Blumenthal notes that Romney is leading in 8 of the 11 most recent Iowa polls and declares him the most likely to win tonight. A caveat: 

[T]hat math assumes two things that may not turn out to be true: First, that the polls are collectively free of statistical bias, and that any shortcomings in the way they identify the likely electorate are not skewing the snapshot in some meaningful way. Second, it assumes that any surges in support for a particular candidate are over and have not changed the standings over the final days of the campaign. The history of polls conducted before Iowa's caucuses says both assumptions are questionable.

Harry Joe Enten likewise bets on Romney:

Overall, polls and models aggregating these polls points to a Mitt Romney victory with Ron Paul in second and Rick Santorum in third. Among the four major models/averages (538Polls and VotesHuffPollster, and Real Clear Politics), the order of the top is the same. 

Nate Silver agrees that Romney is the favorite but cautions that there is room for error:

The critical thing to keep in mind is that a wide range of outcomes remain possible in Iowa even at this late hour. Our forecasts, which are derived from the polling but account for its historical accuracy or lack thereof, project for example that Mr. Romney is most likely to receive about 22 percent of the vote. However, the margin of error on Mr. Romney’s forecast (enough to cover 90 percent of all possible outcomes) is about plus or minus 10 points, meaning that Mr. Romney could plausibly finish with as much as 32 percent of the vote or as little as 12 percent.