>To me the critical question is not what the average
>return is, but what the average is in, say, the top
>quintile of investors, and how stable this is. In
>other words, can you bilk the rest of the crowd if
>your trading practices are good enough?
The finance literature says no, and you can't prove statistically that the consistent top performers are there because of skill or chance. It'd take more periods than a trading lifetime allows for. Using more mundane tests like Hulbert's ratings digest, you can find very few consistent outperformers over the long term. Since you can identify them only after the fact, the info is largely useless for the future anyway.
The "house" can bilk - specialists, market makers, floor traders. An old college friend of mine who trades options in SF has been bearish since the early 1980s, but he still makes money. He just pockets the premium that usually wrong public traders sell him. He & his colleagues also make money from newcomers; when someone steps on the floor for the first time, experienced traders flock to him/her (mostly him) to relieve him/her of "tuition money."
Speaking of sample size, a paper by Philippe Jorion of the University of California-Irvine and William Goetzmann of Yale that assembled long-term stock index for 39 countries <http://www.ssrn.com/papers/9810/98100901.pdf> found that it was impossible to prove that returns in many European markets were significantly different from 0, even with data going back to the 1920s.