Producing content for the Internet requires that one doesn't bury his lead, so I'm going to show you this chart first, but please interpret it cautiously and pay heed to all the caveats and disclaimers below. The numbers you see are not formed by any statistical program; they are informed, but fuzzy, guesstimates.
I periodically get inquiries about why, considering that FiveThirtyEight is a numbers-driven website, I've tended to take a more qualitative take on the 2010 Senate picture. The reason is that -- as the completely unexpected retirement of Byron Dorgan and the somewhat unexpected retirement of Chris Dodd amply demonstrated -- there is still substantial uncertainty about the identities of the candidates. I do try to factor these things into my estimates -- I don't see how you can't -- but it obviously relies upon a fair amount of guesswork. Once the field starts to settle down a bit and some of the races that haven't gotten adequate polling -- like Colorado or Iowa -- begin to get a bit more of it, we will transition into a more systematic approach. That will probably occur sometime in the mid-late Spring. For the time being, however, a good subjective estimate is probably going to be more informative than an "objective" answer that relies on all sorts of potentially false assumptions and treats the ridiculously early polling as the gospel truth.
That doesn't mean, however, that I don't think about the races quantitatively. In fact, I generally have a pretty good working idea of the probability I'd assign to a given seat switching hands. I'm reluctant to publish these as a matter of course, since when you begin to put specific numbers out there, people may assume that you're more certain about a race than you really are. Nevertheless, in light of what I think is somewhat inaccurate conventional wisdom that has developed around the 2010 Senate picture, it is probably worth "showing my work" on occasion. What you see above, then, are my very rough and premature estimates of probabilities of the various competitive and potentially competitive Senate seats switching hands.
If I aggregate my estimates from the individual races, I show Republicans picking up an average of 4.60 Democratic seats, but also, Democrats picking up an average of 2.65 Republican seats, for a net Republican gain of 1.95 seats.
But this is very important: the average is somewhat noninformative here, as these races do not operate independently from one another. It is somewhat unlikely -- though certainly not impossible -- that Republicans will pick up 4-5 Democratic seats and Democrats will pick up 2-3 Republican seats. If the national environment continues to improve for the Republicans, for example, perhaps they'll pick up six or all seven of the seats that are basically toss-ups or better (everything from Illinois upward), and perhaps put another race like California or Wisconsin into play, while defending one or all but one of their own seats.
On the other hand, if conditions improve for the Democrats, perhaps they can hold their losses to 2-3 seats (say North Dakota plus one or two from the group DE/NV/CO/AR) while picking up Missouri, perhaps two from the OH/KY/NH group, and one from the NC/FL/LA group. In that case, Democrats could hold at 60 seats or even improve their numbers to 61-62.
There are an unusually large number of Senate races in play this year and as such there is an unusually large amount of uncertainty surrounding the outcome. It also bears remembering that, although I remain quite pessimistic about what will happen to Democrats in the House, the Senate playing field is intriniscally more favorable to them. The Senators who are up for re-election this year are those who were elected in 2004 -- a good cycle for Republicans. And while Democrats were hurt by their retirements in North Dakota, Delaware, New York, Illinois and probably Colorado (they were helped by Chris Dodd's retirement in Connecticut), the Republicans have created opportunities for them with the retirements in Missouri, Ohio, New Hampshire and perhaps Florida (they were helped by Jim Bunning's retirement in Kentucky). If the 2006 senate class were up for re-election this year, Democrats would potentially face very substantial losses, but fortunately for Democrats they aren't.
As such, I don't think it can be taken for granted that Democrats won't keep their "filibuster-proof" majority, or even expand upon it; I might put this possibility at something like 25-30 percent, following the Dorgan/Dodd retirements. On the other hand, the Democrats might also lose five, six, seven seats ... or perhaps more. I don't think the possibility of their losing their majority rates as higher than a small, single-digit number, although it cannot totally be ruled out if unexpected events (incapacitation of Robert Byrd and/or Daniel Inouye, a party switch from Joe Lieberman and/or Ben Nelson) come into play.