On the House, Stuart Rothenberg says:
the most likely outcome in the House of Representatives is a Democratic gain of 34 to 40 seats, with slightly larger gains not impossible.
Back when I first started looking at methods of estimating the connection between generic polling, the actual aggregate votes for real candidates of the major parties, and the translation of those votes into seats, I suggested Democrats could wind up with as many as 245 seats (at what I considered the high end), which would be a net gain of 42.* At the time, no one else was projecting even close to that much. If Rothenberg’s district-by-district analysis turns out to be correct, was my aggregate method of estimation prescient or lucky? I don’t know. We’ll try it again in 2008. And 2010…
As for the Senate, Rothernberg says control is in doubt, but:
we do not think the two sides have an equal chance of winning a majority in the Senate. Instead, we believe that state and national dynamics favor Democrats netting six seats and winning control of the United States Senate.
* That was when the generic lead was 15 points; the latest estimate from Charles Franklin’s compilation of polls is 17 points.