Technical Analysis, Genetic Algorithms, and Data Snooping : An Empirical Study of Taiwan Stock Market

• Main Authors: Chien-Kuang Chen, 陳建光
• Other Authors: Chin-Sheng Huang
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/17018885455000091590

碩士 === 國立雲林科技大學 === 企業管理系碩士班 === 89 === This study investigates the forecast power of technical trading rules in Taiwan stock markets. The technical trading rules of Brock, Lakonishok, and LeBaron (BLL, 1992) are used as a benchmark to compare with the profitability of genetic algorithm (GA) trading rules. For taking account of data snooping effects, this study employs White’s Bootstrap Reality Check to test the existence of superior rule. Meanwhile, this study conducts break-even cost analysis to ascertain the profit threshold for each trading rule. By control non-synchronous trading effect, this study also calculates one-day-delay strategy and reviews its impact on trading rules. The testing data includes Taiwan stock exchange (60/01/05-89/12/30), Taisddaq (84/11/03-89/12/30), and Taiex futures (87/07/21-89/12/21). The main empirical findings are as follows: 1. GA rules are in general better than BLL’s in terms of usual tests for all testing data. 2. White’s Reality Check indicates that there do exist some superior rules that possess useful economic content. 3. Break-even transaction cost analysis shows most technical trading rules have higher threshold cost than the actual market transaction costs. 4. Non-synchronous trading does have tremendous effect on the profitability of technical trading rules and can cause spurious conclusions. Namely, it affects both of usual tests and White’s Reality Check. It reduces threshold cost of trading rules significantly as well. This study reaches the conclusions that special care should be taken to consider the market institutions when test the technical trading rules.