模組化交易平台與機械化避險操作研究：以台股指數期貨為例

Module Trading Signal for Robot Hedging Trading: The Case of Taiex Index Futures

張鶴霖(Ho-Lin Chang)

指導教授 ： 李存修

國立臺灣大學/管理學院/財務金融組/碩士(2014年)

https://doi.org/10.6342/NTU.2014.10179

摘要

從台灣期貨交易所統計資料可看出，台灣的專業法人在期貨市場的參與率是偏低的約不到1%，再從手握的投資資金的二大業者來看，投信業國內股票型基金規模約2,612億台幣，但在期貨市場的合約值僅佔基金規模1%不到;保險業10,576億投資於國內股市，但在期貨市場統計資料被歸屬和銀行、票券、中華郵政、投資公司在同一類的其他法人的參與率，也不過是0.58%，如何提高這些專業投資機構對期貨市場操作的可能性，本研究提出以避險角度出發，提供避險平台輔助期貨操作，除可保護投資組合外，也可促進台灣期貨市場的發展。

本研究摒棄傳統避險以最小平方法、ARCH或GARCH計量模型預測價格求算避險比例比較避險效果，而是以多支程式交易訊號組合而成的買賣訊，輔以基因演算法配置後，得出何時該空頭避險，該避多少比例，從報酬面來看避險前後的效果，並將實證樣本區間分為樣本內及樣本外，看在樣本外的效果是否一樣有避險的效益。

以1,000張0050為現貨投資組合模擬現貨，實證結果得到在樣本內未避險前的報酬為-393,480，空頭避險後的報酬為52, 976,120，在樣本外未避險前的報酬為8,800,000，空頭避險後的報酬為10,043,600。以增加投資效益的角度，將多頭訊號用來增加0050現貨的投資，合計空頭避險報酬為90,986,964，而在樣本外多頭投資效益加計空頭避險報酬12,744,097。

分析結果在空頭避險上，不論在樣本內或樣本外都比避險前的報酬為佳，顯示有不錯的效果，如再加計增加投資效益多頭部份，效果更為顯著。此研究結果期能提供專業法人依不同自己的需求，自行設計搭配適合自己的程式交易組合避險平台，並用演算法做最佳配置，積極參與期貨市場操作，讓期貨價格發現功能更為顯著。

關鍵字

避險 ； 台指期貨 ； 模組化 ； 基因演算法 ； 機械化操作

並列摘要

According to the statistic of Taiwan Futures Exchange, the participation rate of local professional finance institution investors is less than 1%. The local investment trust companies who are one of the largest fund holder in Taiwan, their local equity fund size is approximately NT$ 261.2 billion, but their futures contract market value is less than 1% of the fund size. The insurance companies who are another of the largest fund holder invest local equity market size is approximately NT$ 1 trillion. Also counting along with “other institutional investor sectors” including the insurance, banking, bill finance, Chunghwa post and investment companies sectors, their participation rate is only 0.58%. How to raise these professional finance institutions participate rate in Taiwan futures market, from the view of hedging purpose in this study, we create a module trading signal platform which not only can hedge the investment portfolio but also facilitate the development of domestic futures market.

Instead of using tradition hedging methods such as OLS, ARCH or GRACH econometric models to predict the pricing and calculate the hedging ratio, we used several program trading signals and genetic algorithms to determine the timing of long/short and ratio of hedging. We take two sample periods, one is inside and one is outside the sample period, and then compared the profit in each period with the effective hedging.

Taking 1,000,000 shares of 0050 ETF as an equity portfolio, the return of inside sample period before and after short hedging is -393,480 and 52,976,120 respectively. The outside sample period before and after short hedging is 8,800,000 and 10,043,600 respectively. From the perspective of increasing investment return, using the bullish signal to enlarge the investment on 0050 ETF and combine with short hedging profit is 90,986,964. On the other hand, the profit on the outside sample period is 12,744,097.

From the result of short hedging analysis, both inside and outside period show the return is better on before hedging, especially increasing the investment on bullish signal. The result of this study expected provides an idea, make the institutional investors to design their own program trading signal and allocate by genetic algorithms to fulfill their needs. This would increase the participation rate of futures market and also discover the real pricing in futures market.

並列關鍵字

Hedging ； Taiex Futures ； Module ； Genetic Algorithm ； Robot Hedge Trading

參考文獻

5. 柯子介，2008年，匯率分離管理─利用總體經濟趨勢建立交易訊號之外匯交易模型，國立臺灣大學財務金融學研究所碩士論文

6. 劉玉琦、何怡滿、許溪南，2009年，權變避險模式在台灣股市應用，台灣管理學刊第9卷第1期P23-46

7. 劉泰山、何怡滿、許溪南，2011年，KD及MACD在避險時機之應用以台股指數期貨為例，東吳經濟商學學報第72期109-138

1. Ghosh, A, 1993, “Hedging with Stock Index Futures: Estimation and Forecasting with Error Correction Model” Journal of Futures Markets, 13(7), 743-752

2. Koutmos Gregory, Kenneth F. Kroner, and Andreas Pericli., 1998, “Dynamic Cross Hedging with Mortgage-Backed Securities”, Journal of Fixed Income, vol. 8, no. 2, 37-51