三檔股票交易設計並與傳統配對交易之績效表現比較

Abstract

傳統配對交易策略從股票集合中抽取合適兩檔股票，一做多一放空建構具備市場中性的交易組合，主要目的在於消除市場變動所造成的風險進而賺取與市場大盤趨勢無關的利潤；但這種方法可能因為市場趨勢消除的不夠完全而導致整體交易績效下跌甚至有可能出現虧損的情況，故本論文欲探討是否能夠利用增加一檔股票的方式使得市場趨勢能夠消除的更加完全，並且藉由原本的股票集合中挑選合適的兩檔改成三檔股票的方式，可以取得更多可行的交易配對，進而提高整體獲勝的機率。由於傳統的配對交易能利用股票間的相關性進行初步篩選提高了組合篩選的效率，但三檔股票的選擇若利用彼此間的相關性做為篩選，對於效率上並沒有顯著的提升而且這種相關性也顯得沒有意義；故本論文在三檔股票選股的過程中利用了將股票集合以行業類股做為分類，使原本的股票集合中轉變為類股集合任取三檔，不僅可確保股票間有相當的關聯性存在亦可大幅減少選股時間，並以此做為三檔股票組合的選股方式與傳統的配對交易進行交易績效的比較。

The traditional pairs trading strategies construct market neutral strategies by longing one asset, says a stock, and shorting another to eliminate β, the systematic risk of the portfolio related to the market trend. The resulting portfolio return is expected to be a stationary process and we can make profit by opening the positions when the return diverges from the average level and closing the positions when the return restores. However, market trendsβcan’t be completely removed that make these strategies risky. This paper alleviates this risk by adding a new stock to the portfolio, call it the triple portfolio, to eliminate β more completely and to construct more tradable stock combination to increase/stabilize the overall winning rate. Possible tradable stock pairs can be filtered the by the correlation between stocks prior to to applying formal cointegration test to save computational time. But dealing with all triples combinations without good filtering mechanism like correlations make the computational problem intractable. This thesis classifies the stock set by the Industry Classification Benchmark (ICB) and then chooses three stocks discretionarily from each classification of industry to reduce the number of possible stock combinations. Note that stocks from the same classification are usually highly correlated and are more likely to form tradable triples. We will compare our method with traditional pairs trading strategy to examine their performance.