Monday, February 2, 2009

Why Trading Systems Don't Work

Years ago, I built a trading system. For years, it worked very well, beating the market by significant amounts before the current market turmoil which began in late 2007. In 2008, the system stopped working and I lost a lot of money (though I didn't fully "blow up"). As a result of losses greater than any period I had back-tested or forward-tested, I no longer have confidence in my system.

Even when my system was generating big profits, I often questioned how one can be sure that his system will continue to work. Unfortunately, I ignored these doubts. But now that these doubts have been confirmed, I believe that it is either impossible or extremely unlikely to develop a winning trading system.

I won't be looking at problems with trading systems, such as curve fitting, but "problems" with the market that kill trading systems

Backtest Period

I used to laugh at people who backtested just 10 years. I had backtested my systems 50+ years so I could include a number of bulls and bears. But the decline of 2008 was nothing like anything in the past 50 years, though the 1973-74 bear market is a close similarity. And the double dip of 2000-2003 and 2008, where the market fell 50%, recovered the losses, and immediately lost 50% again is definitely something we hadn't seen since 1929-37. But even if I backtested back that far, that would only include one such period. In order to model the market, I'd need multiple similar periods to create rules to trade by. To include many market panics, bears, and depression, I'd have to backtest well into the 1800s. But economic and financial data from that period is sparse, making it impossible to develop a trading model that will work in the long run.

Survivorship Bias

My trading system was based on the US stock and bond markets. Why? Because that's where I live and that's the biggest market. But what if I had lived in England in 1914? England was still the financial capital of the world at that point, but was about to lose that status. Or what if I lived in Japan in 1989 which was growing like crazy but about to stagnate for an entire generation. Or Argentina or Southeast Asia before they blew up? You can't select one country that happened to thrive and ignore the losers because you don't know which will thrive and which will die in the future.

This problem could be solved by trading in many countries around the world. If your system is correct and goes long the best and short the worst, you would actually benefit from the fact that many countries don't survive. But trading foreign markets was not really possible until modern telecommunications became affordable a couple of decades ago. Even so, trading overseas is often expensive, difficult, or impossible if the country restricts foreign ownership of stock. Over the past 20 years, China has certainly been the top performing country, but for many years foreign ownership was illegal and even now it is restricted.

Changing Rules

Creating a trading system based on the last 50 years of data implies that the rules of the last 50 years would stay the same. That is obviously not true. Over the last 50 years, interest rates never went negative, until they hit zero and actually went negative in 2008.

Statistical probability states a market is so unlikely to move 10 standard deviation that we should simply ignore that possibility. At least, that is what we were told. But then the market did exactly that in 1987. Long Term Capital Management also suffered one of those "hundred year events" in 1998. The "black swans" occur much more frequently than standard statistics assumes.

Besides market rule changes, there are also legal changes. In 1933, FDR confiscated Americans' gold bullion and broke any contract based on gold instead of the dollar. After World War II, the dollar was again fixed to gold until Nixon took us off that quasi-gold standard. As a result, currencies floated against each other and currency trading boomed. These legal changes had great impact on how markets acted and how trader could profit. Old systems stopped working and new ones must be developed.

Maybe the biggest rule change of all was the closure of the stock market for 4 months after the breakout of World War I in 1914. A profitable trader would have lost four months of income and had his capital tied up. How many traders today would be able to survive an extended market closure, even though that possibility is so unlikely with electronic trading.

And then there is government manipulation. The power to set interest rates, most often lower than they should be. The power to print money. Selling gold from our reserves in order to push down its price. Wars to gain control of natural resources. Repurchase Agreements and Reverse Repurchase Agreement to temporarily move the stock market.

What is to stop the government from shutting down the stock market or making gold bullion or any other investment illegal? How will the government manipulate the market today or this year? Your system, which is always correct, could say one thing and that thing would have occurred if the government didn't manipulate the market against you.

When one looks at 2008, the rules have certainly changed. Interest rates are being set at artificially low rates of 0% for the Treasury Bill and about 2% for the Bond. Banks are being nationalized. The federal government's spending and budget deficit is rising to record peace-time levels. No trading system could have predicted these events and the effect they would have on the market.


Long-term statistical trading systems don't work. As Jesse Livermore/Larry Livingston says in Edwin Lefevre's Reminiscences Of A Stock Operator, "A man may beat a stock or a group at a certain time, but no man living can beat the stock market! A man may make money out of individual deals in cotton or grain, but no man can beat the cotton market or the grain market." A trading system may make money for a while, but no system can "beat" the market. As a result, long term systems are a waste of time and short term system must be kept on a short leash, trading them while they work but being prepared for them to stop working at a moment's notice.

Frank Shostak concludes his "What is Wrong With Econometrics?" with: "Rather than viewing econometric models as a sophisticated technique that can discover the hidden truth about the economy, we should regard them as clumsy and expensive extrapolative devices, which have nothing in common with reality. Anyone who decides to use models as an analytical tool or a forecasting device runs the risk of seriously confusing himself." In the same article, he cites Mises, "As a method of economic analysis econometrics is a childish play with figures that does not contribute anything to the elucidation of the problems of economic reality." The same, I believe, would be true about trading systems.


If you want to trade the market, you have to look beyond the statistics and throw out what has worked in the recent past. You need to study history going much further back. Study the Great Depression and its causes. The Panic of 1907. The many panics of the 1800s. The history of previous empires and their falls. Most importantly, study Austrian economics. With Austrian economics, you will see how the government caused our current crisis. And you will learn how the current government intervention is doing more harm than good. And you will learn how capitalism, if it is allowed to work, will lead to further prosperity.