UBC Theses and Dissertations
Random and deterministic (nonrandom) aspects of athletic behavior with special reference to National League hockey MacDonald, Neil William
Various parts of the question concerning how random and deterministic attributes intertwine during the course of athletic contests have been explored by researchers. This study attempted to extend the research data base and formulate the initial postulates for a model to describe the random/deterministic interaction. The 1988-89 National Hockey League season was the primary focus of attention. Supplemental examination was made of the 1937-38 and 1946-47 NHL, the last 50% of the 1988-89 National Basketball Association season and the 1987-88 Football Association English First Division seasons. The data overwhelmingly supported earlier studies which argued that major outcomes (wins/losses, goals, shots) followed a random sequence. The axiomatic model developed argued that the random pattern of outcomes is quite pervasive (wins/losses, shots and goals for, against or combined are distributed randomly whether home, away or total games are examined). The pattern of outcomes (win/losses, goals, shots) is relatively independent of the size of the unit of measurement: random patterns held whether one period, two period, three period games or four-game sets were examined. Conditional probability tests showed game-to-game outcomes were independent (a win was no more likely to be followed by a win than by a loss). The pattern of outcomes (goals) is dependent on how the data is examined. If all 21 team's goals are plotted time-wise, goals are distributed uniformly minute by minute (except for the last minute of play). If goals or shots per game (or period) are tallied for home, away or both teams, the resultant frequency distribution will approximate the negative binomial distribution. However, if the time-spaces between goals are tallied, a geometric distribution will emerge. Deterministic effects were demonstrated when artificial season outcomes based on first, second or third period only seasons were found to correlate favorably with real season outcomes (wins, losses, points, goals for, goals against). Finally, comparison of hockey, basketball and soccer outcomes suggested that upset rates may vary from one sport to another.
Item Citations and Data