This is fine as far as it goes, but you've redefined free will away from what most people mean by it, which is basically the same thing as saying free will doesn't exist (because libertarian free will seems very unlikely to exist), so the title seems misleading.

I agree with Gordon: I don't think that free will is unpredictability per se.

Determinism needs to be distinguished from predictability. A universe that unfolds deterministically is a universe that can be predicted by an omniscient being which can both capture a snapshot of all the causally relevant events, and have a perfect knowledge of the laws of physics.

The existence of such a predictor, known as a Laplace's demon is not a prerequisite for the actual existence of determinism, it is just a way of explaining the concept. It is not contradictory to assert that the universe is deterministic but unpredictable. But there is a relationship between determinism and predictability: predictability is the main evidence for determinism. Nonetheless, determinism itself is the crux, and predictability only features indirectly as evidence for it.

.

Note that according to both QM and GR you don't need to know "all causes", only the initial state on a given spatial hypersurface, in order to be able to do a Cauchy development of that hypersurface, subject to some rather technical and non-restrictive energy conditions. Whether the initial state can be known exactly depends a fair bit on what one wants to calculate (e.g. a full quantum state, or a classical state). In that sense, one can, at least in theory, predict the behavior of a physical object, including a human, to a high degree of accuracy, though of course never 100%, and the remainder is not related to light cones, just to the uncertainty of one's state, especially the quantum state. 

Still, if another human can predict your actions with high accuracy, though not 100%, how much free will do you really have? 0.1%? Would you be happy with that?

The picture is of a flat space. Weird topologies might break the proof.

If in a deterministic universe, no observer B can 100% correctly predict the behavior of subject A, except when B is in the future of A, we can say that subject A has free will. 

This strikes me as a very unintuitive definition of free will.

We often talk about free will experientially; defining it based on a specific external observer strikes me as odd. But before I critique that in any way, I'd love more clarification on what you mean by "observer." Is this anything capable of prediction (e.g. a faithful simulation of the universe)? 

But more importantly, I think "100% correctly" is doing the bulk of the work here. I fully agree with your claim that if we define free will in a manner similar to this, we will never reach it. But really, very little outside of statements within self-contained axiomatic systems can ever be held to the standard of 100% certainty. If your concept of free will hinges on the realistically minuscule chance that a random event will alter your decisions substantively, then I ask if this conception still resembles anything like the idea of "free will" as we tend to think of it. 

Overall, I concede your claim follows from your definition. But I question the usefulness of such a definition in the first place. I think we can all agree that we cannot have 100% predictive certainty. The question is more whether or not we want to call that shred of uncertainty "free will." Semantically, I think it's confusing to call this "free will" when that is not usually the intended meaning of the phrase, but ultimately the decision is somewhat arbitrary as our experience remains the same regardless.