Monday, January 03, 2011
How To Build A Winning Trading System: A Chat With Nelson Freeburg
Dave: Welcome, Nelson.
Nelson: Thanks; it’s a pleasure!
Dave: Let’s start at the beginning. What first got you interested in trading system development?
Nelson: Almost no one in the field of systematic trading started out as a trader. Even some of the best known names in the profession began in different fields. Gerald Appel was a psychiatric social worker. Martin Zweig was a professor of academic finance. The late Bruce Babcock, who popularized dozens of mechanical timing strategies, was an assistant district attorney (who helped prosecute Charles Manson).
Dave: Wow, that's a shocker. I would have never guessed that is the case. Charlie Manson, the market? More similarities than we would like to admit! (laughter all around). Seriously, did researchers from these diverse fields reach any common conclusions?
Nelson: When all of these people eventually turned their attention to money management, they reached a common conclusion. Buy-and-hold investing simply does not work. There may be periods when the stock market goes straight up. But the good times are offset by frequent bear markets.
Dave: How do the systematic traders earn more than simple buy-and-hold strategies?
Nelson: One way to capture the gains and avoid the risks is to use conventional technical analysis—familiar tools like charting, candlesticks, trendline analysis, etc. But there is a lot of subjectivity involved in interpreting chart formations, breakouts, volume patterns, and so forth. This puts a lot of negative psychological pressure on the trader. Furthermore, there is no easy way to back test trading methods that rely on intuition and judgment.
Dave: Correct; technical analysis is generally very subjective, often an art instead of a science.
Nelson: Yes, so the alternative is mechanical timing methods. Explicit trading rules can be reliably tested in great historical depth. If they don’t work, all you’ve lost is time and effort. You don’t lose real money. Furthermore, much of the doubt and fear inherent in subjective trading is eliminated when the entry and exit rules are specified in advance and rest on sound principles of price behavior.
Dave: How did you start being interested in systematic trading?
Nelson: I was working on a Ph.D. in world politics at Columbia University in the late 1970s. My specialty was strategic arms control. I would use formulas and models to simulate the effects of thermonuclear war. (Please understand, I was trying to REDUCE the odds of such catastrophe.) Around that time, gold and silver were on a historic bull run. My cousin, a professional commodity trader, introduced me to the joys and passion of trading. And I was hooked…just like that.
Dave: You were a scientist, I see. Did you give up on the world politics studies?
Nelson: I quit researching world politics and shifted to the study of price behavior. Like many of us, I had my ups and downs as a chartist. Sometimes the chart patterns worked, other times they didn’t. Meanwhile, around this same time, the personal computer made its appearance. It became much easier to develop and test mechanical timing strategies. For ten years I researched computerized timing methods, trading all along. In 1991, I started Formula Research to share my quantitative findings. We started out with a small but discerning group (including Zweig, Appel and Paul Jones). Today we serve institutional money managers and private investors in 27 countries.
Dave: Let's give systematic trading a clear definition. How exactly do you define systematic trading?
Nelson: It’s simply rule-based trading. Investment decisions are based on specified entry and exit conditions defined in advance. With a systematic timing model, if we all have access the same price data, we should generate identical signal histories—a feat not possible with trading by judgment.
Dave: What are the basic components of a trading system?
Nelson: As is often (correctly) pointed out, some of the best trading systems are the simplest. You just need an entry rule to get you in the market and an exit rule to get you out.
Dave: Interesting, particularly in light of all the "quant shops" opening up.
Nelson: By the way, all things being equal, simpler is indeed better. But I would not exclude in principle trading strategies which embrace more nuance and are therefore more complex.
Dave: Are all trading systems basically the same?
Nelson: Most trend-following systems are surprisingly similar in their governing logic. They will feature distinctive stops, filters, and other individual variations But over a test period going back decades, the equity curves will track each other closely.
Dave: What about intraday systems?
Nelson: As for S&P daytrading trading systems, my friends at Futures Truth have tested dozens of intraday models. They note that almost all S&P daytrading systems incorporate an intraday breakout entry (trend-following) as well as a reversal entry (countertrend). In essence, while S&P daytrading system vary in detail, many adhere to this same dual structure.
Dave: That similarity can be very broad. Is the inner logic basically the same also?
Nelson: No, that is where the differences lay. I must have several hundred trading systems in my personal library of timing strategies. In terms of the logic and structure of these systems, there is a wealth of diversity.
Dave: Do system entry and exit rules rely on the same logic?
Nelson: Oh yeah. But the entry and exit rules don’t have to symmetrical. For instance, one of the best long-term stock market timing strategies ever developed (in this case by the distinguished market analyst Martin Pring) uses asymmetrical entry and exit logic. You buy stocks when the S&P 500 is above its 12-month moving average AND the yield on 90-day commercial paper is below its 12-month smoothing. You sell and exit to the money market when EITHER indicator crosses its 12-month average in the opposite direction.
Dave: When developing a system—is back testing a viable method to determine its potential success?
Nelson: Back-testing is not just viable, it’s indispensable.
Dave: Is there specific criteria that you use to analyze whether a system is successful? In other words, what gains should a system produce in relation to drawdowns?
Nelson: Well, the ratio of return to drawdown will vary greatly depending on how much back-testing is done. The greater the extent of historical testing, the lower the gain in relation to the drawdown.
Dave: Should all traders use the same evaluation techniques?
Nelson: Most institutional analysts evaluate an investment strategy by looking at its compound annual return and maximum draw down on a percentage basis. If the strategy offers a higher compound annual return than the S&P 500 while limiting maximum drawdown to, say, 15% of equity, that would be a promising start. By contrast, in commodity trading, most analysts look at the gains and draw down on a dollar basis. Here a good benchmark is to limit draw down to under 10-15% of net profit.
Dave: Are most trading systems trend based?
Nelson: Many successful trading systems are exclusively trend-following. Others are both counter-trend and trend-sensitive. I would say that most trading systems have a trend-following component.
Dave: How specifically do you determine and define trend?
Nelson: Timing models use many different ways to define a trend: Moving average crossovers, percent swing reversals, channel breakouts including Donchian, Bollinger or Keltner Bands; Wilder’s Parabolic and Volatility formulas. Some people even use countertrend indicators like to RSI to identify a trend. For example, you buy when RSI climbs above 50 and sell on a cross below 50.
Dave: Are there inherent flaws that must be dealt with when determining trend?
Nelson: Well, the main weakness of a trend-following strategy is its susceptibility to false signals. With most purely trend-following systems, the percentage of winning trades is 40%-45%.
Dave: Is there away around the issues with trend-based systems?
Nelson: The only way I know of to reduce whipsaws is to add some external filter. A good example is the Pring stock market strategy I described above. You can only go long stocks when the price trend is bullish and the monetary trend is bullish (as represented by lower commercial paper yields). But when you add such a fundamentally-inspired component, it is imperative that your exit be exclusively trend-following. Why? Because eventually that monetary filter is going to fail. (See Japan in the 1990s, the U.S. in the 1930s).
Dave: Recently, several of the huge trend funds have been suffering large drawdowns. Is this implicit in the system OR is the system probably being managed poorly?
Nelson: I believe strongly that these recurring lapses are inevitable when you bet everything on a strictly trend-following model. Periodic losses come with the territory, no matter how well the investment managers execute their strategy.
Dave: What other aspects, other than trend, can a trading system be based on?
Nelson: You can use countertrend strategies to try to anticipate tops and bottoms. In other words, you rely on indicators like stochastics and RSI to identify overbought and oversold conditions. You can also add fundamental, sentiment, intermarket or other indicators external to the actual price data to supplement and reinforce your ability to capture the trend.
Dave: Getting practical, let's design a basic trading system. First, how much data is needed before we start?
Nelson: In a word, you need a representative data sample, one that incorporates strong trends up, weak trends up, strong trends down, weak trends down, and extended periods of congestion. How much data you need will depend on the time frame and the market. A long-term weekly system for institutional stock investors will require data in greater historical depth than an S&P daytrading system that uses 1-minute bars.
Dave: After the data is gathered, what's the first step?
Nelson: You first have to decide whether the system will be general in nature, designed to trade a diverse portfolio of commodities or stocks. Alternatively, you could develop a profitable strategy that only trades one sector, say energy products or stock index futures.
Dave: Let's go over each of those versions.
Nelson: Sure, if you are trading a diverse portfolio of commodities (Option I above), you will probably use a trend-following strategy. You need historical price data and a testing platform that is capable of simulation across a portfolio of markets in dynamic interaction. If your system is designed to trade a single market or sector (Option II above), say the S&P 500 futures, you will need to decide whether you want to add any predictive inputs to complement whatever price-based logic you start out with.
Dave: OK, what's the next step?
Nelson: The next step is build your system using only a restricted sample of the data. Once your method works on this finite segment of price history, you can test it on the out-of-sample data you prudently reserved for confirmation.
Dave: The Monte Carlo simulation model?
Nelson: Yes, exactly.
Dave: Are transaction costs included in the output?
Nelson: Yes, especially in short-term commodity testing.
Dave: What about slippage?
Dave: How does an effective system handle slippage and transaction costs?
Nelson: Slippage and commissions become more significant constraints as the time frame is scaled to progressively lower intervals, which increases trading activity. If you have a long-term institutional stock market strategy that trades three times a year, you don’t have to worry as much about transaction costs. But if you are developing an intraday trading system for the S&P 500 with 30 entries per day, you will find slippage and commissions to be a major, possibly lethal burden. You must find a way to filter out most of the poor trades.
Dave: Are the above simple system guidelines applicable to most systems?
Nelson: Well, the same general principles apply.
Dave: Wrapping this up, what is the most critical aspect of a trading system?
Nelson: The same as your evaluation of any portfolio manager or investment program. You have to minimize drawdown in relation to investment gains.
Dave: How can our members reach you should they be interested in getting started in system trading?
Nelson: We would be happy to send anyone an information package that includes a sample issue of Formula Research and a digest of all of our timing models. The easiest way to receive it is by emailing us at firstname.lastname@example.org. If you want us to physically mail you the package, just call us at (800) 720-1080 begin_of_the_skype_highlighting (800) 720-1080 end_of_the_skype_highlighting or (901) 756-8607 begin_of_the_skype_highlighting (901) 756-8607 end_of_the_skype_highlighting. I have a modest website (www.formularesearch.com), but I’m too busy doing the actual research to focus much on marketing.
Dave: We are almost out of time. Any final thoughts?
Nelson: Well, the key to system building is testing, re-testing and more testing.
Dave: Thank you for your time today!
Nelson: Thank you, Dave!
Nelson Freeburg is editor of FORMULA RESEARCH, a financial letter that develops systematic investment models for stocks and bonds.
When he first came to the financial markets, Nelson was pursuing a Ph.D. in world politics at Columbia University. Wholly taken by the excitement and promise of trading, Nelson said good-bye to the academic life. Soon the markets would give him an education. Despite reading widely in finance, Nelson's investment results fell short of expectation.
Determined to persevere, Nelson began researching the markets full-time. Eventually he would build a financial database that reaches back to the last century--and from this, a library of advanced trading strategies. Today Nelson uses these timing models to advise institutional clients and manage his own investments.
Fifteen years ago Nelson started FORMULA RESEARCH, where he began sharing his findings with a small nucleus of traders. With time the research effort found a wider following. Today FORMULA RESEARCH serves over investors in 27 countries, including many of the leading names in global trading and finance.
Posted by marketsurfer at 4:52 PM