Custom Search

Technical analysis has been severely criticized for reasons other than imprecise definitions and inadequate historical testing. Even the weak form of the efficient market hypothesis (EMH) states that past market prices provide no information about the future prices which would allow a short-term trader to earn a return above what could be attained with a naive buy-and-hold strategy.

A data series is considered to be random precisely because it does not exhibit a pattern or structure. Obviously the very concept that patterns do not exist in the data contradicts the foundations of both technical analysis and pattern recognition decision rules when applied to price series (as in the upcoming gold market example).

Weak EMH strictures do not apply to pattern recognition decision rules when the indicator series are different and not derived from the series for which the identification of patterns is being tested. For example, a pattern recognition decision rule model which used interest rates as the indicator series for identifying patterns in the stock market would not contradict the weak version of the efficient market hypothesis, but the postulate that short-term patterns exist in the stock market does.

Concerning the existence of short-term patterns in financial time series, the key word for the weak EMH theory is short-term. The weak EMH theory does not deny the existence of long-term patterns and trends such as those associated with the business cycle5. This fact is often overlooked in discussions of the efficient market hypothesis. For example, long-term, cyclical trend changes in the Standard & Poor's 500 Stocks Average have been observed to be one of the most reliable leading indicators of the gross national product (GNP). Thus, if the existence of cyclical patterns and trends can be defined in a series, both technical analysis procedures and pattern recognition decision rules can be used to identify them. This article will deal with such long-term patterns.

A technical analysis procedure that is developed and tested with charts for only one business cycle has a very weak foundation, but a set of pattern recognition rules tested successfully over 27 business cycles represents a significant empirical finding. This leaves open the question as to what constitutes the short-term and long-term. As a result, the time horizon limits of the short-term conditions in the weak EMH are undefined. Because the duration of business cycles is sufficiently long to be long-term under EMH theory, there is at least the feasibility of developing pattern recognition decisions that identify trends associated with the business cycle.