There are things that science can never "prove wrong", because they aren't set up as falsifiable hypotheses (eg: lucky charm hats). The best that science could

There are things that science can never "prove wrong", because they aren't set up as falsifiable hypotheses (eg: lucky charm hats). The best that science could do is establish a lack of correlation (assuming a certain set of control factors). Interpreting the result of such an experiment is strongly a function of your null hypothesis (or bayesian prior). It is only by applying the law of the excluded middle that one "reasonably" concludes that a lack of indication is an indication of there being no correlation. Also, the cost of wearing a lucky hat is negligible compared to the happiness gained from your team's victory (and often adds to the atmosphere/excitement). That is why people can "believe" in things that could reasonably be dismissed.