近年來於股市、期貨投資領域中，運用程式交易自動下單越來越為常見。然而當程式交易策略在進行回測時，難免因過度參考已發生之歷史資料，而常有過度最佳化之疑慮。在交易策略相關研究議題裡，對於策略實際的有效性、穩定性分析也並沒有完整的驗證方式與流程，大多是對不同策略、不同參數因子、不同標的等方面進行比較，然而這樣較片面的分析方式對於交易策略於實務上的幫助並不大。實際上要判別策略是否長期有效且穩定，必須透過移動窗格法，對樣本內、樣本外的區間不斷進行時間推移與績效評估，來避免一般回測下因使用未知資料，而產生的過度最佳化問題。 而本研究為解決上述問題，設計出一套完整的策略驗證流程，並依據此流程設計與實作一平台。能通用於常見程式交易平台（e.g. Multicharts、Amibroker），自動化幫助投資者驗證策略之有效性與穩定性，並找出穩定的參數因子與該商品於該策略適合之窗期大小，供投資者作為實際下單交易前的參考依據。 ;In recent years, in the field of stock market and futures investment, it is more common to use program trading to place orders automatically. However, it is difficult to avoid curve-fitting when the program trading strategy is conducting back-testing owing to over using historical data. In past research topics related to trading strategies, there is no complete verification process for the effectiveness and stability of trading strategies. In most of the related research, only discusses and focuses on the changes of the trading strategy parameters, the comparison of different trading strategies, the selection of different symbol, and so on. However, these analysis methods do not help the actual transaction. In fact, we must to use walk forward analysis to continuously evaluate strategic performance within the every period of window to avoid the problem of over optimize. In order to solve the above problems, this research designs a complete strategy verification process and implements a platform based on this process. This platform can be used with common program trading platforms (e.g. Multicharts, Amibroker) to automate verification the effectiveness and stability of the strategy, and help investors find the suitable parameters before the actual order transaction.
