, and voters' notions of the state of the country. Here's the conclusion, though it's worth reading the whole piece to understand his methodology:

If history is any guide, voters in at least two-dozen districts will agree that their local Democratic candidate is "part of the problem" and that the Republicans have fielded at least a slightly better alternative. But the Republicans need at least 40 districts to make a change. Will that happen? That remains to be seen, and that's not a trite equivocation. Congressional elections are a strange brew of national and local forces, which means that each is a unique world unto itself. The national forces have sorted themselves out pretty well, but strong Democratic performances on the local level could very well result in the party holding its House majority, albeit it by a slim margin. The best case scenario for Democrats at this point is a nominal majority where the median member is not a terribly reliable ally of the party's liberal leadership. Something similar is set to occur in the Senate, where a Republican gain of at least five seats will push the filibuster to more conservative ground, from Brown/Collins/Snowe to Alexander/Cochran/Murkowski. Barack Obama ran for and won the Presidency in 2008 based upon a pledge to pursue bipartisanship, and the results in 2010 are effectively going to force him to do just that, at long last.