The idea of superdeterminism is not really about free will. Free will is a concept that is very hard to define in a logical-positivistic way. If you don't believe me, try to define it! If you can't say exactly what you mean by a notion, in terms of "If I do this and that, what happens?" then it is not clear that the notion is well-defined. There are many questions which are just your brain fooling you into seeing sense where there is none.
First, I want to say that 't Hooft's original ideas were profoundly nonlocal and would not resemble the local cellular automata ideas of, say, Wolfram. In recent years, 't Hooft has considered the idea that there is a deterministic theory that is local in space-time which reproduces quantum mechanics. This idea is clearly wrong, and 't Hooft is talking nonsense.
't Hooft's old ideas of an underlying realistic theory were not so silly, since they came on the heels of the holographic principle. Once you realize that gravity is defined far away on a holographic screen, the idea of hidden variables becomes more plausible, because the physics of gravity is nonlocal in a way that suggests it might fix quantum mechanics. There is no real proposal for doing this, however, just vague speculations.
Holographic hidden variables could conceivably even be holographically local, meaning that they are local on a holographic screen. There is very little that can be said without a precise proposal for what these variables are.
But if you want hidden variables without holographic non-locality, then you are in trouble. You are trying to get out of the problem using "superdeterminism", the idea that the polarization settings are determined in advance, and so that Bell's inequality violations do not necessarily mean that local hidden variables are logically impossible.
Superdeterminism is silly
The proof of Bell's theorem tells you that measurements of different polarizations of far-away particles have statistics that are not reproducible in advance by the electrons alone, making crib-sheets when they are close about what the answer is going to be for the different experiments.
If you want to use the superdeterminism loophole, you need to assume that there are electron crib sheets which tell the electrons how to behave, and further, there is some mechanism which links the electron crib sheet to the choice of the experimental apparatus of which direction to measure, so that the direction one chooses to measure is somehow determined by these crib sheets.
To understand how ridiculous this is---- I could program a computer to run a random number generator, and set the polarizers according to the outcome. Then whichever random number generator I choose to use, the result must be correlated in the exact same way with the crib sheets.
If I use a thermal random number generator (a heated chip which reads out random 0s and 1s), the result will have to be correlated with the crib sheets. This correlation cannot change even if I change the temperature, altering all the Avogadro's number of particle positions and velocities. It doesn't change if I touch the chip to a hot liquid, introducing new atoms. The correlation doesn't care if I flip a coin and switch out the random number generator for another one, or if I rewire the experiment to make different outcomes correspond to different polarization settings. The nature of the conspiracy is so implausible, that it requires an intelligent agency which knows exactly what I am doing, and rearranges all the crib sheets and correlations to make everything come out right.
It is just plain impossible to imagine such a mechanism. It defies common sense that for any randomization procedure one can dream up, the results are correlated with the electron crib sheets. Further, if you have a beam of different correlated electron pairs, you might measure this electron pair or that. The mechanism has to be correlated with all the electron crib sheets. I think that this is sufficiently ridiculous that to call superdeterminism a loophole is just an abuse of language--- it is a loophole in the same way that we could all be dreaming in the matrix and the aliens have set up the outcome to look like quantum mechanics is true. It's no more plausible than this sort of nonsense.
You brought up the issue of free will, and this is an old saw from philosophy. The actual history of the idea is important--- it comes from a religious paradox:
- If God knows what is going to happen, and made it all happen outside of space and time, how can God punish people for doing evil by sending them to hell?
This is the essential question that bothered people about Christian theology, and led to free will debates. The notions in this question are very hard to define in a logical-positivistic way, and when you do define them this way, the problem evaporates. There is no problem here, and there never was, independent of the fact that the notion of God has not been properly defined, nor its properties in any way deduced from a framework which is capable of persuading anyone in any way except by force of social convention.
To make a logical-positivistic description, you have to define all properties in terms of sense experience. So one can take a definition of free will as follows:
Free will (version 1): If I have a ham sandwich and a cheese sandwich, and I place them in front of me, and I am told to "take one and eat it", then I end up holding one of the sandwiches and eating it.
This is clearly no good. The idea we have of free will is not that we do things, but that we could have done something else. So try this:
Free will (version 2): If I have a ham sandwich and a cheese sandwich, and I am told "If you disobey this prediction, you will get $1000. I predict that you will take the ham sandwich and eat it", then I will take the cheese sandwich and eat it. Likewise, if I am told "you will eat the cheese", I will eat the ham.
This definition is not so great either. It is saying that I am capable of spiting any prediction about my behavior, if I am motivated to do so. This is independent of determinism: if the universe is deterministic, like a being in a computer simulation, you can still have this type of spiting behavior. All it says is that if you predict the outcome, and then tell the person the outcome you predicted, you change the outcome, so that it cannot be predicted anymore.
But this is the closest I can see to making sense of the concept of free will. So free will for me means the following:
Free will (version 3): Given access to the predictions any algorithm that purports to predict your behavior, and gives you incentives to spite the prediction, an agent has free will to the extent that the predictions will not come true.
This is true of people: if you tell a heavy smoker "Do not smoke for a year, and you will get a million dollars", then it is likely that the person will not smoke for a year. But this is clearly nothing to do with what people's intuition about the thing is. The intuition is that the prediction can be spited even if it is behind a curtain, hidden from the agent.
But if you don't tell the agent the prediction, there is no sense in saying the agent is somehow behaving non-freely in doing what you predict. I predict that you will get out of bed tomorrow, but this doesn't mean that you don't get out of your own free will. It only means that if I tell you that I predict this, and give you a big incentive to stay in bed, like a million dollars, and you still get out of bed.
The God business at the beginning is really resolved by defining God properly, but even supposing you believe that God is an external agent that knows the future and punishes sinners in the afterlife (something which I can't make logical positive sense out of), the fact that God knows the future in this metaphysics does not mean that you didn't choose it, since God didn't tell you the prediction and ask you to spite it!
In a certain sense, actually, in many religious traditions, God does tell you some predictions about human nature and ask you to spite them--- the prediction that human beings will be cruel and capricious, for example. This type of thing is asking human beings to spite predictions about the general nasty character of human relations in a Darwinian world, and the insistent demand that one spite these predictions, despite there being no incentive to do so, is the major purpose of religious belief.
Anyway, this is a major digression. The point of this is that the concept of free will is not well defined, and any way of defining it positivistically, it is either obviously true that human beings have free will, obviously false, or obviously meaningless to ask the question. The fact that free will has no definition, or at least, no consistent agreed-upon definition, should make one pause whenever someone discusses the concept, since this person can impose whatever definition he or she likes on it, and argue from this metaphysical position.
The position that superdeterminism means we have no free will is only true in a sense that it is determinism. This is not conflicting with free will in the definition I gave above, since determinism doesn't tell you what your choice is going to be and ask you to spite it, less does it give you incentive for doing so.
One way to try to violate the definition of free will above is to send correlated electrons to distant experimentalists, Alice and Bob, and try to predict Alice's polarization settings by capturing the electrons along the way and measuring their crib sheets (imperfectly, by measuring their spins). The issue of course is that in the superdeterministic view, the person capturing and measuring the electrons is not predicting anything about Alice and Bob, because the electron's crib sheets are now correlated with this person's polarization settings, and not Alice's or Bob's anymore. This sort of nonsense makes all experimental thinking and scientific hypothesis testing impossible, and it is really a form of magical-universe hypothesis. One must reject it a priori.
So, unlike the concepts of even God and religion, I can't see any way to make sense of the concept of superdeterminism in a logical-positivist framework. I personally consider the answer to the question as a resounding "no". No, it is not true that there is a superdeterminism loophole, and local hidden variables are just plain ruled out by Bell's inequality violations.