Pang模型(2009)下的股價預測與配對交易之實證研究─以台灣市場為例

Abstract

本篇論文旨在探討當市場現象不符合股價預測模型假設時，能否運用其他方法達到預測的目的。並進一步希望藉由得以預測股價的優勢，找出從市場上獲利的方法，因此模擬回測配對交易的績效。回測後，發現若以台積電、聯電為投資組合，並佐以買進台積電、賣出聯電且每半年結算一次的方式，將可以得到夏普指數為0.5115、勝率高達78.26%的成效。這結果顯示了在長期多頭、空頭市場交錯下，運用此策略仍能獲利。

The main purpose of this paper is to see if there’s another way to predict the stock price when the usual assumptions for the forecast model are no longer applied to the market. Once we have a better ability to forecast, finding opportunities to profit from the market would be our goal; thus, we did some back-testing. From the result we had, if TSMC and UMC were in our portfolio, the best strategy would be buying TSMC and selling UMC with a holding period of half a year, which would give us a sharpe ratio of 0.5115 and 78.26% of winning chances. This showed that no matter which kind of market, bull or bear, we were facing, profiting from the market was not out of reach.