Body mass weighed far less on flat time-trial speed than on climbs: flat speed was best predicted by power divided by mass raised to only 0.34-0.48 (the authors back about W·m^−0.32), hill speed on 6% and 12% grades by power divided by mass^0.91, close to watts per kilogram

Nevill and colleagues asked whether the oxygen-to-mass ratios earlier found to predict flat and hill-climbing cycling speed carry over to power outputs that are quick to measure in the laboratory: maximal aerobic power (W_MAP), power at the ventilatory threshold (W_VT) and average power held in a 1-h performance test (W_AVG).

They fitted proportional allometric models linking each power measure and body mass to time-trial speed. On the flat, the best models were (W_MAP·m^−0.48)^0.54, (W_VT·m^−0.48)^0.46 and (W_AVG·m^−0.34)^0.58, explaining 69.3%, 59.1% and 96.3% of the variance in speed.

In cross-validation, W_MAP together with body mass explained 94.6% of the variance, with a standard deviation about the regression line of 0.686 km/h.

The authors read this as support for a flat-road ratio of about W·m^−0.32, as earlier VO2 work found; the power exponents (0.54, 0.46, 0.58), however, sat above the cube root (0.33) that physics would predict, which they say could be explained by other confounding factors such as bicycle geometry, "tractional resistance" or a tailwind.

For the 6% and 12% climbs, speed was proportional to (W_MAP·m^−0.91)^0.66, a mass exponent of 0.91 that also agrees with earlier research.