APBRmetrics

I had posted something about this just before the board went down about 10 days ago, and it appears to have disappeared. I will try again:

I've created some unique datasets related to the NCAA tournament and have successfully fit a survival (event time) model to the data. It appears to preform better out of sample than most other predictors, correctly picking the champion every year from 2006 to 2010.

The problem that I am running into is that the predicted rankings are in fact relative to a baseline hazard function that is treated as unknown (i.e. Team X has Y odds ratio of survival compared to the baseline). What I would like to do is be able to turn those hazard rates into probabilities so that I could have predicted odds of Team X beating Team Y. Does anyone have any experience with this or thoughts on where I might find some help?

It doesn't sound like you have a model for predicting Team X beating Team Y, but it sounds like you may be able to use properties of odds ratios (http://en.wikipedia.org/wiki/Odds_ratio) to get the probabilities that Team X survives longer than Team Y.