August 25, 2010 he predicted that the Democrats would lose 6 to 7 seats.
In his final prediction, he predicted that the Democrats would lose 7 seats.
They ended up losing 6 seats.
In the gubernatorial races he predicted on September 3rd that the Republicans would control 30 seats.
That prediction held leading up to the elections:
In the end he ended up correctly calling 36 of the 37 races.
In the House race on September 10 he predicted the house had a 2/3 chance of going the Republican's direction.
In his final House prediction he under estimated the number of seats the Republicans would take but it was still within the margin of error:
All things considered, we are certainly in the time range where his predictions have been meaningful and accurate in the past. It should be known that his models also consider the likelihood of election changing events, such as changes in unemployment between now and the election, in his predictions.