It stops working nicely like that once you go to slightly more complicated than measuring two photons with spin up and spin down. Then you find correlations that make no sense at all if you assume that each photon has a definite state before you measure them (for example, it's as if some events have negative probabilities).
