A behavioral view of the head-and-shoulders technical analysis pattern

The technical analysis approach to predicting stock returns is based on the identification of recurrent patterns in the way stock prices evolve in time, (a) using optical examination of price history, and (b) a host of technical indicators. Optical examination of price history materialises as the recognition of certain patterns that supposedly have predictive power. “Head-and-Shoulders” (H&S) is a well known technical price pattern, which is said to occur seldom but with high predictive power. In this paper we examine whether or not the H&S pattern can be identified in price series generated stochastically. A rule-based mechanism for the identification of the pattern is applied on half a million price series generated stochastically with the use of the Geometric Brownian Motion and with different combinations of volatility and drift rate. We find that the H&S pattern can be identified almost 11 times out of 100 in randomly generated price series. However, the expected payoff for a short-long strategy including transaction costs is negative (as expected). After examining the characteristics of the pattern (frequency of occurrence, profitability, and so on) we look with a behavioral view the results of our simulation. The question we bring into discussion is whether investors selectively focus on specific profitable cases of the pattern. This results in the misleading conclusion that the pattern is seldom observed but profitable. According to our findings it is possible that the pattern is being identified more often having no predictive power at all. Various cognitive biases affect the way investors make decisions and result in the aforementioned misleading conclusion. This is confirmed by the answers that graduate students from the University of Macedonia gave at a questionnaire which was set based on the results of our simulation.