Election Predictions is run by a high school student named Tom. He works very hard at this and has one of the best prognostication sites on the Web.
Tom is also predicting six Senate seats and is predicting a very conservative 26 House seats.
My feeling is that all these prognosticators are low-balling the House because there is no way to predict what the electorate is actually going to do in the relatively small House district races. The opinion polls are fewer, have worse internals, and a small change in turnout can have a disproportional effect, especially if the race is close, which many are.
If turnout is good, which it seems to be, when the night is done Dems could easily pick up upwards of fifty House seats. There are going to be some real surprises. Keep an eye on KS-02 D-Boyda vs R-Ryun for a bellweather.