Princeton Election Consortium is once again diverging pretty markedly from most other aggregators and as of today has Dems back to a 50% probability of holding 50 seats. Pollster has gone the opposite direction over the same week--moving slightly from 62% to 66% R takeover--and so far I can't determine the basis for the divergence; though I assume it comes down to different calculations of polling 'house effects' that would give us GA and maybe hold IA and CO. For what it's worth--and to some readers it may be worth a lot---538 has followed the same trajectory of change as Sam Wang over the past week, though since Nate et al were starting from a much more pessimistic point a week ago, they're now at essentially 60/40 R takeover.
Both Nate and Sam would, of course, stress that the difference between 50% and 40% probability is very, very slight. However, both are also supremely invested in being right come election day, so movement over the next week on both sites will be very much worth watching. But don't spend too much time watching. Spend your time helping to GOTV in GA, KY, IA, CO, LA and wherever else you're so inclined. (And whatever happens in the Senate, it's a certainty that almost the entire '10 class of excremental R governors are staring at the abyss, so let's push them over...)
2:13 PM PT: I should add that even the (to me) absurdly bearish NYTimes model has followed the same positive trajectory in the past week, sitting now at 64/36 vs about 75/25. They also have a handy aggregator-of-aggregators at http://www.nytimes.com/... Again, I would stress that it's the trajectories that interest me more, not the absolute numbers (though obviously i'm happiest with Sam's)