This study aimed to quantitatively investigate the influence of effective visual searching behaviors on recognizing extraordinary events during the acceleration process of railway driving. Simulator training data was analyzed, including eye movements data from 128 actual drivers. The given driving scenario was to recognize a major extraordinary event (a subsidence of a railway track) after dealing with four minor events. Participants who braked before passing the subsidence were identified as part of the recognizing group, and those who did not brake until after passing the subsidence were identified as part of the nonrecognizing group. Logistic regression analysis revealed that the possibility of recognizing subsidence increased by 1.41 times when the percentage of gazing at the landscape increased by 1%, and the possibility increased by 1.21 times when the percentage of gazing at the railway track increased by 1%. The percentage of gazing at the landscape had the largest influence. The regression model with these variables can classify the participant group, with the percentage of correct classifications being 86.7%.