2018 年 63 巻 1 号 p. 315-325
The present study was verify of reliability and validity the regression equation constructed using the game footage from the 2010 FIFA World Cup. Formulated using multivariate logistic regression analysis, this equation’s purpose was to predict the probability of a goalkeeper’s failure to stop shots at goal. The 2014 dataset consisted of 587 shots at goal from within the frames during 64 games played by 32 teams that participated in the FIFA World Cup in Brazil. The results showed that for the probability of failing to block a shot predicted by the regression equation, the 2 inter-rater interclass correlation coefficient was high (0.91), thus demonstrating reliability. Furthermore, when a contingency table was created with a cut-off value of 0.5, accuracy was high (85.9%); when an ROC curve was drawn and the area below the curve was measured (0.874), it was notably distinct, indicating generalizability. Therefore, the regression equation predicting the probability of failure to block a shot has high generalizability for predicting world-class goalkeepers’ probability of failure to block shots. This is considered a useful formula for evaluating goalkeepers’ blocking ability.