The problem is here "beyond what chance would dictate". It's not as easy as it sounds to find out what chance would dictate.
Mauboussin and Arbesman used a Monte Carlo simulation and compared the experienced streaks with the simulated streaks. Problem is that a Monte Carlo simulation generates a Gaussian distribution. Trading results are not Gaussian distributed. They tend to cluster. Therefore experienced streaks are longer. You can most likely see this already when you analyze your own trade results. The "fat tails" effect that generates long streaks is a characteristic of the markets, and is no proof of the existence of trading skills.
A proof of the existence of trading skills would be when a significant number of trading institutions with human traders would not have streaks, but consistently perform above market. This is not the case.
http://www.elitetrader.com/et/index.php?threads/we-are-the-95.228628/page-5#post-3327399
Even the smartest FAIL big emg
http://en.wikipedia.org/wiki/Long-Term_Capital_Management
They also cite authors who use bootstrap analysis to conclude the same thing, e.g. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1974577