As you can see, by the time you get up to 50, the Democratic defense list has started to get kind of uninteresting, while there are still some hotly contested offense seats left to discuss. It's a pretty good illustration of how lopsided the playing field for the two parties was this year. For instance, there's only one Democratic defense seat that we had left on our big board that fell off the list: Tim Walz in MN-01, who was Likely D but won by 29.6% (good for 67th place).

On the other hand, here's the continued list for offense seats!
51) NV-02, Heller, -10.4, Lean R
52) TX-10, McCaul, -10.8, Lean R
54) AZ-03, Shadegg, -11.1, Lean R
57) NJ-05, Garrett, -13.5, Lean R
58) TX-07, Culberson, -13.5, Likely R
59) WV-02, Capito, -14.2, Lean R
60) NY-26, Open, -14.5, Lean R
67) NC-10, McHenry, -15.2, Likely R
68) IN-03, Souder, -15.3, Tossup
72) FL-18, Ros-Lehtinen, -15.8, Likely R
73) FL-21, L. Diaz-Balart, -15.8, Tossup
77) OH-07, Open, -16.4, Likely R
80) NC-05, Foxx, -16.8, Likely R
81) PA-15, Dent, -17.2, Likely R
83) FL-13, Buchanan, -18.0, Likely R
89) VA-10, Wolf, -20.0, Likely R
95) IA-04, Latham, -21.2, Likely R
103) MO-06, Graves, -22.5, Likely R
128) LA-07, Boustany, -27.6, Likely R
142) LA-01, Scalise, -31.4, Likely R

The first thing I notice is that there are only six places where we got it "wrong," where wrong means we felt that, rather than leaving a race as "Tossup," we could move it to "Lean" or even "Likely..." only to see it go the wrong way. On the defense side, that means Bill Jefferson at Safe D, whose loss I think absolutely no one saw coming (the NRCC's four-digit campaign expenditures notwithstanding). It also means Nancy Boyda at Lean D. Although she seemed to have a comfortable edge in polls, her surprise loss provides a nice object lesson for incumbents defending tough districts: don't try to run a campaign that actually appeals to your constituents' logic and good judgment. Accept the DCCC's money, and use it to run negative TV spots, instead of trying to engage them intellectually with policy-specific newspaper ads.

On the offense side, the big screwup is Don Young at Lean D; again, this is one that basically no pundit saw coming, thanks to extremely consistent polling in favor of Ethan Berkowitz. The lesson here: never underestimate Alaskans' willingness to vote for more pork, even if it means supporting a felon (or soon-to-be felon) in the privacy of the voting booth.

We also had something of a crisis of faith in Bobby Bright in AL-02, in the face of tepid campaigning and a crimson district. Despite our dropping him late in the game to Lean R, his name rec and DCCC spending seemed to pull him over the line. Finally, we were caught off guard by the magnitude of the Obama coattails in Virginia, where we left Glenn Nye (VA-02) and Tom Perriello (VA-05) at Lean R. The polls just weren't there for them, in GOP-leaning turf, but the bluening of Virginia lifted them far enough. (If there's one candidate I'm personally shocked that won, it's Perriello; I was miffed to see the DCCC pouring money into a guy who seemed way too progressive for such a rural and downscale district. Here's one race where I'm super-happy to eat some crow.)

Where else did we whiff? IN-03 and FL-21 seemed like Tossups at the time, given the very close polling and baffled-seeming incumbents, but these ones are languishing up around #70. Apparently the constituents decided late in the game that, in IN-03, they had a challenger they just didn't know enough about (Mike Montagano), and in FL-21, probably a challenger that they just knew too much about (Raul Martinez).

We may also have been a little generous on the Louisiana challenges in LA-07 and LA-01 (both listed as Likely R). Jim Harlan, with a conservative profile and his own fat pocketbook, seemed like the best possible candidate for LA-01; however, given that this is one of the nation's most right-wing districts, I guess we have to take a 30-point loss (instead of the usual 50-point beatdown that we take in that district) as some sort of moral victory.

On the flipside, we missed a number of strong performances in California, especially the near upsets of Ken Calvert in CA-44 and Dan Lungren in CA-03. What's most interesting is that the rising blue tide in California seemed very evenly distributed throughout the state and probably tied to an Obama-driven boost in infrequent voters voting straight-ticket D, as higher-profile challenges to Dana Rohrabacher and David Dreier did only slightly better than completely under-the-radar challenges to guys like Buck McKeon, Wally Herger, and Elton Gallegly.

Where did we buck the odds? I'm pleased with how well we did at moving the right people to "Lean D" in the weeks before the election; at the time, it seemed a little audacious to call a win in advance for Gary Peters, Larry Kissell, Suzanne Kosmas, Betsy Markey, and Eric Massa in their fights against (lame) incumbents, but they all pulled it out... as did last-minute change Mary Jo Kilroy, who finally managed to pull it out in overtime and save us a lot of egg on our faces.

On the whole, we ran up a pretty good track record (while using the ass-covering category of "Tossup" a lot less than certain other prognosticators). The lesson here is that prognosticating is more art than science; your predictions are only as good as your polls and your scuttlebutt.

DavidNYC: This was indeed a very fun project and a tremendous learning experience, and I expect will continue to do race ratings in the future. It was also remarkably time-consuming, especially as we got toward the end - as Crisitunity suggests, there was a lot of back-and-forth as we pored over Google spreadsheets - plus the occasional bit of smack talk. But I think we'd all gladly take more cycles like the one just concluded!

I just have a few additional thoughts. I think our Senate ratings hit the mark, and I think we were in general pretty disciplined in not moving races until we had sufficient evidence to justify a change. Some examples I'm thinking of include OR-Sen and CT-04, where we insisted on seeing polling before concluding that popular, "moderate" Republican incumbents were truly in jeopardy.

On the flipside, I think sometimes you just have to acknowledge an open seat is gone, as we did early on by moving VA-Sen to Safe D in August, and later NY-13 in October. (Both of these were thirty-point races.)

One of our biggest flubs, though, was NY-24. We had the race as Likely D until a week before election day, when we moved it to Safe. A lack of polling, zero outside spending, and a seemingly unimpressive Republican who had been substantially outraised all convinced us that there was nothing to see here.

We couldn't have been more wrong. In the end, Mike Arcuri raised "just" $1.6 million (unimpressive compared to fellow freshmen like Chris Murphy, Patrick Murphy or even Paul Hodes), while Richard Hanna took in almost $1.1 mil. The final four-point margin was hair-raising, and suggests Arcuri still has a lot of work to do to establish himself. It also tells us that there will always be surprises - and that absence of evidence is not evidence of a Straniere.

James: Crisitunity and David touched on a lot of key points above, but I'll just add that I think that we all were a bit caught off guard by just how much of a focal point our race ratings exercise became in the day-to-day operations of this blog.

When I first drafted a preliminary set of race ratings at the tail end of 2007, David's response after I asked him for his thoughts was merely: "Nice work!" David later admitted to me that he felt as if he were a busy parent being handed a crappy piece of crayon art by a proud six year-old son. But once that crayon drawing was slapped on the fridge, if you will, we all realized that we would have to put in a great level of care into making sure we felt that each rating had a strong leg to stand on. In that sense, our race ratings project became the engine of SSP: we all had to step up our game to make sure that no major (or even minor) developments in the key House races would slip past us unnoticed -- especially after we achieved some early success by noticing MS-01 before anyone else did.

Our goal of making this ratings project as honest and credible as possible, I believe, had a great impact on our front page coverage, and I know we caught on to a lot of stories and developments that we may have otherwise missed had it not been for our relentless commitment to stay on top of things. There's no doubt in my mind that our ratings exercise, even if it provided no great revelations to anyone else, helped improve the work and quality of this blog immensely over my output in the summer and fall of 2006.