Now think about what would happen if Every. Single. Choice. was simulated until you look at it, at which point half of the work would be wasted. It would be so slow.
Half the work would be wasted after one choice, potentially 3/4 after two choices. Potentiall 7/8ths after a third choice. Less if some outcomes of some choices preclude later choices being made, or conversely remove from play prior branches.
Like Something might or might not happen the 'virtual' Qubit (this not being run on a true quantum computer, but being emulated) represents as 0 or 1. SomethingElse might or might not happen, resulting in a further 0 or 1 (we now have 00, 01, 10, 11). If a ThirdSomething must(/could) happen only if neither of the prior two happened then only 000 and 001 are viable branches. If a ThirdSomethingElse could (though need not) happen if only one of the prior two happened then 010, 011, 100, 101 are also viable branches, though arguably you'd be talking about a total of 000-, 001-, 01-0, 01-1, 10-0, and 10-1 as valid states of four seperate qubits. Whether there'd be 11--0 or 11--1 combinations (something arising from the conjunction of both initial possibilities both being true, assuming the fifth qubit position is allocated to the actions only possible on that). The nature of the quantum calculation would result in there never being any valid 11111 solution (it would 'collapse out' of the calculation, being a nonvalid state in classical terms).
But, anyway... The way I see it is that it would work upon the "Hidden Variable" level. The game would have a definite idea of the 'sub-planck' activity. It would be data intensive (not so much as if emulating true superposition information), but apart from the game's induced randomness (a deviation from the how the Hidden Variable idea is supposed to work in the Real World, of course) otherwise linearly deterministic. It would then hide it behind a layer of 'macro-world' abstraction which obscures the pure determinism and 'fudges' macro-explanations that appear to indicate a quantum-level multiplicity of states. i.e. the "split experiment" or "quantum tunelling" phenomena is explicitly coerced from the background data on a "coulda, shoulda" basis.
If you see what I mean. And I wouldn't blame you if you don't.
(For the record, I
am a believer in the Real World being utterly deterministic, below the Planck-dsitance and Planck-time resolution which is the limit of our detectability. Apparently instantaneous "Action at a distance" effects are just a result of this data sloshing around affecting things accordingly, or perhaps something more exotic that is beyond the wit of man... Imagine a 'creature' within Conway's Game Of Life. Complex enough to be 'conscious' (to itself, though to us 3D, non-gridded creatures, it would just be a trivial fully deterministic automata), it could not possibly correctly assimilate a conscious awareness of individual on/off cells, and probably not even individual 'glider' combinations (save for the disturbances they set up in its cellular "structure") which might pervade the cellular 'universe' and fulfill the same function as photons, or even gravitons in certain circumstances.
Apart from being able to detect an "orthogonal bias" to the structure of the universe (for much the same reason as all edge tiles of a square area in DF are equidistant from a central point), it would be hard to understand how a Conway-esque 'flatlander' creature (to bring to bear Edwin A. Abbott's novel, as well as Ian Stewart's 'sequel', "Flatterland") would imagine the structure of the universe 'he' resided in. It certainly wouldn't feature "oh, I'm just a simulation ticking away on a massive supercomputer, whose transisters operate by means of doped silicon connections altering the availability of electron-holes", for that would be something way beyond the ken of the Conway-creature. Ditto, doubtless, the 'mechanics' behind our quantum effects and whole universe (not for one minute suggesting that there's a Great Programmer who set it up, though!), and so the DF universe (whose Great Programmer we ourselves know to exist) would be trivially deterministic to ourselves, and yet exhibit 'quantum' effects to the inhabitants as far as the detail of the simulation allows.)
But I think I ramble too much. I wasn't going to write the above two paragraphs, but it looks like I ended up doing so. (That's determinism for you!