Any science that doesn't have the luxury of proving by experiment is always on shaky grounds. I work in finance and have seen enough beautiful backtestings that perfectly predict the past but run into walls as soon as they go live, to take a prediction purely based on backtested mathematical model with a pinch of salt.