This paper shows that short horizon stock returns can be predicted to a much greater degree by past price movements than would be anticipated given their low autocorrelation. This raises doubts over the reliability of the autocorrelation statistic as a measure of stock market predictability.

The predictive ability of technical trading rules and the presence of calendar anomalies are well known, but theoretically anomalous, features of equity markets. We show that while some rules exploit calendar effects they are primarily being driven by other factors.