Prediction of world records in athletics and swimming by a time-series analysis / by Marilyn Ruth Nelson. --

Abstract

In an age of a flourishing emphasis on sports and a high frequency of individual record breaking, a detailed description of performance trends would provide a better understanding of what might happen in the future. In this study, world records in swimming and athletics were analyzed to relate the time of occurrence to their magnitude in order to predict future record performances. Records were considered from 1945, or the earliest date after 1945, to 1977 and subjected to a time-series analysis (Box-Jenkins method) to determine predicted values for 1978 through 1984. Predictions and their confidence limits were developed for all events. A 5% error rate was considered as the widest acceptable degree of error. Only some track events fell within this criterion range and therefore, contained adequate predictions. Swimming and field events were mainly unacceptable in light of the predictions which were made. Several variables affecting predictions were discussed. Otherwise the prediction of world record performance trends in swimming, track, and field was found to be unsatisfactory when world records served as source data.