Nelson Freeburg On Developing The Perfect Trading System, Part II
Last week we covered the first part of our interview with Nelson Freeburg. Click here for Part I of this interview. Here is Part 2:
Freeburg: Here’s an example of how I implement this is my own personal investing: I’ve got three children and I started a trust for my youngest about five or six years ago. I just calculated what the total annual return was for it and it was 11%, which is about 50 basis points above what the S&P produced during the same period. So I wasn’t swinging for the fences, but the drawdown was negligible. And that’s the key thing you’ve got to do. When you’re managing a child’s portfolio, if you take a 50% hit, it might take 12 years to recover from it.
Connors: The five-year performance may be in line with the S&P, but if you take a look at what you’ve done over the past couple of years, you could have just as soon started this money three years ago and the S&P performance would be in the negative right now. You’d greatly outperform it over a three-year period.
Freeburg: Yeah, probably, especially on a risk-adjusted basis. That’s the operational perspective I bring to the task of strategy development.
Connors: So really in a sense then, management of risk is the Holy Grail for you. It’s making sure that you’re minimizing risk at all times.
Freeburg: Yeah, I just did a study. We used one of Mark Boucher’s timing models to trade the S&P going all the way back to 1928. I don’t care what kind of strategy you’ve developed, you’re going to have some rough sailing during the period from 1929 to 1932. That was the most volatile and the most risky period for the stock market in history. My point is, if you were a buy-and-hold investor, you would have suffered an 84% drawdown between 1929 and the market bottom in 1932.
But using Mark’s method — or actually a blend of two of Mark’s models, a bond model and a stock model — the drawdown was reduced to 20%. Now that’s a pretty steep drawdown, but this was a once-a-century event, the Great Depression. If you have a 20% drawdown, you need a 25% gain to recover. If you have an 84% drawdown, you need a 525% gain to recover. This kind of extended historical testing through every conceivable range of price behavior to me reinforces confidence in a timing model.
Connors: When we can, we try to look at strategies to see if they work in other decades. We ask “Did this work in the ’30s? Did this work in the ’50s? Did this work in the ’70s?” You really get a feel if you have a robust system. Do you recommend people doing this? If they have a methodology they’re trading, to try go get data that goes back decades to see if it holds up during those times?
Freeburg: It’s controversial. The knock on that approach is that the character of the markets may change, and the nature of price behavior may evolve over time and therefore if you’re extrapolating from trends that are no longer present, your performance will suffer in real-world application, and that’s a good theoretical point. A contrasting, or countervailing point is if you don’t test your model through very very difficult, very very challenging times of economic turbulence, you could be exposed coming in blind to a horrible collapse in the market that you were just unprepared for.
That’s exactly what’s happened to so many money managers who were children in the ’70s and then reached their professional standing in the ’90s. All they remember is from 1982 on. And they just fell over a cliff with this bear market which, of course, is the second-worst of the last 110 years. And so many people were just unprepared for the catastrophic collapse of the stock market. I’ll give you an example: I’m on the board of the only growth fund of thousands in the entire country that showed a profit in calendar year 2002; it’s the Hussman’s Strategic Growth Fund. My point is that all the other money managers — including people with a lot more experience, and John is still a young man — none of them could make a profit in all of calendar year 2002. It’s that kind of historical testing that prepares you for a wide range of contingencies.
Connors: How do you go about putting together a portfolio to minimize volatility?
Freeburg: Well there are two ways. You can diversify across markets and choose the proverbial non-correlated asset mix, so that you’re not trading all within one family of markets. And second you can actually trade a portfolio of diverse markets with a mix of different trading strategies; you can have several longer-term trend-following systems, you can have shorter-term trend-following systems and you can throw in a sprinkling of countertrend systems. This has been the approach of a number of successful portfolio managers and I think it might be a prudent alternative.
Connors: So they’re trading both breakout and countertrends at the same time?
Freeburg: They’re doing that, and they have different rule-based algorithms in either case and furthermore to reinforce the principle of diversification they apply these strategies to a wide range of markets.
Connors: You didn’t answer this question completely the first time. Is money management the Holy Grail to you?
Freeburg: I have done a lot of testing of systematic money management algorithms, fixed fractional…all of them, and I do think it is critical to smoothly running a portfolio but I don’t endorse the view advocated by some, that money management is more important than trade entry and exit. There are people who say that where you get in and out of a position, whether it’s in stocks and futures, is less important than than money management rules you embrace, in other words, bet size. I don’t really believe that although I certainly agree that position management has a decisive impact on equity curve.
Connors: So the best money managers that you’ve seen out there and the best traders out there have superior entry and exit strategies as opposed to pure money management strategies, is that what you’re saying?
Freeburg: Yes, I am saying that. But you need both. You can’t take a flawed entry and exit logic and make the system work by better money management.
Connors: There’s a fairly well known individual out there who basically says you can flip a coin and make money of you have proper money management. You’ll completely disagree with that?
Freeburg: If there were no transaction costs…maybe. Unlikely, but maybe. But with slippage and commissions, real world constraints and that kind of thing, any money money management algorithm is likely to exacerbate the losses.
Connors: So to you, entry and exit is critical.
Freeburg: I think it’s critical. Money management also critical, but it is not the Holy Grail.
Connors: And then keeping volatility within that portfolio to a minimum is also critical, but that comes from position size, correct?
Freeburg: That’s a value choice. There are other people that are less risk averse than I am. To some degree, you and I are alike. Our personalities call for containing risk. Other people maybe don’t care as much. I don’t ever bet at Las Vegas, for instance.
Connors: You wimp.
Freeburg: Ha. Can’t stand it. The most boring activity in the world. I prefer the stock market.
Connors: Some people trading their own money are more aggressive or less aggressive than if they trade other people’s money. I know for my own money, I’m a heck of a lot more aggressive. We sometimes joke about some of the swings that we take, if we were managing public money there would be no way we’d be taking these type of swings, we would lessen the position size. It really becomes a personality choice. It becomes what the marketplace wants. I don’t think today you can have a money management company that draws down 50%, 60%, 70%.
Freeburg: It’s amazing, I follow all sorts of performance statistics, both for people in the cash stock market and obviously the CTA listing on the Internet, but many of these market timers can’t even beat the S&P over a five-year period even after suffering the losses of the past two years. I don’t know how they stay in business with drawdowns that have ballooned to exactly those proportions. It’s amazing, in the real world, there are practitioners out there with that kind of track record. I’m with you. I would certainly be more prudent in maintaining custody over other people’s assets than with my own account, although as I get older I’m a lot more conservative than I was when I started out trading in the early ’80s.
Connors: What are the top mistakes people make when they go to construct a system?
Freeburg: The first one is unrepresentative testing, where they look at only a selected period of time, or price behavior.
Connors: What does that mean, a selected period of time? What is your definition?
Freeburg: The best example would be somebody in 1985, let’s say, let’s go back 17 years, and they’re just introducing computerized software for testing and so he takes all of the available S&P futures data which goes back 2 1/2 years, let’s say. And basically the S&P went straight up from the time it got started in the spring of 1982 to the time this guy’s developing his system. Let’s make it even better: Let’s say he starts testing in August 1987 and essentially, with the exception of some turbulence in 1984, the stock market went straight up from 1982 to 1987. If he extrapolates from that limited sample of data, he’s going to get killed six weeks later when the market crashes in October of 1987. So that would be one hazard in terms of developing trading strategies. Somewhat similar is letting the computer optimize 1000 different parameters and develop a rule set that perfectly reflects performance in the past with very little capacity to perform well in the future.
Connors: It reminds me when we had a guy working for me in 1994. He used to come up to my house and write code all day. And then he would have the computers run all night looking for the perfect optimization over thousands of different variables. I think my wife and I are still paying off the electric bill he generated from doing this.
Freeburg: Yeah, it’s a common problem that I see all the time. Guys will take the optimized version of these tests and then attempt to trade them. They rarely succeed. The markets are far more complicated and it’s impossible to think you will put a whole number of variables into a computer and the Holy Grail system will pop out.
Connors: Thanks very much, Nelson. This has been great.
Freeburg: My pleasure Larry, thank you.
For a FREE issue of Nelson’s publication, Formula Research, call (800) 720-1080, or for international callers, (901) 756-8607, or e-mail sigma20@midsouth.rr.com.