Comparing and Forecasting Performances in Different Events of Athletics Using a Probabilistic Model

• Activity Number: 54
• Type: Invited
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Sports
• Author(s): Brian Godsey*+
• Companies: RedOwl Analytics
• Keywords: athletic performance ; bayesian ; probabilistic model

Abstract

Subsequent, and sometimes prior, to great athletic achievements, there is much discussion about the quality and rarity of such a performance. Commentators commonly compare performances to a current or former world record, to a ranking on a list, or to a pre-calculated table of points such as the IAAF Scoring Tables. While these methods are informative, none establishes how likely such a performance is to occur, this year and beyond. Also, none are robust, in that they rely on a few key performances (e.g. world records, top 10 lists). I present a method that utilizes lists of the best marks in history (as many as are available) for each Olympic event, to estimate a probability distribution of elite performance. Though data exist only for the high-performance tails of the distributions, parameters can be estimated meaningfully via MCMC simulation, and for each individual mark, a p-value can be calculated. These p-values can be used directly to predict the number of performances that will exceed this mark in subsequent years (at which task they out-performed the IAAF tables), and more specifically, they can can be used to calculate the probability of a world record being broken.