Sanford Gordon of NYU has an interesting paper that forecasts the House results based on district-by-district ratings by Charlie Cook and Stu Rothenberg. The forecasters categorize every House seat as "solid" for one party, "likely," "leaning," and "toss-up." Gordon compares those forecasts with recent elections to translate each category into a percentage chance -- i.e., "lean" means a 93% chance of victory -- then plugs in the current district-by-district ratings. He comes up with a somewhat surprising result: Republicans are expected to gain just 28 to 25 House seats, not even enough to control the majority, and considerably below what most of us expect.
Now, Gordon does continue with a very strong caveat: it is possible that the forecasters deliberately make conservative calls. They underestimated the scale of the recent Democratic waves, and it's possible the same sense of caution is causing them to underestimate the coming Republican wave. And the curious thing, pointed out by Gordon, is that both Cook and Rothenberg predict higher aggregate gains for the GOP than their district-by-district ratings would suggest. So either their district ratings are hedged, or their aggregate prediction is hedged.
But the interesting takeaway is that if the Cook and Rothenberg district-by-district ratings are correct, Republicans will probably fall short of winning the House. Again, it's not clear if those ratings are themselves hedged.