Prediction Models for Track and Field Performances

Abstract

The specific purposes of this study were (a) to identify the best mathematical model and the best data set for predicting men's future world-best track and field performance, (b) to predict ultimate performances for seven selected track and field events, and (c) to develop a new "random sampling" model to predict performance for those events that have exhibited asymptotic-like behavior over the last 15 years. Linear and nonlinear models were used to fit world record data and best performance per year data to identify the best fitting model. Extreme value theory and Monte Carlo simulation methods were used to derive predicted future performances under the random sampling model for the 1,500-m event. The results showed that (a) an exponential model relating running time and historical year with yearly best performance data is the most valid deterministic model for the prediction of world records and ultimate performance, and (b) the random sampling model is an effective method to predict future world records for the 1,500-m event, the only event to exhibit asymptotic performance over the last 15 years.