That's based on my experience few months ago testing different pools. This may change along time and it's probably not exact. But maybe you can provide a better aproximation to this value.

Maybe we can assume time between solved blocks / network power as a proper proportion to know how many time it takes to solve a block having X hashing power.