How Did DCCC Polls Perform?

The following are "independent expenditure polls" performed for the the DCCC in October 2010 and posted on the DCCC blog. Here's how they did.Poll (10/26, CA-20): Rep. Jim Costa 47%-Andy Vidak 41%. Actual Result: Undecided, Vidak currently leads 51%-49%. The poll appears to have gotten the winner wrong and to have overestimated the Democrat by about 8 points. Poll (10/18, AL-2): Rep. Bobby Bright 51%-Martha Roby 39%. Actual Result: Roby won with 51% of the vote. The poll was wildly off on both the winner and the margin of victory. Ouch. Poll (10/18, AR-1): Chad Causey 44%-Rick Crawford 42% Actual Result: Crawford won by 9 points, 52%-43%. The poll got the winner wrong and was off on the margin by 11 points. Poll (10/11, NC-07): Rep. Mike McIntyre 52%-Ilario Pantano 41%. Actual Result: McIntyre won by 8 points, 54%-46%. The poll got the winner right and was pretty close on the margin of victory Poll (10/11, HI-1): Rep. Colleen Hanabusa 48%-Charles Djou 44%. Actual Result: Hanabusa won by 6 points, 53%-47%. The poll got the winner right and was within 2 points on the margin of victory. Poll (10/11, NC-11): Rep. Heath Shuler 54%-Jeff Miller 41%. Actual Result: Shuler won by 8 points, 54%-46%. The poll got the winner right, overestimated the Democrat's margin of victory by 5 points Poll (10/11, IA-3): Rep. Leonard Boswell 49%- Brad Zaun 41%. Actual Result: Boswell won by 4 points, 51%-47%. The poll got the winner right, overestimated the Democrat's margin of victory by about 4 points Poll (10/11, AZ-5): Rep. Harry Mitchell 46%-David Schweikert 39%. Actual Result: Schweikert won by 9 points, 52%-43%. The poll was wildly wrong on both the winner and the margin of victory. Poll (10/11, PA-15): Charlie Dent 45%-Rep. John Callahan 43% Actual Result: Dent won by 15 points, 54%-39%. The poll correctly predicted the winner, but was wildly off on the margin of victory Poll (10/11, IL-14): Rep. Bill Foster 48%-Randy Hultgren 38%. Actual Result: Hultgren won by 6 points, 51%-45% The poll was wrong on the winner and off on the margin of victory by 16 points. Poll (10/4, NY-20): Rep. Scott Murphy 51%-Chris Gibson 38%. Actual Result: Gibson won by 10 points, 55%-45%. The poll was wildly off the mark, getting the winner wrong and also the margin of victory by 23 points. Wow. Summary: These 11 polls were all over the place, performing well in several cases but abysmally poorly in others. Of the 10 races that have been called, the polls got the winner right in 5 races and wrong in 5 races (and probably another one as well). Margins were wildly off the mark in several races as well. In sum, I wouldn't bet the ranch - to put it mildly - on DCCC "independent expenditure polls." UPDATE: A reader emails me to point out that at least some of this could have been the result of "selection bias," in that the DCCC "would only release the favorable polls and not the unfavorable ones, so you get a bias towards the Democratic candidate." What do you all think?