Ice hockey players have long been given different monikers to describe styles of play (e.g., "enforcers," "grinders," "skilled players"). The purpose of this paper was to empirically determine whether there is any basis for these categories using the APBA (American Professional Baseball Association) computer simulator for playing the 2003-2004 NHL season. Forwards and defensemen were analyzed separately, while goaltenders were not considered. Variables were initially standardized separately by position to assist interpretation of the resultant clusters. Meaningful clusters of players were generated based both on interpretability and sample size. Cluster centers were examined and descriptors were attached to each cluster depending upon relative variable importance. The descriptors were then used to assign subjective labels. There were seven forward clusters: (1) skilled attackers; (2) second-tier skilled attackers; (3) grinders; (4) enforcers; (5) least effective; (6) defensive specialists; and (7) face-off specialists. There were six defensemen clusters: (1) skilled attackers; (2) second-tier skilled attackers; (3) defensive defensemen; (4) serviceable defensemen; (5) most intimidating defensemen; and (6) least effective defensemen. From these descriptors and the specific players contained in each cluster, it was determined that there is good empirical support for the use of labels to classify hockey players from the computer simulator.
© Copyright 2009 International Journal of Computer Science in Sport. Sciendo. Alle Rechte vorbehalten.