Pollster.com looked back at the polls right before the 2004 and 2000 caucuses.

In the last competitive Republican caucus, the final results were Bush 41, Forbes 30, Keyes 14, McCain 5, and Bauer 9.

Now, these pollsters may have improved their methods in the past eight years. But the Des Moines Register had Bush 43, Forbes 20, Keyes 8, McCain 8, Bauer 6.

The University of Iowa had Bush 55, Forbes 12, Keyes 9, McCain 5, Bauer 8.

Research 2000 had Bush 46, Forbes 23, Keyes 9, McCain 7, Bauer 7.

The L.A. Times had Bush 43, Forbes 25, Keyes 10, McCain 8, Bauer 7.

It’s tough to see any clear patterns or lessons we could take from those results, other than maybe everyone underestimated Forbes’ momentum. Most of them had Bush with a bigger lead than he actually finished with on caucus night. Keyes did a few points better than anyone expected. (Is Huckabee the closest one to Keyes in this race? Or is Keyes still the closest one to Keyes in this race?)

Still, it’s a different voter pool than eight years ago, and this race has a different dynamic – no one is a huge, clear frontrunner like Bush. While Forbes and Romney are both northeasterners with a lot of money and experience in the business world, I think it’s safe to say Romney’s a better retail politician than Forbes was. (No offense to Forbes.) Maybe Romney is Bush, as he’s been the frontrunner much of the year, and won the straw poll. Clearly, this time McCain isn’t skipping the state, so it’s tough to expect his 2000 performance to be predictive.

Still, perhaps one lesson we can take from 2000 is that we shouldn’t be stunned if the final results show somebody overperforming their poll position, and someone underperforming by a significant margin.