As an aside, learning, like outperformance, is incompatible with the efficient markets hypothesis, according to which the markets follow a random walk and you can no more learn to trade them than you can improve at flipping coins. Our data therefore suggest the markets are not in fact random.