Shorter Nate Sliver: Between 20 to 80 seats!

A gain of as large as 70-80 seats is not completely out of the question if everything broke right for Republicans. Conversely, if Democrats managed to see a material rebound in their national standing over the final two weeks of the campaign, they could lose as few as 20-30 seats, as relatively few individual districts are certain pickups for Republicans.

haha, that's one way to cover your bases.

The problem with predicting House races is that there are too many to base it on polling because not enough polls are done. I'll predict right now that there will be upsets that very few saw coming, simply due to the lack of any public polling being done.

There are going to be at least 100 races that Democrats either lose, or come within single-digits of losing. In both 2006 and 2008, I went with predicting what the wave looked like, and it turned out that way. If you go with what the wave looks like now, it winds up in Jim Geraghty's lap.

If you look at those Gallup numbers though, in this redistricting decade, you'll see that it does tend to tighten as the election nears during the month of October.

Democrats had a 23 point lead that closed to 7 points in 2006.

Democrats had a 1-3 point lead that closed to a Republican 6 point lead in 2002.

After that 2002 election, the makeup of the House of Representatives was 229 Republican seats and 206 Democratic voting seats.

So, if the Democrats are able to close the current 17 percent difference by about 12 percent, making it 5 percent, a good marker would point toward a swing from the current 255 seats held by Democrats to the 206 seats held after 2002. Which happens to be, at 49, right near the current consensus.

That's not too surprising either-- to see prognosticators gravitate toward the to-date extreme for the current CD makeup to predict the mid-term results when it looks like a blowout is pending (it did get a bit worse for the Democrats after the 2004 GE when the makeup was 232 Republicans to 203 Democratic voting).

A 49 - 52 seat loss (bringing the Dems back to 203 - 206 seats), seems to be as bad as conventional wisdom could imagine.

A 17 point generic Republican lead however, like the current Gallup shows, means that any seat where there's a less than a less than a D+10 makeup is more than on the table. ~70 is just grabbing a higher number. There is no sure way to know. It should close from here I'd imagine, but we'll see.

U.S. voters' preferences for the party they will support in this year's House elections have been quite stable over the past three weeks, with Republicans leading by low single digits among registered voters. They lead by substantially more than that among likely voters, including both high-turnout and typical-turnout scenarios.

It is not clear from the historical record how likely these patterns are to continue through the end of the month. Gallup's pre-election polling in prior midterm years documents that there were some years when the structure of voter preferences by this point in October was generally maintained through the elections (1998 and, to a lesser extent, 1994), but others when it changed substantially (2002 and 2006).