When you think about it, there are really really only two ways to trade. You can trade statistically or you can trade analytically. All technical trading is in fact statistical in nature. Whenever you are coding a technically-based system you are basically saying that if A occurs, do B and over 100 or 1000 or 1 million samples, and you will succeed because according to past price behavior the profits will exceed losses.

That’s it. That’s all that technical trading comes down to. And therein lies the problem. Modeling past behaviour successfully is extraordinary difficult. Mark Twain once said that history does not repeat itself but it rhymes. The same can be said for the markets. Unfortunately, those rhymes can contain just enough of a difference to turn profitable strategies into losers. Ultimately, that is what happens to all good technical ideas from such simple algos as "Dogs of the Dow" to much more complex strategies like stat arbitrage.

Analytical strategies on the other hand don’t suffer from this fate, but they have problems of their own. If you think about the great hedge fund managers of our time, most of them follow the analytical model. But while the weakness of statistics is that it applies a consistent approach to an ever changing environment, the problem of analytics is to produce consistent accuracy in judgement, which is extraordinarily hard to do. For one George Soros, there are a million John Paulson’s -- guys who are good for a year or two, but who flame out with the destructive intensity of the worst of the MT-4 robots.

So what's the answer? Hell, if I know. At its core, trading is about price and price behavior, so a statistical understanding of price action is paramount if you want to succeed. But it is not enough. Ultimately, I believe that every stat strategy must have an analytical overlay. The more I trade my strategies the more I realize that the key to success has less to do with taking trades and much more to do with NOT taking them.

The NOT taking part is the really difficult aspect of trading because this is where the analytics come in. The computer is a dumb machine. It does not understand context, that why trusting all your money to a computer is idiotic. Judgement is crucial if you want to thrive and survive, which is why purely technically-based systems always fail in the end.

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