Trading Study Bus 4502

[pic 1]

TRADING STUDY

BUSI4502

[pic 2]

[pic 3]

Contents

Introduction        

Literature Review        

Synthesis        

Market Efficiency        

How a Filter Works and how optimal filters are selected        

Trading Filter Study        

Stock Choices        

Development Period        

Results        

Testing Period        

Results        

Commentary        

Conclusion        

References        

Appendix        

Development Period        

Test Period        

Introduction

The intent of this study was to empirically test the efficient market hypothesis through a filter test. A high level summary and synthesis of some of the relevant literature is discussed. The six stocks that were randomly selected were: (a) Amazon; (b) Yahoo; (c) Berkshire Hathaway Class B; (d) Priceline; (e) Biogen and; (f) Facebook. The criteria for selection included stocks that were devoid of stock splits, dividends, or other events that could result in adjustments. Each stock in the filter analysis was evaluated using increments of 0.5 – 8. The sample period was one and a half years of daily data. This was compared to the buy and hold returns over the same period.  The filter with the highest returns was selected for use during the testing period of one and half years. The aim was to discover if the use of a filter could earn greater returns than a passive buy and hold strategy of the same stocks. If this turned out to be the case, it would help to chip away at the efficient market hypothesis.

Literature Review

Cheng-Lung Huang, Cheng-Yi Tsai, (2009) main objective is to explore filter-based feature selections to choose important input attributes in able to improve prediction accuracies for the financial daily stock index predictions. This study conducts an experimental data test by collecting a sample from the FITX data set, which is a daily index prediction for stocks and future markets and runs a regression against three different filter variables. The FITX data set collected covered a six-year period, pairs of daily observations correlated against three variables that are MSE, MAE, and MAPE; other variables such as RSI, MACD, MA, and PSY were analyzed but never tested against the hypothesis. The authors concluded that the regression resulted in an overall improvement in the average predictions accuracy and training time.