Purpose: This study aimed 1) to examine the accuracy with which W' reconstitution (W' REC ) is estimated by the W' balance (W' BAL ) models after a 3-min all-out cycling test (3MT), 2) to determine the effects of a 3MT on the power-duration relationship, and 3) to assess whether accounting for changes in the power-duration relationship during exercise improved estimates of W' REC .

Methods: The power-duration relationship and the actual and estimated W' REC were determined for 12 data sets extracted from our laboratory database where participants had completed two 3MT separated by 1-min recovery (i.e., control [C-3MT] and fatigued [F-3MT]).

Results: Actual W' REC (6.3 ± 1.4 kJ) was significantly overestimated by the W' BAL·ODE (9.8 ± 1.3 kJ; P < 0.001) and the W' BAL·MORTON (16.9 ± 2.6 kJ; P < 0.001) models but was not significantly different to the estimate provided by the W' BAL·INT (7.5 ± 1.5 kJ; P > 0.05) model. End power (EP) was 7% lower in the F-3MT (263 ± 40 W) compared with the C-3MT (282 ± 44 W; P < 0.001), and work done above EP (WEP) was 61% lower in the F-3MT (6.3 ± 1.4 kJ) compared with the C-3MT (16.9 ± 3.2 kJ). The size of the error in the estimated W' REC was correlated with the reduction in WEP for the W' BAL·INT and W' BAL·ODE models (both r > -0.74, P < 0.01) but not the W' BAL·MORTON model ( r = -0.18, P > 0.05). Accounting for the changes in the power-duration relationship improved the accuracy of the W' BAL·ODE and W' BAL·MORTON , but they remained significantly different to actual W' REC .

Conclusions: These findings demonstrate that the power-duration relationship is altered after a 3MT, and accounting for these changes improves the accuracy of the W' BAL·ODE and the W' BAL·MORTON , but not W' BAL·INT models. These results have important implications for the design and use of mathematical models describing the energetics of exercise performance.