Predicting finishing times

I would like to know how others go about predicting their finishing times for trail ultras, especially those that they have never run.

Recently, I predicted it would take me 12 hours to finish Voyageur. I ended up finishing in just under 11 hours. While I am certainly satisfied with the time, I can't help but think I possibly could have done better if I hadn't straddled myself with the lower expectations. There certainly is an element of self fulfilling prophecy in predicting race times, at least if those predictions are relatively conservative. You predict a certain time, so you go out at that pace, and lo and behold, you finish near the predicted time.

One strategy I have used is to look at the previous results for a race, and shoot for the median time. I have noted from my previous races that I generally finish right in the middle, near the median. So for example, before running the Wild Duluth 50K last year I looked at the 2009 finishing times, noted that there were 58 finishers and the 29th runner finished at 7:18. I used that as a ball park for my own time, and sure enough, finished in 7:05. But again, I can't help but wonder if I could have done better if I had predicted a 6:15 instead.

It is worth noting that this strategy works best with more data. Using a single year's race data is of limited use unless the weather and trail conditions are going to be identical, and there was a relatively large number of finishers. I just signed up for the Marquette 50 mile race on Aug. 20th. There were only 9 finishers last year. Number 5  finished in 11:32, but number 4 finished at 10:08.

This strategy is also limited by my own self confidence. I see my self as a median runner. Maybe if I saw myself as a top 25% runner my times would improve?

Well. What are your strategies for predicting finishing times?