本文應用 Rubinstein & Leland （ 1981 ）的研究，進行各種選擇權部位的複製 。 有鑑於臺灣股市巨幅的波動性， 整合在多頭、 空頭下獲利能力較佳的 strap 策略與 strip 策略，佐以濾嘴調整法則，提出 strap 與 strip 混合策略，並與 straddle 策略、 複製性賣權策略、買入持有策略進行績效評估。另外，由於股價波動性的估計在選擇權複製 策略中扮演關鍵性的角色， 本文捨棄傳統的歷史資料法，以移動平均法（ moving average ）及 GARCH 模型估計之。本研究的結論有三：（ 1 ）混合策略與 straddle 策略的績效顯 著優於複製性賣權策略與買入持有策略，但是兩者期末報酬的分配具有相當高的波動性，顯 示這兩種策略都屬於高風險高報酬的投機性策略。 （ 2 ）在多頭時期， 混合策略與 straddle 策略因為沒有自我融資的限制，因此績效顯著優於其他策略。 在盤整時期，混合 策略與 straddle 策略的期末資產均明顯低於買入持有策略。在空頭時期，複製性賣權策略 的確能發揮保險功能，而混合策略與 straddle 策略更能獲得正值報酬率。 （ 3 ） GARCH 模型的績效表現略優於移動平均法，但均未達統計上的顯著性。

　　　　　This paper applies Rubinstein and Leleand's (1981)method to replicate a variety of option positions. Due to the characteristics of large volatility of the Taiwan Stock Market, we integrate strap and strip strategies to develop a mixed strategy. The filter rule plays a "switch" function which controls the alternate applicaton of strap and strip strategies in bull and bear markets, respectively. The performance of the mixed strategy is compared with those of straddle, synthetic puts, and buy-and-hold strategies. Furthermore, due to the importance of volatility estimates in replicating strategies, this paper uses moving average and GARCH models to estimate volatilities. The conclusions of this paper are as follows: (1) The performance of the mixed strategy and straddle strategy are significantly superior to those of synthetic puts and buy-and-hold strategies, but the distributions of terminal values for mixed and straddle strategies are highly volatile, indicating that they are speculative strategies of high risk and high return. (2) In bull markets, the mixed and straddle strategies perform better than other strategies due to no limitation of self-financing. In correction markets, the performance of mixed and straddle strategies are significantly lowe than that of buy-and-hodl strategy. In bear markets, though synthetic put will provide insurance furnction, the mixed and straddle will provide positive return. (3) The performance of GARCH model for volatility is slightly superior to that of moving average model.