A Study of Profitability of Technical Trading Rules in Global Futures Markets

• Main Authors: Wan-LinHaung, 黃婉玲
• Other Authors: Meng-Feng Yen
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/32836845070348019367

碩士 === 國立成功大學 === 財務金融研究所 === 99 === This paper discusses Profitability of Technical Trading Rules in Global Futures It has been a disputation for a long time in the past.Based on Hansen’s SPA test (test for superior predictive ability), we examine the profitability of a universe of 20,970technical trading rules, which are selected from previous studies, in twenty twenty futures markets The transaction cost and the margin are taken into account for a practical manner. We also imitate the measure used by Marshall, Cahan, and Cahan (2008) for researching the whole interval and two equal sub-periods.We use SPA test with two-side to examine whether the best trading rule can significantly beat the benchmark of the risk-free rate in the full period, as well as in the first sub-period and the second sub-period. The results of the SPA show that the best trading rules are superior to the benchmark for almost all the futures markets; however, the outperformance vanishes after adjusting the data snooping bias, providing a consistency with the previous study of Marshall, Cahan, and Cahan (2008). The SPA test identifies only one statistically outperforming technical trading rule in the S＆P 500 futures market during the full period. Containing numerous poor models (technical trading rules) may weaken its testing power, we adopt 7,846 technical trading rules used by STW to test the first sub-period and the second sub-period. During the first sub-period, three statistically outperforming ones are discovered in the Euro FX, FT-100 and TW futures markets. During the second sub-period, four statistically outperforming ones are found in the Stoxx-50, FT-100, and S＆P 500 futures markets. The results of the SPA show that the worst trading rules are inferior to the benchmark for all the futures markets; however, the underformance exists after adjusting the data snooping bias. The fours values of smooth parameter we conduct for the geometric distribution for the stationary bootstrap block size of the SPA test make p-value differ from each other, causing some influence for analysis. Adopting more values of this smooth parameter comparing to the previous works hence enable this paper more complete.