台灣加權股價指數型態辨認演算法交易獲利性研究

Abstract

技術分析廣泛的被投資人採用多年，而其中形態學因其易於肉眼辨認，無須繁雜運算等特性，而成為最主流的技術分析理論，相對於如此大量的採用而對形態學中如等幅擺盪等獲利的特性卻未有人以客觀的方式加以檢驗，本文採用Lo, Mamaysky, and Wang (2000)中的研究方法，以非參數核回歸作為基礎建形態辨認演算法，針對「雙重頂」，就型態定義、進出場規則等加以改進，另外針對演算法中的參數如:寬帶(Bandwidth)、形態形成時間等進行分析。實證結果顯示，在特定的參數設定之下，台灣市場展現出明顯的超額報酬獲利性，另外在不同參數下，也確認了等幅擺盪發生機率非常高。

Technical analysis, especially “charting”, has been a part of financial practice for many decades, but this discipline didn’t receive the same level of academic scrutiny and acceptance. In this paper, we impose the pattern recognition algorithm of Lo, Mamaysky, and Wang (2000) with several modifications to determine whether ‘‘Double Top’’ (DTOP) price pattern has profitability for TAIEX returns. Which modification include the implementation of typical filters and treats which technical analyst considered as common sense. With data from TAIEX over the period 2005-2014, we find support for the profitability of our own revised DTOP strategy under some parameter settings. On the other hand, we also find there is obvious presence of equal amplitude after breaking the neck line under all parameter settings.