Keith Fitschen, best known for his Aberration commodities trading system, has done every would-be mechanical trader a tremendous service by writing Building Reliable Trading Systems: Tradable Strategies That Perform as They Backtest and Meet Your Risk-Reward Goals (Wiley, 2013). It is clear, thorough, and, perhaps best of all, provocative. With it also comes access to the TradeStation Easy Language code and daily signals for the systems developed in the book.
Fitschen has been developing trading systems for more than 25 years and is fully aware of their potential pitfalls. Among other things, they can perform brilliantly in back-tests only to fail because they are overly curve-fit, although Fitschen is quick to point out that “there’s always a certain element of curve-fitting because we don’t have an infinite data set and we’re developing a solution based on historical data.” (p. 23) They can work well in one market and bomb in another. In fact, in this book the author uses a Donchian baseline entry signal for both stocks and commodities—but “a counter-trend Donchian entry for stocks and a trend-following Donchian entry for commodities.” (p. 65) Moreover, trading systems normally view the world in limiting, self-constraining black and white: “you can only enter when the entry logic conditions are fully met, and only exit when an exit logic condition is fully met. There are no shades of gray like: this is a weak entry; this is a very strong entry; or this is a good entry even though one logic condition isn’t met.” (p. 119)
To help counter the curve-fitting problem, the author developed a method called “Build, Rebuild, and Compare,” or BRAC, where the second step uses part of the initial historical database. Why not out-of-sample testing? “The main problem with out-of-sample testing is there is no performance reference for the out-of-sample results.” (p. 24)
To address the black-or-white problem, Fitschen introduces the idea of bar-scoring as an adjunct to fairly good entries and exits. For instance, he uses one standard deviation below the 10-day average as baseline entry criteria for stocks. This yields 595,806 trades, which averaged 0.288 percent per trade. To this baseline he adds five criteria—two price measures, a volume measure, a volatility measure, and an eight-bin breakdown of bar type. He uses 20 bins for his bar-scoring criteria. Without going into any of the details, the upshot is that bar-scoring allows the trader to significantly outperform the basic strategy. Of course, this outperformance must be verified using BRAC.
Fitschen tests (and thereby challenges) some widely accepted claims, such as “always look to fade the gap,” “always trade in the direction of the trend,” “a divergence is a strong signal,” and “trade Fibonacci retracements.” He shows, for instance, that “divergence trading doesn’t have a significant edge in either stock or commodity trading; in fact, for commodities you’d be much better off trading in the opposite direction on each divergence signal.” (p. 137) To Fibonacci buffs he says, “I think we can conclude that with very noisy data, like stock market data, numbers are just numbers; there are no magic ones.” (p. 147)
The money management section of this book is particularly strong. The author distinguishes between large ($100,000+) and small ($20,000-$100,000) accounts and develops money management techniques for each—both commodity and stock strategies. As for the distinction between large and small accounts, he explains: “A small account is … defined as one that cannot be traded risking a small fixed percentage of equity on each trade. … Small-account traders are forced to take greater relative risk than large-account traders. Small-account traders are always a small number of adverse trades away from a margin call, while large-account traders who risk a small percentage of equity on each trade are always a very large number of adverse trades away from a margin call. … It should be the aim of every small-account trader to grow his equity to large-account status so he can enjoy the benefits of large-account trading. Besides walking in from the edge, the large-account trader enjoys money management strategies not available to the small-account trader.” (p. 183)
Fitschen gives the small-account trader ammunition to help increase the size of his account and offers the large-account trader systematic skills to avoid lapsing back into small-account trader status. There are no magic bullets, just as there are no magic numbers. What is required is a lot of hypothesis creation and careful testing to devise systems that match your risk/reward profile. Building Reliable Trading Systems is an excellent place to start.