Originally posted by

**JRM**
The purpose of that function is to model the initial drive phase of the athlete. The constant b is chosen to fit the observed velocity curves, and as a result its effect largely goes away by about 1 second into the race. Earlier models didn't include it, but did use non-exponential, velocity-dependent decay terms (see e.g. Keller's optimization model). The problem is that the latter produces unrealistic speed curves.

A velocity-dependent term would certainly be well-motivated from a biomechanical point-of-view, as it would more appropriately represent the transition in posture during the early drive (since velocity is a universal determinant in gait-related motion). The flip side of this argument is that during this phase the cross-sectional area of the athlete is smaller than for upright running, and the athlete's velocity is well under 10m/s. So, drag effects are much less important.

I would be surprised if changing that drive term would alter the results in any significant way (for the averaged winds). But that being said, one could always investigate it. It would be a relatively easy thing to check out. If you're up to it, email me.

A velocity-dependent term would certainly be well-motivated from a biomechanical point-of-view, as it would more appropriately represent the transition in posture during the early drive (since velocity is a universal determinant in gait-related motion). The flip side of this argument is that during this phase the cross-sectional area of the athlete is smaller than for upright running, and the athlete's velocity is well under 10m/s. So, drag effects are much less important.

I would be surprised if changing that drive term would alter the results in any significant way (for the averaged winds). But that being said, one could always investigate it. It would be a relatively easy thing to check out. If you're up to it, email me.

## Leave a comment: