- 學位論文
少數者賽局在財務時間序列的預測與交易策略–以台灣加權指數期貨為例
Financial Time-Series Modeling and Trading Strategy Design through Minority Games. Using TX of TAIFEX as Example
摘要
本文利用少數者賽局模型為基礎，發展混合不同類型交易代理人的演化少數者賽局，希望利用簡單的代理人模型捕捉台灣指數期貨的價格變化，並以此為基礎發展交易策略，並將此一交易策略實際應用於台灣指數期貨的交易上，驗證其績效與有效性。本文中以2001到2010年的台指期歷史資料做為樣本內研究對象，發展符合台灣指數期貨市場的少數者賽局模型，而在交易策略部分，則先以2001到2010年的資料做回測，檢視其績效，然後以2011/01~2011/06六個月的資料做為樣本外的資料，驗證交易策略的有效性。實證結果發現，基於少數者賽局的市場模型的日報酬機率分佈與真實市場相近。而交易策略部分，搭配隱馬可夫模型做為濾網，可以有效增加績效，在計入交易成本的考量下，仍可以達到年化報酬約13%，而樣本外的資料，其年化報酬也有14%，最後我們利用蒙地卡羅模擬法來驗證交易策略的有效性，獲得相當的統計顯著性，因此利用少數者賽局模擬市場並以此發展交易策略並程式化交易是有一定潛力的。
並列摘要
In this paper, we discuss the minority game and develop a mix-agent evolutionary minority game model for TX of TAIFAX which is the index future of TAIFAX. Then we design a trading strategy based on the signal triggered by the model mentioned above. We test the profitability of the strategy by using back testing through history dataset from 2001 to 2010. And then we use out of sample data which are from 2011/01 to 2011/06 to verify the empirical result. We discover that the model based on our research has quite similar statistical properties with Taiwan stock market. Therefore, the trading strategy we design by using both MG model and HMM is quite profitable. According to the back testing, we can earn about 13% annual return rate during 2001 to 2010. Even more, we can also have about 14% annual return rate during 2011/01 to 2011/06 when there are out of sample data. We also verify the result by Monte Carlo simulation test by which we have high statistical significance so that our trading strategy is possibly useful.