They played three times at the Australian and nadal won each time. But the law of averages would favour Federer in tomorrows final.
I hate to mention this, but if those are uncorrelated events (like in a Markov process), the law of averages does not help on an event by event basis. Worst, in this case, if there is correlation, it would help to maintain streaks.
What the law of averages actually tells us, is that if their number of matches goes to infinity (definition of hell to our fellow poster Denis), on the limit the actual win/loss ratio will reflect the correct probabilities of each man wining. The distinction between isolated events and their infinite collection is a tricky one.
All this to say: it is against all odds, my friend.