| Here is the comparison of my 2012 predictions against the results. I know that not all the results have been finalized, but it doesn't seem likely that any result will change at this point.
Most of my predictions were really close, though there were some surprising results where my prediction was way off, and some of the safe races had no polls at all, including some states with presidential polls (ie. Alaska).
For the presidential race, I:
(1) Weighed state polls and national polls for the presidential race. Here is an example, using Hawaii.
Merriman River Group/Civil Beat 10/24-10/26: Obama 61, Romney 34 (D+27)
Monmouth 11/01-11/04: Obama 48, Romney 48 (EVEN)
Rasmussen 11/01-11/04: Obama 48, Romney 49 (R+1)
Washington Post/ABC News 11/01-11/04: Obama 50, Romney 47 (D+3)
Politico/GWU 10/29-11/01: Obama 48, Romney 48 (EVEN)
NBC News/WSJ/Marist 11/01-11/03: Obama 48, Romney 47 (D+1)
Public Policy 11/01-11/03:Obama 50, Romney 47 (D+3)
United Technologies/National Journal 10/25-10/28: Obama 50, Romney 45 (D+5)
Fox News 10/28-10/30: Obama 46, Romney 46 (EVEN)
North Star Opinion Research 10/23-10/25: Obama 47, Romney 48 (R+1)
(2) I then averaged the national polls and used them against the state's margin in 2008 (Obama won Hawaii by 45.26%).
National poll average: 1.11
2008 margin: 45.26
Adjusted margin with 2012 national polls: 46.37
(3) Finally, I weighted the national polls against the state polls. (If there were no state polls, the national-adjusted number would be the prediction.) If there was 1 state poll, I would multiply each by half. If there were 2 state polls, they'd get 2/3 of the share; 3 would get 3/4, etc.)
State polls: 27.00 x 0.50 = 13.50
National polls: 46.37 x 0.50 = 23.19
Prediction: Obama will win by 36.69%.
For the governor and Senate races I simply averaged the polls. Races without polls I put as safe for the incumbent party.
For the House races, I weighed pundit ratings with polls (if applicable). No polls meant the average pundit rating (Cook, Sabato, Rothenberg, CQ) would get 100% of the weight, 1 poll would get 50%, 2 polls 67%, etc.