I've always wondered how learnable an understanding of unlikely events is but never investigated further. I guess a population of online poker players with increasing number of hands played (10k, 100k, 1 million etc.) (preferably winning ones) and others could yield interesting results.