Detecting a Stock Market Anomaly with a Classifier System

This paper presents a classifier system to detect stock market anomalies. The classifier system groups the last 15 years' daily data of the ISE100 Index of the Istanbul Securities Exchange into classes of fixed size, and computes for every observation in each class the return over the succeding T days. Next, the average return in each class is calculated. Confidence intervals for average returns are constructed using bootstrap re-sampling method. It is observed that for an investment period of at least one year, average classified returns becomes positive at all levels of the ISE100 Index, strengthening the now well-known assertion that the ISE is not weak form efficient.