But the dice experiment is: one leaves variables out of the model bc of its overwhelming complexity for what we know so far.

No, I do not buy this, vote me negative but I do not. If it was true, the weather forecast should be guessable.