THE SPECULATOR -- "Don’t Bet The Rent On Technical Analysis: Common Sense Sits Head And Shoulders Above Even The Most Widely Accepted Charting Strategies. Here Are The Numbers To Prove It"
By Victor Niederhoffer and Laurel Kenner
In times of great disaster and stress, we like to think of great ship commanders and generals heading toward their goals with cool calculation while the guns are blazing. Therefore, rather than succumb to the temptation to join in the general angst as Americans watched their lives pass before their financial eyes on Tuesday, we tried to stay calm in order to figure out how to survive and make our next buck. Among other things, we were thinking about a post that came across our monitors two weeks from a hedge fund manager whom we call Miss X. “The technical action in the S&P since the big-cap rally last week is being threatened by a massive head-and-shoulders going back to January 1998,” she wrote. Since then, the S&P took out the January 1998 low and headed straight down to April 1997 levels. On Tuesday, it broke through 800.
Our correspondents point out that if one uses the classic Edwards-Magee head-and-shoulders method to project the “downside target” for S&P futures, 540 or so is the likely next stop. As for the Nasdaq 100, it’s headed for negative 2,690 under the same method of reckoning. Does the apparent success thus far of Miss X’s gigantic pattern mean that head-and-shoulders trading works? More to the point, how does knowing this help right now, as the foundations of the financial world appear to be crumbling?
Worthless Nostrums And Superstitions
First of all, it always helps to know what works and what doesn’t. In times of plague, worthless nostrums and superstitions flourish. It’s important to keep a clear head and take the long view. As to whether head-and-shoulders patterns work well enough in the stock market to be profitable -- they don’t, as we’ll show. Head-and-shoulders trading has been around since before 1930. The pattern consists of three peaks, the highest being in the middle. A horizontal line -- the “neckline” -- is drawn to connect the troughs between the shoulders and the heads. The crossing of the neckline is supposed to signal that prices will continue down away from the head. An inverted pattern is read as bullish.
45 Years Of Testing
We’ll start by saying that we avoid all vaguely defined “technical” indicators requiring visual evaluation by a gifted interpreter. Moreover, in 45 years of trading, Vic and his staff members have tested every indicator to which value is ascribed. If it works, he would be using it. Head-and-shoulders trading is a trend-following strategy -- and as we have explained, “the trend is not your friend.” We believe that any wealth that accrued to Miss X over the past two weeks as a result of trading this pattern -- and we do congratulate her for it -- was a lucky event, not a sure indicator of future success. However, any important question deserves to be tested, and the results made public. Haphazard anecdotes, confident assertions and appeals to authority -- even, or especially, our own authority -- will not do. As Steve Stigler writes in his magisterial “Statistics on the Table”:
"If a serious question has been raised, whether it be in science or society, then it is not enough merely to assert an answer. Evidence must be provided, and that evidence should be accompanied by an assessment of its own reliability. We have tested the head and shoulders strategy on S&P 500 futures and will report the results below. But we also will take the opportunity of passing along to our readers the results of a remarkable study by Carol Osler of the New York Federal Reserve that concluded head-and-shoulders trading in individual equities is, on balance, unprofitable."
Osler’s tests were rigorous, and she presented her conclusions with great clarity in a study called “Identifying Noise Traders: The Head-and-Shoulders Patterns in U.S. Equities.” Osler wrote a computer program to identify head-and-shoulders patterns, based on the descriptions in eight technical manuals. She applied it to 100 companies with price data spanning July 2, 1962, to Dec. 31, 1993, selected at random from the Center for Research on Securities Prices at the University of Chicago. All of the manuals were ambiguous about the criteria for exit, but they agreed that a head-and-shoulders pattern signified a major change of trend. Osler therefore wrote her program to require that a position be held until the price stops moving in the predicted direction, with a stop loss of 1%.
Irrational Speculation
Her conclusion: Head-and-shoulders trading is unprofitable and “does not qualify as rational speculation.” Amazingly, head-and-shoulders trading is quite popular. Osler estimates that such trades account for as much as one-quarter of an average day’s volume around the time of the “neckline crossing,” the signal to put on a trade. How to account for the popularity of an unprofitable trading strategy? Certain peculiarities of the human mind may account for its acceptance, Osler says. People are prone to see nonexistent connections between groups of things. They tend to be overconfident in their own judgment. And they remember pleasant or successful experiences (e.g., profitable head-and-shoulders trades) with far greater clarity than they do unpleasant experiences.
A few successes may bring the head and shoulders trader fame and funds, encouraging new entrants. As Osler notes, cognitive psychologists have shown through experiments that beliefs and behaviors are difficult to “extinguish” when they are randomly reinforced. Osler’s data ends in 1993, and we wondered whether the picture might since have changed. Fortunately, we were able to interview her earlier this month. She told us that she is updating the database. We have heard many market players say they don’t believe in head-and-shoulders trading or any other technical analysis patterns, but like to know what such traders are doing so they can eat their lunch. Not likely, says Osler. Even before transactions costs, trading against head and shoulders traders is just not profitable in individual stocks.
Next month, Osler will leave the Fed to take a professorship in the international economics and finance department at Brandeis, where she plans to research the role of stop-loss and take-profit orders in the currency market. One area of interest is the possible clustering of orders at round numbers. She doesn’t have any plans to extend her head-and-shoulders work to S&P futures. “Technical analysis is just not a hot topic,” she explained. “It’s not the sort of thing you can make a big splash with.”
Follow The Algorithms
Always ready to jump into the breach, we tested the head and shoulders strategy on S&P futures prices beginning in 1996. Shi Zhang, Vic’s computation assistant, wrote a program based on Osler’s description of her head-and-shoulders algorithm. Zhang’s program looked for six points on a bearish head and shoulders pattern -- the right shoulder, the right trough, the head, the left trough, the left shoulder and “neckline” crossing. The time between points had to be at least five days and no longer than 180 days. Various distances were used to make sure that no patterns were missed. After running his computer nonstop for 10 hours, Zhang came up with seven patterns, which he then evaluated for profitability 1, 2, 3, 4, 5, 10, 20, 30 and 60 days out. The results appear below:
Number of days after neckline crossing -- 1 2 3 4 5 10 20 30 60 Average profit/loss -- 0.3% 1.6% 1.2% 1.1% 1.3% -0.7% 0.7% -0.2% -2.6% Standard Deviation -- 0.01 0.03 0.01 0.02 0.02 0.02 0.02 0.04 0.1
We conclude that head-and-shoulders trading does not work either as a signal of a trend change or as a profitable strategy. We will therefore go back to picking up good stocks at prices unseen in five years, and we’ll try to restrain ourselves from using the rent money.
Study on pattern: http://www.ny.frb.org/rmaghome/staff_rp/sr42.pdf
