I'd start with either TTs or cyclocross. TTs are reasonably easy to model. If a rider takes it easy in a TT will he do well in the stage the next day or is he tired. Or if he does better than expected will he do well in the next stage. Might even be able to correlate it with the biological passport data if it's available.

Lap times are available for a lot of cross races. WHo will do best, the steady lap times or the more variable? often the guy who has the ability to hang on and then turn up the heat for the last lap (a la Vervecken) seems to do well.

Just some thoughts. Fascinating idea.

http://www.amazon.com/High-Tech-Cycl.../dp/0736045074

Read that book. He quotes a lot of statistics similar to what you're looking for with lots of references to university studies and publications that can get you even more info

Another thing you might try is digging through palmares. You could try to find the peak age for winning or patterns from year to year. You could try to predict if a rider wins X races in 2010 how many he will win in 2011. Maybe try to look for a pattern that predicts that a young rider is on the verge of something big or an older rider is losing it. Maybe compare GC riders to sprinters. Does one age better than the other?

This won't help the OP's case but, was Versus showing a percentage of work certain teams and/or riders were doing at the front during last year's TDF, or am I dreaming? I wonder if they really kept track of that or if they just kind of eyeballed and guessed.

Anyway, with the rise of the Sloan Sports Analytics Conference and stats in sports, I wonder how much goes on behind the scenes in cycling that we don't know about. I know it's very common in football for teams to track their own "secret" stats and the Mavs and Rockets do it in basketball. With big-budget, forward-thinking teams like RadioShack and Sky, I wonder if they keep track of how often Team X sends a guy in the break, works to bring the break back, etc. Or are they more content to track power data and spend time in wind tunnels?

I wonder if you could come up with something like an adjusted plus/minus for domestiques.