Predictions, anyone can make them but few are ever correct. Jay Cost has a great set of predictions for this year's elections. It is interesting in that it relies upon some strong social science theory. Read the whole article for more data.
What is portrays is a rather pragmatic picture of mid-term elections. Congress is elected locally, by populations whose views do not seem to be subject to significant change. The electroates are more driven by economic factors than the performance of their elected officials or external ideology.
And the big question is, so what? There are only a few possible elements which may cause significant swings in the present political landscape, almost none of which are within either party's control.
One, Iran gets hot and military action is required. Look for this to be a benefit to the incumbents all around. No one wants to swap leaders in the middle of a war. For reference, see 2004. Two, the gentleman shot by VP Cheney dies, unlikely. This would (in my opinion) cause the VP to resign. You can not have an elected official in office (except for Ted Kennedy) who is responsible (while in office) for the death of a citizen. Even an accidental death. Three, the economy tanks. This would be the result of a devastating storm season, more problems in the Middle East or something dire in Latin America.
All of these items are possible, some more probable than others. But nothing from our friends on the right or left will influence the mid-term elections one way or another. For all the wionging on the left about wiretaps, Abramoff and the Valerie Plame affair, none of these circumstances rises to any level of seriousness in the eyes of the electorate.
As has been said by many in the past, all politics is local.