The Sensitivity of Moving Average Trading Rules Performance with Respect to Methodological Assumptions

AbstractThe question whether the tools and methods characterizing technical analysis can lead to superior performances when it comes to forecast the future evolution of speculative prices (mainly stock prices) represents a matter of permanent debate in the specialized literature. Various surveys show long term equilibrium between the number of empirical studies that confirm these superior capacities and those failing to do so. In the same time, it is suggested that a possible explanation for these heterogeneity resides in the various methodological approaches when testing technical analysis rules. It seems that the subjective choices of the parameters can bias substantially any empirical results through the negative effects of data mining. The paper's objective is to test the sensitivity of the returns generated through the use of the dual moving average crossover rule with respect to the dimension of the investment decision space and the moment in which the investors do the actual trading suggested by the rule. Based on a price of 2,942 observations characterizing the Bucharest Stock Exchange, we ran simulations for 5,670 different trading rules and we found that particular approaches of the trading rule consistently leads to different results, approaches which are more or less prone to boost the rule's profitability. For example, using a two element space of investment decision will always outperform a three element one mainly because of the heavy trading commission fees generated by more frequently changing the investment exposures. We also considered the impact of the trading commissions when evaluating the total returns and that of the net interest that can be earned through borrowing and lending at a differentiated cash rate (namely ROBOR-3M and ROBID-3M)