One of the best things about "Inside the Black Box." the book about quants that I am reading, is the authors refreshing candor in discussing the shortcomings of algorithmic trading. The quant’s Achilles heel is something called "regime change." Regime change is basically a fancy way of saying that the future looks nothing like the recent past. In fact, it becomes almost the polar opposite of what you would expect. Black becomes white, up becomes down, value companies get shorted, junky nearly bankrupt stocks rally frenetically. This is a nightmare for computer-based traders whose whole investment model is built on the idea that past patterns will repeat themselves in the future. When market behavior becomes asymptomatic, algos lose money.

I was contemplating this issue with respect to my own simple daily scalping. The irony of the matter is that I have railed for years against algorithms proudly describing myself as a "purely discretionary trader." Yet the older and better I get, the more systematic I become. Fact of the matter is that on a short-term time frame markets are mindlessly repetitive and once you observe a certain pattern in the price action you tend to take advantage of it continuously until and unless it begins to fail.

Yet discretion still matters. No matter how solid your setup, human judgment should still be the final determinant of action. Case in point was my scalping last Thursday night. I’d taken a short in GBP/USD because my round number setup triggered a signal. However, UK housing data had just come out and it was extraordinarily hot with prices rising more than twice the market expectation. Normally this would be very bullish for the pair, but it actually fell on the news and I smugly thought to myself, "Aha! See you should never doubt your setup!" as I watched my P&L grow bigger. Needless to say, the fall in the pound did not last. After a few moments the pair reversed itself and I watched in dismay as it climbed against me for about half an hour, finally putting me out misery like rabid dog as it hit the stop.

The next trigger came from a EUR/USD short which again I took. This time however I was a lot more wary. We were very bullish German unemployment data which was going to be released in half an hour time. Why do I want to be short euros into this data point? I asked myself. I didn’t. So I decided that I would cover the trade if it retraced back to my entry. That was a smart move, because after the unemployment report, the euro traded in a straight line to the upside and would have forced me to take another stop out.

There is of course the danger of constantly second guessing your system and such behavior is as bad as blindly following its every command. However on balance there is and I believe always will be room for discretion in trading. Human behavior cannot be perfectly modeled by a bot, and just as it takes a thief to catch a thief, so too it takes a trader to stop your system from making a reckless trade. Discretion, used sparingly is still the better part of valor.

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