基於F-score以及修正式F-score的交易策略

Abstract

本研究透過近20年來仍屹立不搖的選股模型，Piotroski F-score，建構不同交易策略並進行投資組合比較。此外，本研究以F-score為基礎，並利用公司對於每一項指標的達成率給予F-score進行分數調整，修改為Revised F-score。本文以台灣上市櫃公司為研究標的，研究期間從2013年5月15日至2022年3月31日。\n透過交易策略的設計，我們將逐步調整並增加不同條件，希冀能接近實際交易情況。首先，本文將先討論在固定選取30或50家公司的投資組合下，比較只考量F-score以及同時考量F-score以及Revised F-score的績效表現。在固定選取30家公司並以市值加權的投資組合中，同時考量F-score以及Revised F-score在統計上顯著優於僅考量F-score，說明了修改後的評分方式有助於績效的提升。後續在同時考量F-score以及Revised F-score的模型基礎下，以固定選取30家公司並以市值加權的投資組合進行各種條件測試。實證結果顯示，不論在何種策略下，絕對報酬都優於元大台灣50 ETF以及台灣加權股價報酬指數。此外，本文最後也考量資金規模並建構投資組合，年化報酬更提升至36.6%。此方法將提供個人以及機構投資者在此研究的模型中，找出適合的投資標的

This research uses the Piotroski F-score, a stock selection model that has been remain strong for nearly 20 years, to construct different trading strategies and compare performance of portfolios. In addition, We use the F-score as the basis and calculate the company`s achievement rate for each indicator to adjust the F-score, which is called Revised F-score. This paper considers companies traded in Taiwan Stock Exchange and Taipei Exchange, and the research period is from 15 May 2013 to 31 March 2022.\nThrough the design of the trading strategies, we will gradually adjust and add different conditions into strategies in order to close to the actual trading situation. First, this research will compare the performance of the F-score with that of the Revised F-score. Under the portfolios with market value weighted and 30 companies in the portfolio, the performance of considering the F-score and the Revised F-score is significantly greater than that of considering the F-score only. The statistical result shows that the revised scoring method can improve the performance significantly. We will conduct further researches on the basis of the portfolio with market value weighted and 30 companies in the portfolio under the consideration of the F-score and the Revised F-score. The empirical results show that, regardless of any strategy, the absolute returns are better than the Yuanta Taiwan Top 50 ETF and TWSE Capitalization Weighted Stock Index-Total Return Index. In addition, this research also considers the size of funds to construct the portfolios and the annualized return can be 36.6% at most. This method provides individual and institutional investors to find suitable investment targets in the model of this research.