Identification of Elite Swimmers' Race Patterns Using Cluster Analysis

Abstract

The identification of elite swimmers' race patterns at the same time is an important unresolved problem. This problem appears to be of fundamental importance for coaches in training promising elite swimmers. To address this problem, a system of cluster analysis for studying group structures on the basis of elite swimmers' race results and various available race components, such as lengths, speeds and times, is described that uses standard statistical algorithms to arrange elite swimmers according to similarity in their race patterns. The outputs are displayed graphically using tree-like dendrograms, conveying elite swimmers' race patterns in a form intuitive for coaches. As an illustration of this, we have used this method to uncover a number of key race components in connection with Ian Thorpe's race results in the men's 400m freestyle final at the Sydney 2000 Summer Olympics

A hierarchical cluster analysis to determine whether injured runners exhibit similar kinematic gait patterns

“…Unsupervised machine learning works by discovering underlying patterns and associations in datasets without any a priori information aside from a set of input variables . These statistical methods have demonstrated success in uncovering underlying structure within datasets in previous sport science research . For example, race patterns in elite swimmers, different golf swing patterns, and different gait patterns in injured and healthy runners have all been successfully identified using unsupervised machine learning methods.…”

“…12 These statistical methods have demonstrated success in uncovering underlying structure within datasets in previous sport science research. [13][14][15][16][17] For example, race patterns in elite swimmers, 13 different golf swing patterns, 14 and different gait patterns in injured and healthy runners [15][16][17] have all been successfully identified using unsupervised machine learning methods. Another study used unsupervised machine learning to create a movement profile for healthy controls walking and assessed the deviation of patients with gait problems from this profile.…”

Freestyle race pacing strategies (400 m) of elite able-bodied swimmers and swimmers with disability at major international championships

“…Although 400m freestyle pacing can be described in general terms that appear to be supported empirically, idiosyncrasies are evident when the specific approaches adopted by swimmers are examined in more detail. For example, an application of cluster analysis to components of performance at the 2000 Olympic Games 400m freestyle swimming final by Chen, Homma, Jin, and Yan (2007) revealed distinct groupings that reflect differing 'race patterns' of the individual swimmer. Further, using a computer algorithm to examine 264 national and international 400m freestyle swims Mauger, Neuloh, and Castle (2012) found that the 'fast-start-even' and 'parabolic' pacing strategies were most prevalent but did not result in improved performance times when compared to the alternative positive, negative or even pacing strategies.…”

“…Although the extant literature regarding pacing strategies adopted within the 400m freestyle swimming event has provided useful insight these conclusions have been based upon elite able-bodied swimmers (Chen et al, 2007;Maglischo, 2003;Mauger et al, 2012;Robertson et al, 2009;). These findings potentially have limited application to other PACING STRATEGIES IN ELITE 400m FREESTYLE SWIMMERS 5 populations, such as Paralympic swimmers, where unique biomechanical and physiological demands can be present (Fulton, Pyne, Hopkins, & Burkett 2009).…”

A Machine Learning Approach to Predicting Winning Patterns in Track Cycling Omnium

“…Data mining and machine learning-based approaches have also been used widely in this domain. Examples include modeling and predicting competitive performances in swimming using Neural Networks [4], using Self Organizing Maps for classifying coordination patterns of participants in different types of basketball shots [5], cluster analysis of elite swimmers' race patterns [6], predicting player injuries in soccer using data mining [7], and knowledge discovery techniques implemented in Advanced Scout software used for finding series of consecutive events that may lead to scoring in NBA basketball games [8].…”