There were predictive estimates.

Washington Post estimated Ahmadinejad victory by margin of 2:1 about 3 weeks before the election based on polling data.

There are also rumors sifting out of the intelligence community that the margin was expected to be about 2:1.

Obviously the rumored intelligence report I cannot verify.

The Washington Post report is elaborated on here:

http://news.bbc.co.uk/2/hi/middle_east/8101841.stm (Just search the article text for 'Washington Post' if you are not interested in the rest.)

I'm starting to get the feeling that this was not a rigged election.

That Washington Post survey you mention is flawed for a number of reasons, the main being that while it showed Ahmadinejad ahead 2:1, it was only with 33.8% saying they'd vote for Ahmadinejad, 13.6% for Mousavi, and a full 42.5% of the respondents saying that they were either undecided or would not answer the question.

Obviously a survey in which the amount of undecideds is far greater than the amount saying they will vote for a candidate is hardly evidence of strong polling for that candidate.

More: http://www.fivethirtyeight.com/2009/06/did-polling-predict-a...

Let's suppose, however, that 100% of the undecided went for Mousavi. Well, he would have just barely won, about 55-56%. It is statistically improbable that 100% went for Mousavi. So while we are not talking about something that was mathematically impossible, we are talking about something that was statistically improbable.

There was at least one forecast, but as noted it was hard to draw any real conclusions from it.

However, there is also a fair bit of data on past elections and voting trends - which the data from this election departs from remarkably.

Thinking about it, making a forecast in an unfree society might be quite difficult. People might be reluctant to tell random strangers who they would vote for - could be spies from the government.