Performance of Stock Market Prediction – A study on prediction accuracy and realised return

Abstract

Traditional portfolio theory stated that diversified portfolio is optimised regarding returns. It can generate the highest return with relatively lowest risk. Market risk cannot be diversified, so the most intelligent approach is to buy and hold assets in the long run. Therefore, the market return is treated as the benchmark return. Return added in advance to the benchmark return is alpha return. As such, the objective of trading actively is to beat the benchmark, i.e. throughout trading, alpha return is expected. Otherwise, active trading does not make any sense. Active trading is based on prediction, whether by fundamental analysis, by technical analysis or by applying the principles of Chinese feng shui. This study found that with a certain level of accuracy of prediction, it can achieve alpha return. It suggests a model for estimating the level of accuracy. Also, the model is enhanced by including transaction cost. Some implications or corollary can be concluded by this study. Firstly, a model is proposed which can be treated as a baseline for formulating the required accuracy of different computer aided prediction models such as SVN, Neural Network, and GARCH. In other words, to examine if one prediction model works, we can examine the required level of accuracy by the proposed model and then compare the required accuracy and prediction accuracy. Secondly, there are several corollaries on market behaviour the evidence of which supports the finding that most active traders are unable to exploit market opportunities, i.e. to buy and sell better and make more buys in a bull market.