The development of a stochastic model for predicting championship squash performance

• Main Author: McGarry, James T.
• Format: Others
• Language: English
• Published: 2008
• Online Access: http://hdl.handle.net/2429/1799

Sport analysis provides objective data regarding individual or team performance in the widespread belief that the information available is applicable to a future setting. Recent matches, often against different opponents, of two competitors scheduled to meet next are frequently analyzed in some form for a priori match information. An underlying assumption is that consistency in athletic performance exists, to the extent that these analyses duly inform the coaching process in preparation for the forthcoming contest. The hypothesis of this study was that an elite player's match “profile" characterizes individual playing habits in squash and can be used to accurately predict future sport performance. The data comprised the quarter-finals through final of the men's Canadian Open Championship, 1988.The hypothesis was tested through the development and application of a stochastic Markov model for predicting future competition squash match play from past performance. The results, using two way chi-square analysis, revealed consistency in individual athletic behavior between games (and also between matches) against the same opponent (p>.25), but not between matches against different opponents (p<.25). It seems that a player's profile is dependent on the particular interaction between two individuals, and that this pattern irreproducible within this interaction. Importantly, however, the interaction is unique to the two players and not generalizable to other opponents. This suggests that additional information from the opponent is required if the model is to successfully predict future sport performance. The present model possesses no predictive validity beyond the scope of its previous observation(s), a restriction which unfortunately consigns the past to forever repeat itself. The application of the model for the identification of future optimal strategy has therefore little present utility. The attempt to model competitive sport performance is nonetheless a worthy pursuit, not least because it directs attention to the critical aspects of data which determine sport success. Further research is recommended to investigate whether an underlying order which characterizes player performance can be established. If no such individual signature can be identified in sport, the utility of commonly applying tactical strategy from previous observation(s) against different opponents must be questioned. === Education, Faculty of === Curriculum and Pedagogy (EDCP), Department of === Graduate