The purpose of the present study was to create a probabilistic neural network (PNN) to clarify the understanding of movement patterns in international judo competitions by gender. Analysis of 773 male and 638 female bouts was utilized to identify movements during the approach, gripping, attack (including biomechanical designations), groundwork, defense, and pause phases. PNN and X2 tests modeled and compared frequencies (p=0.05). Women (mean[interquartile range]: 9.9[4;14]) attacked more than men (7.0[3;10]) while attempting a greater number of arm/leg lever (women: 2.7[1,6]; men: 4.0[0;4]) and trunk/leg lever (women: 0.8[0;1]; men: 2.4[0;4]) techniques, but fewer maximal length moment arm techniques (women: 0.7[0;1]; men: 1.0[0;2]). Male athletes displayed one-handed gripping of the back and sleeve, while female executed a greater number of groundwork techniques. An optimized PNN model, using patterns from the gripping, attack, groundwork, and pause phases, produced an overall prediction accuracy of 76% for discrimination between men and women.
© Copyright 2017 Motor control. Human Kinetics. All rights reserved.