The main purpose of this paper is to examine the predictability of daily, weekly and monthly stock returns in ten emerging markets based on out-of-sample forecasts. We use the Variance Ratio test and other models in our study. However, the Variance Ratio test does not provide clear evidence of stock return predictability. Application of ARMA (Auto Regressive Moving Average), GARCH-M (Generalized Auto Regressive Conditional Heteroscedasticity-in- Mean) and ANN (Artificial Neural Network) models provides strong evidence of stock return predictability. Different models are found to be superior to others in different emerging markets under different measurement intervals. Predictability of stock returns has been found to enable investors to increase the accuracy in predicting direction of change in stock prices and significantly enhance their returns over a weekly horizon. The success of the ANN model in forecasting in majority of the markets suggests that prices in these markets may be captured by non-linear processes.