One of the more popular clips of me on YouTube is called “Trading is not Logical”. It was done during the meltdown of 2007 when many algorithmic driven funds sustained massive losses on their strategies. I was being interviewed by CNBC Europe and very emphatically stated that we were “huge adversaries of algorithmic trading because “trading is not logical, it’s physiological” and machines would never be able to identify the subtle shifts in market behavior.

At the time, Kathy and I were working at 32 Old Slip in the same building at Goldman’s Alpha fund which was one of the algo hedgies most heavily hit and I admit to having some very smug satisfaction that the “masters of the universe” had finally tripped up. However a few years older and somewhat wiser, I have amended my view. Algorithmic trading can be a legitimate part of your portfolio and in fact I believe will become ever more popular in the retail FX market.

Computer assisted trading is of course huge already. Anyone who has a Metatrader account will attest to that fact. The pitch for automated traded is devastatingly seductive. No need to sit at your desk watching every tick. You can just program an algo and let the computer go to work. You can enjoy a life of leisure while your account grows like your very own FX piggy bank.

The reality is of course far different. The toxic combination of badly designed trading setups and sub-par broker execution often result in bigger drawdowns self traded accounts. Nevertheless, the concept of computer assisted trading is quite valid as long as you recognize it limitations. One of those limitations is execution. A setup that looks nice and orderly on a chart can be a cluster*** in real life in fast moving markets. Entries will be missed, phantom stops will be triggered and one dangling order on your books could ruin a good week’s worth of profits. The first rule of business when it comes to algo trading is to understand that just because a computer is executing your setups you still need to supervise and audit the account very closely.

The second limitation to appreciate is that most of the algorithms that traders use are horrible. I bring this topic up because the September issue of Active Trader magazine has a very interesting interview with Marco Dion who ran quant trading strategies for JP Morgan and other institutions in Europe. The key takeaway from Mr. Dion‘s interview is that much of his quant analysis was based on fundamental rather than technical data. Contrast that with most retail robots out there who use nothing more than glorified versions of moving average crossovers and you begin to see why most retail traders lose money with their algos.

Think about it for a moment. Price is a derivative of news and sentiment. Technicals are a derivative of price. When you trade techicals you are therefore trading a derivative twice removed from reality. You are inevitably late in your analysis and are reacting to the past rather than anticipating the future.

I don’t mean to imply that technicals have no role to play in algo trading. I merely suggest that they should not be the sole reason for your setup. Price action always has a context and Mr. Dion has a very clever way of dealing with the invertible change of context that destroys the validity of your algo. I will discuss that issue next week.

Boris Schlossberg serves as director of currency research at GFT, and runs bktraderfx.com.