Dataminers - a question for you.

How big a dataset do you need to have a good read on a good number of opponents that you might play against regularly?

To break that down a little, if I always play SNGs for the same 4 hours a day at the same buy in at PP, how much datamining on the same 4 hours would I need to do before I have good data on 10% of the players, or 20% etc. And how often would I play against one of those players I have good data on?

Anyone done any research on this or have experience of it?