DETECTING PATTERNS IN STOCK RETURNS USING ALGORITHMIC INFORMATION THEORY

Abstract

The purpose of this study was to develop a rule for the time of trading an stock exchange asset based on an indicator of degree of randomness of a sequence of characters. It was applied the Lempel and Ziv standardized complexity measure and a measure of relative efficiency of the market, based on algorithmic information theory, to high frequency data of returns for different periods of negotiation. The results were obtained for an asset, in which it was possible to determine the set of periods in which there might be more likely to detect predictability in returns. We were able to describe, in probabilistic terms, in which intervals may occur standards for a particular asset. The measure defines a standard, but that suffers from statistical fluctuations can be, in some observations, wrong despite a 1% chance of its occurrence.