Prediction of 2000-m on-water rowing performance with measures derived from instrumented boats

Rowing instrumentation systems provide measures of stroke power, stroke rate, and boat velocity during rowing races, but how well these measures predict race performance has not been reported previously.

Data were collected per stroke from 45 2000-m races using Peach PowerLine and OptimEye S5 GPS units. The boat classes assessed were nine male singles, eight female singles, three male pairs, and six female pairs. Random effects and residuals from general linear mixed modeling of stroke velocity adjusted for stroke power, stroke rate, and mean headwind provided measures interpreted as technical efficiency, race conditions, and stroke-velocity variability. These measures, along with mean race power, mean stroke rate, and mean headwind were then included in multiple linear regressions to predict race velocity from official race times. Effects were assessed for 2 SD changes in predictors and interpreted using interval hypothesis tests.

Effects of mean race power, mean stroke rate, and mean headwind on race velocity ranged from small to extremely large and were mostly decisively substantial. Effects of technical efficiency and race conditions ranged from trivial to extremely large but were generally unclear, while stroke-velocity variability had trivial-small and mostly unclear effects. Prediction error was small to moderate and decisively substantial. Men's pairs lacked sufficient data for analysis.

On-water rowing race performance can be predicted with mean race values of power, stroke rate, and headwind. Estimates from stroke data are potentially useful predictors but require impractical numbers of boats and races to reduce their uncertainty.