Mathematical models that predict athletic performance in sprint and jump events for current and future world championships

The prediction of future athletic performance is a recurring theme as sports scientists strive to understand the predicted limits of sports performance. Predictive models based on Olympic data for athletics have derived some accurate predictions of performance in the 2000, 2004 and 2008 Olympic Games. The aim of this research was to develop predictive models using IAAF World Championship data in athletics outdoor championships 1983 to 2009 for men's and women's 100m, 400m, long jump and high jump. Data were average scores of the top six performances in the finals of each event. The approach utilised regression-curve estimation by evaluating fit to linear, logarithmic, inverse, quadratic, cubic, compound, power, sigmoidal, growth exponential and logistic functions. The mathematical models were poor predictors of past, current performance and future performance in these events (low R2, non-significant p-values and large residuals or error). The one exception was the women's long jump, which predicted a significant decline in performance. The low predictive ability may be due to a short timeframe data set and insufficient to develop substantive predictive data, a timeframe unable to detect phylogenetic trends in human athletic performance or excessive variability in athletic cohorts, which mask trends if they occur.