It's way too soon to start obsessing on this stuff. But I was impressed with a careful (to the degree that any attempt to forecast election results a year and a half in advance can be careful) two-part look by Sean Trende of Real Clear Politics at the shape of the race for control of the U.S. House in 2014.
If you read the whole thing (part 1 here, part 2 here), you'll learn quite a bit about midterm election history and see quite a bit of debunking, or at least de-hyping, of the "sixth-year-itch" narrative, which suggests that the president's party usually gets clobbered in the midterm election of a president's second term. Trende shows that this usually doesn't happen if the president's party already got clobbered in the midterm of the president's first term, as happened to Obama and the Dems.
Trende also finds that the vast majority of House districts are already aligned between the party of the current representative, the party that carried the district in the 2012 presidential election and the party's partisan lean as evaluated by those who obsessively evaluate districts for that purpose. Or, to let Trende explain that point:
There are only nine Democrats in districts Mitt Romney carried, and only 17 Republicans in districts Barack Obama carried. If we look at things in terms of Partisan Voting Index (how a party performs relative to national forces), there are only eight Republicans in districts with Democratic PVIs, and 15 Democrats in districts with Republican PVIs.
Obama's approval ratings have recently hovered around 50 percent. The economy continues its agonizingly slow (but impressively steady) recovery. If either of those factors changed significantly, the outlook would also change in obvious ways. As of today, pending the outcome of the imminent special election in South Carolina, Dems need a net pickup of 17 seats to take over the House. Trende's bottom line:
Somewhere between a five-seat Democratic pickup and a 15 seat Republican gain seems a safe prediction for now.
No way do I endorse this as a prediction, but I felt smarter after reading Trende debunk and historically contextualize a great many of the rules of thumb that one hears about midterms.