Predicting a nation's Olympic-qualifying swimmers

Sian Allen, Tom Vandenbogaerde, David Pyne, Will Hopkins

    Research output: Contribution to journalArticle

    10 Citations (Scopus)

    Abstract

    Talent identification and development typically involve allocation of resources toward athletes selected on the basis of early-career performance. Purpose: To compare 4 methods for early-career selection of Australia's 2012 Olympic-qualifying swimmers. Methods: Performance times from 5738 Australian swimmers in individual Olympic events at 101 competitions from 2000 to 2012 were analyzed as percentages of world-record times using 4 methods that retrospectively simulated early selection of swimmers into a talent-development squad. For all methods, squad-selection thresholds were set to include 90% of Olympic qualifiers. One method used each swimmer's given-year performance for selection, while the others predicted each swimmer's 2012 performance. The predictive methods were regression and neural-network modeling using given-year performance and age and quadratic trajectories derived using mixed modeling of each swimmer's annual best career performances up to the given year. All methods were applied to swimmers in 2007 and repeated for each subsequent year through 2011. Results: The regression model produced squad sizes of 562, 552, 188, 140, and 93 for the years 2007 through 2011. Corresponding proportions of the squads consisting of Olympic qualifiers were 11%, 11%, 32%, 43%, and 66%. Neural-network modeling produced similar outcomes, but the other methods were less effective. Swimming Australia's actual squads ranged from 91 to 67 swimmers but included only 50-74% of Olympic qualifiers. Conclusions: Large talent-development squads are required to include most eventual Olympic qualifiers. Criteria additional to age and performance are needed to improve early selection of swimmers to talent-development squads.

    Original languageEnglish
    Pages (from-to)431-435
    Number of pages5
    JournalInternational Journal of Sports Physiology and Performance
    Volume10
    Issue number4
    DOIs
    Publication statusPublished - 2015

      Fingerprint

    Cite this