An experimental test of all theories with predictive power beyond quantum theory

According to quantum theory, the outcomes of future measurements cannot (in general) be predicted with certainty. In some cases, even with a complete physical description of the system to be measured and the measurement apparatus, the outcomes of certain measurements are completely random. This raises the question, originating in the paper by Einstein, Podolsky and Rosen, of whether quantum mechanics is the optimal way to predict measurement outcomes. Established arguments and experimental tests exclude a few specific alternative models. Here, we provide a complete answer to the above question, refuting any alternative theory with significantly more predictive power than quantum theory. More precisely, we perform various measurements on distant entangled photons, and, under the assumption that these measurements are chosen freely, we give an upper bound on how well any alternative theory could predict their outcomes. In particular, in the case where quantum mechanics predicts two equally likely outcomes, our results are incompatible with any theory in which the probability of a prediction is increased by more than ~0.19. Hence, we can immediately refute any already considered or yet-to-be-proposed alternative model with more predictive power than this.