Identifying the reasons for which some individuals and teams achieve success is one of the most common goals in the research literature on sport performance. The methodologies used to study sport performance have been established to compare the actions of successful and unsuccessful teams and athletes, and the outcomes include a range of statistical data of discrete actions performed by teams and players during competition (see Hughes & Franks, 2004). Despite the importance of these data, a significant criticism of the notational analysis of sport performance is that it does not identify the reasons for those discrete actions in order to explain the difference between successful and unsucessful teams. The data typically inform us of what happens, but not of how and why it happens.To understand the dynamical processes in sport performance, McGarry and Perl (2004) used a specific type of neural network: the Kohonen feature map. They argued that the neural networks approach must be used first to recognize situations and analyze processes, and only then to identify decision-making processes. However, with that approach, a 2-D information structure (e.g., a pair of coordinates) was implemented to represent a pattern, which decreases the ability of a neural network to analyze the complex processes that emerge in sports contests.Additional problems of network learning processes include their dependency on input patterns and the fact that those patterns change continuously as a result of a player's tactical behaviors. A theoretical rationale from the field of ecological psychology allows us to understand that tactical behaviors are dependent on the information available in specific contexts, and that that information is geared to each player's tactical behavior. Clearly, the use of Kohonen feature maps to study performance might incur problems in reliably recognizing patterns in sports contests. Another method proposed by McGarry and Perl (2004) is the dynamical controlled network (DyCoN), already tested successfully in different areas of sport (Perl, 2002). This method deals with the problem of dependency on input patterns. The DyCoN has the ability to learn continuously or in discrete phases, so input patterns need not be presented to the network in a single data set. McGarry and Perl (2004, p. 241) argued that the information acquired during the learning of the DyCoN could be used in "complementing our understanding of the processes that take place in the learning of cognitive and/or motor skills In previous attempts to identify dynamical systems properties in patterns of play in team sports, only 2-D analysis methods have been used, implying that the plane of motion must be preselected and that movements out of the chosen plane are ignored. In the present study, we examined the usefulness of 3-D methods of analysis for establishing the presence of dynamical systems properties, such as phase transitions and symmetry-breaking processes in the team sport of rugby. Artificial neural networks (ANNs) were employed to re...