Comment by bigiain

Completely uninformed speculation:

Wind drag goes up with v squared, so power required goes up with v cubed.

If you run at 105% speed downhill,that requires almost 16% more power to overcome wind drag. You might be better off running at 100% speed downhill (and "saving" that 16% power), and pushing harder to run as close as you can to 100% speed on the uphill stretches that would otherwise have you running slower than 100%. The power used to increase your potential energy going uphill is "zero sum" because you get it back when you go back downhill -n there no pesky v squared or v cubed non linearity there (assuming the race starts and finishes at the same elevation).