中華職棒2014年至2016年得分期望值相關研究

Abstract

本研究以中華職棒2014年至2016年球季約6萬筆play by play的紀錄為樣本，藉由得分期望值相關矩陣對攻守戰術進行分析，可供中華職棒各隊解讀賽場上的狀況，於臨場做出最合理的調度與戰術選擇。以馬可夫鏈的概念為基礎，依據比賽中不同的出局數和佔壘情形，定義出24種狀態，將比賽紀錄重新編碼後，算出各種狀態下的得分機率、打席數期望值、得分期望值等矩陣，可針對觸擊、盜壘、代打調度策略等項目進行分析探討。本研究結果發現，中華職棒2014年至2016年球季間，得分機率、打席數期望值、得分期望值都逐年上揚，雖然2014年與2015年犧牲觸擊後的得分機率略有提升，但2016年卻出現下滑，在三個不同球季中，犧牲觸擊戰術同樣會造成得分期望值的下滑，而且幅度越來越大；在盜壘戰術的部分，兩人出局後的盜壘成本最低，一人出局最高，無人出局次之。綜合上述研究結果，本研究建議，當球隊的攻勢順暢時，應減少下達犧牲觸擊與盜壘戰術，以免對球隊的得分產生負面影響。To help the four CPBL teams recognize the conditions and make the most sensible tactical choice, this study analyzed the offensive and defensive strategies by the run expectancy matrix which is derived from the 60,000 plays of CPBL 2014-2016 season. Based on the concept of Markov chain, we defined 24 kinds of situations according to the base-occupied/out state on the field for re-coding the game scores and calculating varies of matrixes such as the probability of scoring, plate appearance expectancy and run expectancy. These matrixes are useful tools for tactics analyzing like sacrificed bunt, stealing and pinch-hitting move. The research results found that all the probability of scoring, plate appearance expectancy and run expectancy numbers kept growing year by year. The bunting tactic in 2014 and 2015 season raised the probability of scoring, but it overturned in 2016, and it also decreased the run expectancy in all of the three seasons. In terms of base stealing, the tactic cost the lowest after two outs, second in no out and the highest in one out. According to the above results, we suggest teams should reduce bunting and stealing tactics when the offense was flowing, or it might have a negative impact on the team's offense.