Relationships Between Model-Predicted and Actual Match-Play Exercise-Intensity Performance in Professional Australian Footballers During a Preseason Training Macrocycle.

Abstract

To assess and compare the validity of internal and external Australian football (AF) training-load measures for predicting preseason variation of match-play exercise intensity (MEI sim/min) using a variable dose-response model. A total of 21 professional male AF players completed an 18-wk preseason macrocycle. Preseason internal training load was quntified using the session rating-of-perceived-exertion method (sRPE) and external load from satellite (as distance [Dist] and high-speed distance [HS Dist]) and accelerometer (Player Load [PL]) data. Using a training-impulse (TRIMPs) calculation, external load expressed in arbitrary units was represented as TRIMPsDist, TRIMPsHSDist, and TRIMPsPL. Preseason training load and MEI sim/min data were applied to a variable dose-response model, which provided estimates of MEI sim/min. Model estimates of MEI sim/min were correlated with actual measures from each match-play drill performed during the preseason macrocycle. Magnitude-based inferences (effect size [90% confidence interval]) were calculated to determine practical differences in the precision of MEI sim/min estimates using each of the internal- and external-load inputs. Estimates of MEI sim/min demonstrated very large and large associations with actual MEI sim/min with models constructed from external and internal training inputs (r [90% confidence interval]; TRIMPsDist .73 [.72-.74], TRIMPsPL .72 [.71-.73], and sRPESkills .67 [.56-.78]). There were trivial differences in the precision of MEI sim/min estimates between models constructed from TRIMPsDist and TRIMPsPL and between internal input methods. Variable dose-response models from multiple training-load inputs can predict the within-individual variation of MEI sim/min across an entire preseason macrocycle. Models informed by external training inputs (TRIMPsDist and TRIMPsPL) exhibited predictive power comparable to those of sRPESkills models.