Feedback loops


A recent article in Wired magazine titled “Algorithms Take Control of Wall Street” has gotten a lot of attention recently.  After an almost 5 month long rally, the subject of HFT and May 6th seems to have all but been forgotten.   We profiled this article about a week ago and highlighted its main points.  One of the authors, Felix Salmon, wrote a piece on his blog yesterday where he included some more info that wasn’t in the original piece.  Most interesting were the comments by Prof.  Michael Kearns of the University of Pennsylvania ( ):

“There’s a growing understanding and belief on Wall St, especially intraday, that there’s a strong game-theoretic aspect to the market: the performance of a given strategy depends on what other strategies are trading in the market at the same time. My payoff is a function of what I do and what the other players in the game do…The concepts of strategic interaction are important. Game theory is hard to understand even in simple cases. And now strategies are adaptive, so it’s especially difficult.”

“Our financial markets have become a largely automated adaptive dynamical system, with feedback

 “The quant meltdown demonstrated the unexpectedly high correlation between quant strategies who believed they were doing different stuff from each other…There’s some echo of that in adaptive algos: if we’re all using somewhat similar adaptive algos on the same data, then the data itself correlates us”.

What Prof. Kearns is basically saying is that once again Wall St is chasing its tail.   When the herd gets an idea, they all chase it.  Somebody comes up with a winning strategy like HFT, it drives in lots of new entrants to replicate the same strategy and then becomes one giant feedback loop till it pops.  And then everybody stands around scratching their head, wondering what the heck just happened.  They will say, “Why didn’t the regulators see this coming?”  Well, unfortunately, they did see it coming.  They are just outgunned and overmatched and have almost no chance of stopping it now before something bad happens again.