勝場貢獻值之研究與應用-以超級籃球聯賽為例

Abstract

本研究以勝場貢獻值 (Win Shares, WS) 為理論基礎，探討第12季超級籃球聯賽(Super Basketball League, SBL) 七支球隊共117名球員的勝場貢獻值及相關問題。方法：首先藉由文獻探討方式分析目前被使用的各種評價球員模型及公式，說明其理論及運用的方式。接著從中挑選並解釋為何本研究使用勝場貢獻值作為理論基礎。進而分析2014年11月22日開打至2015年3月1日止，第12季超級籃球聯賽105場的例行賽賽事中，七支職業球隊各隊登錄並有上場紀錄之球員共117名，分析各球員對球隊的貢獻程度為何。接著說明勝場貢獻值於球隊的實際勝場數之間的關係為何。最後探討12季SBL年度各項獎項獲獎球員的勝場貢獻值內容。結論：研究發現部分球員之個人勝場貢獻值為負值，但因影響貢獻值的結果不只一項，很多個人能力很好的球員，勝場貢獻值不一定會高於能力較差的，原因在於能力好的球員他的隊友不見得夠好，球對整體戰績不佳，便會間接地影響到勝場貢獻值，反觀能力較差的球員，若是其隊友及球隊的整體表現有達到水準，同樣的會提升這名球員的勝場貢獻值。其次，SBL第10-12季的球隊勝場貢獻值與實際勝場之間的相關係數為0.802，表示相互之間具高相關性。以第10-12季7支球隊的勝場貢獻值為自變數與實際勝場之關係經由逐步迴歸分析得出結果顯示，勝場貢獻值對預測球隊勝利有顯著相關，有80%的相關性及62.4%的解釋力。以第10-12季7支球隊的團隊基本攻守數據為自變數(上場時間、兩分球命中數、兩分球出手數、兩分球命中率、三分球命中數、三分球出手數、三分球命中率、罰球命中數、罰球出手數、罰球命中率、進攻籃板、防守籃板、助攻、抄截、失誤、阻攻、個人犯規、得分)，與勝場貢獻值之關係經由逐步迴歸分析得出結果顯示，主要影響貢獻值的為兩分球命中數，有高達94%的解釋力，相關係數則為97%，具高度相關。預測能力方面，利用貢獻值來預測NBA的兩個球季，迴歸出的結果分別為50.2%及52.6%的解釋力，相關係數則為70.5%及72.9%。The study uses win shares as the theory foundation, exploring the twelfth season of Super Basketball League, along with seven teams, 117 players in total and other related questions. Method: First of all, this study analyze the models of various types of players and the equations, which demonstrate the theory and how it was applied by using documentary analysis. Then we select a few and explains the rationale of applying win shares as the theory foundation for the study. Therefore the game stats from November twenty-second, 2014 to March first, 2015 was analyzed. According to the twelfth regular season of Super Basketball League, total one-hundred and seventeen players were registered by seven professional teams, and calculate how each player contributes to their team. Afterwards we explain the relation between win shares and the actual wins. Finally, the awarded players' win shares of the twelfth season of SBL were explored. Conclusion: The study finds some players have negative win shares, but due to the affecting factors of win share is more than one, many efficient players don't always have the higher win share than the less efficient players, the reason is the efficient players may have less efficient teammates, with an overall unsatisfactory record, which directly affects win share. On the contrary, if a less efficient player is on an average team with a satisfactory record, it'll also improve his win share. Secondly, the correlation coefficient between the win share and actual wins is 0.802 from the tenth to twelfth seasons of SBL. According to the gradual regressive analysis of these seasons, the seven teams' win shares as the independent variable along with the actual wins, the result shows, win share is indeed a siginificant factor of predicting the winning, with 80% of correlation coefficient, also 62.4% square multiple correlation. According to the tenth to twelfth seasons, the seven teams' basic team offensive and defensive statistics are independent variables (playing minutes, field goals made, field goals attempted, field goal percentage, three point field goal made, three point field goals attempted, three point field goal percentage, two point field goals made, two point field goals attempted, two point field goal percentage, free throws made, free throws attempted, free throw percentage, offensive rebounds, defensive rebounds, total rebounds, assists, steals, turnovers, blocks, personal fouls, points), win shares through the gradual regressive analysis obtains the results that indicate the win shares are mainly affected by two point field goals made, nearly 94% square multiple correlation, and the correlation coefficient is 97%. As for predictions, by using win shares to predict two seasons of NBA, regressive analysis values are 50.2% and 52.6% square multiple correlations; in addition, the correlations coefficients are 70.5% and 72.9%.