應用邏輯斯迴歸建構中華職棒大聯盟優質先發之模型

美國運動作家John Lowe在1985年時針對美國大聯盟，所提出投手先發時至少須投6局且責任失分不得超過3分的優質先發（Quality Start）概念。這項統計數據剔除了得分支援、救援失敗、以及失誤等的影響，僅參考先發投手個人的表現並與棒球統計學上越來越多試圖反映個人成就的趨勢相符。優質先發這項指標適用於美國職棒大聯盟，但未必適用於中華職棒大聯盟。因此，本研究依此概念，利用中華職棒大聯盟官方網站所提供的2005年至2010年球季每場先發投手相關資料，透過邏輯斯迴歸分析建構中華職棒大聯盟優質先發的模型。本文分析結果發現：2005年至2010年中華職棒大聯盟先發投手勝投者在各因素上的表現優於非勝投者。另外，先發投手的投球局數、自責分率及被上壘率是影響先發投手勝投與非勝投的關鍵因素，並藉由關鍵因素的產生，再搭配邏輯斯迴歸分析，建構出中華職棒大聯盟的優質先發模型為0.282x投球局數-0.303x自責分率-0.313x被上壘率-0.845，其準確性頗佳。藉此模型瞭解各隊所屬先發投手較真實的能力，以利球團評估球員的實力。

關鍵字

棒球統計學；先發投手；關鍵因素

並列摘要

In 1985, the sportswriter John Lowe developed quality start statistic for MLB which is a concept that a starting pitcher allows no more than three earned runs at least six innings in a game. According to the quality start, it eliminates fielding errors, blown saves, and poor run support; on the other hand, the quality start focuses on starting pitcher evaluation and corresponds to the personal success which consistent with concept of Sabermetrics. The statistic works well at MBL; however, it may not apply on Chinese Professional Baseball League (CPBL). Therefore, on this research, collecting data from Chinese Professional Baseball League official website and evaluating starting pitchers from 2005 until 2010 are used to develop the quality start statistic model at CPBL. As the result, starting pitchers as winning pitchers perform better than losing pitchers from 2005 to 2010. Also, recognizing innings pitched, earned run average, and on base against are critical factors which affect starting pitchers performing. In conclusion, combining affective critical factors and using logistic regression develops the model which applies for CPBL. Therefore, using this model to analyze starting pitchers in CPBL will more affect; also, it will show players' strength and quality more accurately.