Lately I have been doing some reading on statistical arbitrage - largely for my work. Several interesting ideas exist in the domain. The practitioners in the advanced markets have explored various aspects of this idea and have in fact implemented a large variety - with varying degree of success. What interested me at a diffent level though was the curious fact that it exists.
It is quite debatable whether statistical arbitrage is arbitrage in the true sense of the term. Most practitioners anyway declare that it is not. Unlike "regular" arbitrage, the returns and success of a trade are not guaranteed. However the idea is that one is looking for an edge - success ratio of 70% will do excellently in most cases!!
Let us assume it exists and the total monies deployed in statistical arbitrage are making returns which cannot be explained by noise or randomness a la Taleb! The very existence of this sort of arbitrage is quite intruiging though. Why does it exist? Pattern theory may have some insights. Behavioral finance may have some ideas to throw at the question too!
My own limited thinking on the matter has reached thus far. It exists because there are not enough market paerticipants chasing it! Unlike the more visible movement in spot markets, the movements in the spreads between different underlings and in some cases even real and synthetic assets is not visible to a screen watcher. If and when sufficient automation spots these opportunities in time and exploits them at the very onset, the easy returns will go away. Something already underway in the developed markets!