Developing a profitable trading system requires extensive research and backtesting, but only when a system succeeds in meeting these four qualifications is it truly ready to be applied under real market conditions.

I have received several E-mail questions recently about developing a trading system that not only works on backtesting, but also in the real-time, real-world markets. These are common questions and worth taking some time here to answer in detail.

The first:

“If you’re talking a pure technical (trend-following or countertrend based on indicators) system on highly liquid vehicles like the PowerShares QQQ Trust (QQQ), ‘getting a real world system to work’ is insanely difficult….any advice?”

And the second:

“It’s simple enough to create a system that backtests to show gains. It would take me about ten minutes on Tradestation. How do you get a real-world system to work? That’s the $64,000 question.”

I completely agree that anyone can spend a few minutes to get good results on a single asset in a single time frame under a single set of market conditions, and that leads us to the subject of developing a robust system. In order to be robust, a system must meet all of the following criteria:

  1. It must perform well across asset classes
  2. It must perform well with an arbitrary list of stocks, currency pairs, commodities, and so forth
  3. It must perform well in multiple time frames
  4. It must perform well in all market conditions

And that takes more than ten minutes of Tradestation work. It is not easy—insanely difficult, possibly–but definitely not easy. Our primary trading system finds its roots in a paper we wrote over ten years ago, and after major enhancements in 2004, 2007, 2009, and again just a few months ago, we have a system that meets all the criteria of a robust system. But even though the results are good, we work every day to make them even better, while at the same time testing new ideas that are completely different from our existing systems.

I guess you could say we are seeking the “Holy Grail,” but we know we will never find it; so think about this as you start down the road toward building a robust system:

  1. It all starts with an idea. It might derive from your personal trading, or an idea from a blog, or taking an existing system and making it better
  2. The next step is turning your ideas into rules, and the key is to have a platform such as Tradestation, MultiCharts, eSignal, TradeSim, Ninja, or a long list of other platforms that give you the basic tools to build a system and test it and does all the data handing for you so you can focus on just one thing: translating your ideas into a trading system
  3. But then comes the testing. And it is far beyond just seeing if your idea is profitable; it must also be robust and consider drawdowns, win/loss ratios, frequency of trades, and so on. Most importantly, it has to fit your personal style. If you can’t suffer ten losers in a row, then a system that has lots of small losses with a few home runs isn’t going to work for you
  4. There is no grading on the curve here—it is a pass/fail for every test you do. If it fails by any of the criteria you have established right up front, then move on to the next idea. If it passes, then we might have the beginnings of a trading system that could be robust

And that’s what we will be talking about over the next many weeks; how to turn an idea into a set of rules; how to decide which parameters we should test; how to avoid curve fitting; what to look for in a test; how to use stops/targets to improve a system, and many more things that will get you ready to take on the insanely difficult world of trading systems.

It will be fun because we are building our own personal casino one brick at a time, and eventually we will have the house odds on our side of the table.

By Rick Martin, trader, SMB Capital