中華職棒比賽勝負預測模式之建構

Establishing models to predict the outcomes of baseball games in CPBL

目的：本研究之主要目的在於辨識影響比賽勝負之關鍵要素，並建構職棒比賽勝負之預測模式。方法：本研究以中華職棒為研究範圍，以第18年之對戰組合攻守紀錄為研究主軸，運用文件分析與德爾菲法，建立影響中華職棒比賽勝負之影響因子，再用邏輯斯迴歸分析，建構中華職棒比賽勝負之預測模式。結果：一、影響中華職棒比賽勝負之關鍵預測要素，攻擊成績指標共計22項、投球成績指標32項、守備成績指標7項。二、影響攻擊勝負因子中，解釋力高低之因子分別為得分、打數、打點、失誤、犧牲短打（p＜.05）。再由勝數比分析發現，打數、失誤為負面因子（odds ratio＜1），其中又以失誤因子之勝數比最小（odds ratio＝.577）；而得分、打點及犧牲短打為正向因子（odds ratio＞1）。三、影響投球勝負因子中，解釋力高低之因子分別為投球局數、面對打席、被安打、四壞球及自責分（p＜.05）。再由勝數比分析發現，面對打席及自責分為負面因子（odds ratio＜1）；而投球局數、被安打與四壞球為正向因子（odds ratio＞1）。結論：本研究結果可提供未來預測球賽勝算或球團編制戰力時，可參照攻擊與投球之預測指標，加強其戰力的提升。Purpose: The purpose of this study was to construct prediction model for winning and losing of baseball games, and also differentiate the key factors of winning and losing. Method: The research area would mainly focus on Chinese Professional Baseball League (CPBL) and the eighteenth year scoreboard, using content analysis and Delphi research methods and building the factors that influenced the winning and losing of CPBL game. The statistics of logistic regression was used in the analysis in order to construct CPBL prediction model. Results: The results indicated that: 1. According to the expert opinions in this study, key factors which influenced the winning and losing of the game included 22 offensive items, 32 pitching items, and 7 defensive items. 2. Among the attribute factors of offensive, scoring had the highest explanation power which followed by runs, at bats, runner batted in, errors, and squeeze play (p < .05). According to the odds ratio analysis, it discovered that the odds ratio of at bats and errors were less than 1, which meant they were negative factors, and the least was the errors factor (odds ratio = .577). Odds ratio of runs, runner batted in, and squeeze play were larger than 1, which meant that they were positive factors. 3. Among the attribution factors of pitching, game had the highest explanation power which followed by inning pitched, hit batsmen, hits, base balled / walk, and earned runs (p <.05). According to the odds ratio analysis, it discovered that the odds ratio of hit batsmen and earned runs were less than 1, which meant that they were negative factors. Odds ratio of inning pitched, hits, and base balled / walk were larger than 1, which meant that they were positive factors. Conclusion: Based on the findings, hitting and batting predictive items would be a good reference for prediction of winning and losing, and also for team management to increase competitive ability.