I realize W/kg aren’t directly used, although of course equations can be recast in terms of W/kg, or whatever else. That said, W/kg is a convenient parameter which is particularly closely correlated to climbing speed. I claim W²/kg is better for flat speed.
That Keuzotter website is excellent. Here’s the estimate for CdA:
This calculates the cross-section of an inclined rider. My crude model is simpler, the case where λ = π/2. In this case, sin λ = 1, and the m / (h · ρ) terms cancel. And sure enough, just as I claimed, CdA ≈ sqrt(h·m). If I divide watts by CdA I get something proportional to watts/sqrt(kg), or I can use watts²/kg as a benchmark for riders of a given height.
The m / (h · ρ) terms give the correct answer if the rider is fully recumbent. In that case, cross-sectional area matters, not height directly.
So for example, consider the following two riders:
A: 400 watts, 80 kg
B: 300 watts, 50 kg
which is faster on the flat? If I square the watts/100, I get 16/80 for the first rider = 0.2, and for the second, I get 9/50 = 0.18. So the first rider is faster.
If i look at W/kg, the first rider = 5, the second rider = 6, so the second rider has more W/kg.