There's a pretty good essay from Scott Aaronson [1] that tries to disentangle the question of randomness from unpredictability, by saying that unpredictable events (which he says are the only things we know of physically that might permit something like free will) are events to which we can't assign a probability distribution.

It separates out the notions of "random" events like coin flips, which we know will go heads with p=.5 and tails with p=.5, from truly unpredictable events, where we can't say even in principle what the probabilities of the outcomes might be. These latter events are where he finds some wiggle room for free will.

[1] http://www.scottaaronson.com/papers/giqtm3.pdf