I read the entire article, and the problem is that it cannot distinguish between an effect and chance, and many have an incentive to call chance an effect. Also there's the announcement effect to consider -- once a player has a track record, people start reacting to his moves by buying the same stocks, after which the track record is no longer meaningful data.

The WSJ Dartboard Contest was a terrific test that could have shown a real effect, if there is one, and anyone who produced reliable returns would get a lot of free publicity. So there was a big incentive, but no one could show an effect. It was the closest thing to science I've seen, and it supported the null hypothesis.