Chaos Theory and Market Reality, Part Two

Chaos Theory and Market Reality, Part Two
In my last article I discussed the natural progression of learning for a new trader. He or she quickly determines that in order to be successful, one must master market predicting. After reading some books in the conventional literature, he attempts to find repetitive market patterns and cycles using price bars or mathematical indicators. He may fall prey to various expensive system promotions. In spite of the abundance of such prediction methods in books, systems and software, in the long run, probably 95 percent of traders lose. Nevertheless, almost no traders question the proposition that exploitable, repetitive price patterns and cycles exist.
People are naturally susceptible to wishful thinking. They believe what they want to believe in spite of obvious evidence to the contrary. Short-term luck causes many such faithful traders to reinforce their invalid beliefs. Unsuccessful traders have a distorted view of the markets, themselves and what they are really doing when they trade. It is very difficult for them to shed these misconceptions so they are doomed to long-term failure.
It turns out that it is possible to examine historical market price action with mathematical and statistical tools and determine whether such repetitive patterns and cycles exist. Chaos Theory is the mathematics of analyzing systems such as market price action.
For those willing to plod through some fairly technical, jargon-loaded language, I recommend Edgar Peters' two books on Chaos Theory and the markets, Chaos and Order in the Capital Markets (992) and Fractal Market Analysis (1994). Both are published by John Wiley & Sons.
Chaos analysis tells us that market prices are highly random with a trend component. The amount of the trend component varies from market to market and from time frame to time frame. Short-term patterns and repetitive short-term cycles with predictive value do not exist. The patterns of prices and indicators traders use to predict occur as readily in random data. Thus, you have about as much chance to predict short-term market prices using technical analysis as you do to predict future numbers on a roulette wheel.
In writing his second book Edgar Peters examined four years of tick data in the S&P. He concluded that while short-term data is not totally random, the deterministic element is so small as to be barely measurable. He concluded that "it is highly unlikely that a high-frequency [short-term] trader can actually profit in the long term." He also found that there are no cycles in intraday data.
As I read the literature, this is not opinion. It is scientific fact. Traders who ignore it do so at their financial peril. Does this mean the markets are a random walk and that eventually all traders will lose because of the costs of trading? No.
Traders can exploit the longer-term trend component of commodity market price action to obtain a statistical edge. This is precisely what trend-following systems do. It explains why good trend-following systems traded in diversified market portfolios tend to make money year after year while day-traders invariably lose in the long term. To be a successful speculator, you must put yourself in the same position as the house in casino gambling. On every bet the house has a statistical edge. While the house may lose in the short term, the more gamblers bet, the more the house will eventually win. If you trade with an approach that has a statistical edge and if you follow your approach rigorously (a big if), like the casino, you cannot lose in the long term.
My calculation of the trading success quotient is that one-third depends on the system, one-third on the portfolio of markets traded and one-third on the trader's discipline to follow the system precisely. We can never know for sure whether our system has a statistical edge. The best we can do is create it without over-curvefitting and test it historically. If the system has been over-curvefitted, any historical testing will be pre-ordained and worthless. A simple way to guard against over-curvefitting is to use exactly the same rules for all markets and test your system on as many markets as possible. If it is profitable in a wide variety of markets in a long historical test that generates a large number of trades, it is probably not over-curvefitted.
To maximize your edge while minimizing your risk, it is crucial to select an optimal portfolio for your system and account size. Since a market's price trend component is what gives you a statistical edge in the first place, you can increase your edge by concentrating your trading in markets with the highest historical trend component. I have written a book called Trendiness in the Futures Markets which is a systematic examination of the tendency of 29 popular markets to trend in all time frames between 5 and 85 days.
While we can measure the various markets' tendency to trend in history, we cannot know for sure which markets will trend the most in the next six months to a year. Thus, to reduce short-term risk, we must diversify as well as concentrate. My research has shown that optimum portfolios for trend-following systems have between 10 and 20 markets. I personally trade 19 different markets with various trend-following systems.
Another aspect of managing risk is keeping drawdown in relation to account size under control. Because drawdown and profitability are closely related, to reduce drawdown, you must accept smaller profits. It is impossible to evaluate the optimum portfolio for your system unless you compute the joint historical drawdown when trading the entire portfolio. I suggest a starting account size representing twice the maximum historical portfolio drawdown plus the total margin for the portfolio.
Contrast this rigorous, scientific approach to the way most traders operate. They have no idea whether their methodology has a statistical edge. They assume that if it is in a book or came from a famous guru or cost a lot of money, it must be good. They trade with highly subjective methods that can never be tested. They are too lazy to create some hard and fast rules and perform proper historical testing. If this is you, don't be surprised if your trading produces losses. For you, trading may be fun, but you will pay for your entertainment.