This among many things would be difficult. Many riders are unwilling to give out power numbers, or V02 Max numbers, because they allow opposing teams an advantage. If people knew what Contrador's 5 min power number was they have a target to work at.

Secondly there is simply to much variation in races. Weather plays a huge difference in the outcome of a race. How are you going to account for the weather, let alone the variations in weather over a 180 mile bike ride. Routes in many of the grand tours where betting would be most common change every year. Teams change every year, and any cycling statistic is going to be heavily influenced by the riders team.

You could probably collect data on how often a break away succeeds. However to produce a useful model yo have to take into account all of the variables. Here is a list of variables to review, and I'm sure there are some that I left off:

-Race
-Distance to finish breakaway begins
-Weather conditions at breakaway
-Weather conditions in peleton
-Power output of riders in breakaway
-V02 max of riders in breakaway
-Aerodynamics of riders positions in breakaway
-Level of cooperation in breakaway
-Power output of riders in peleton
-V02 max of riders in peleton
-Aerodynamics of riders positions in peleton
-Level of cooperation in peleton
-GC position of riders in breakaway
-Race stage within tour
-Course profile
-Road Conditions
-Fatigue level of riders
-Number of riders in breakaway
-Likelihood of a crash
-Likelihood of a mechanical
-Affect of GC leaders getting a mechanical

Theres many many more. Simply too many variables to have any kind of statistically significant model