What is grade-adjusted pace?

How GAP is calculated

Every GAP model starts from a curve of how much harder running gets at each gradient. Divide your actual pace by that factor and you have the flat-ground equivalent. The curve gpx2pace uses comes from Minetti and colleagues (2002), who measured the energy cost of running on a treadmill at slopes from −45% to +45%. Here it is from −20% to +20%, as time per kilometre relative to flat:

The exact numbers for every gradient from −10% to +10% are in the pace by gradient table.

GAP after a run, and before a race

Training apps such as Strava show GAP after a run, to tell you how hard a hilly session really was. The same idea works in reverse for a race you have not run yet: pick one flat-ground pace, multiply it by the curve for every kilometre of the course, and you have a plan whose effort is level from start to finish.

That is what a graded pace plan is. Every split has the same GAP, even though the paces on the band look very different. Three miles of the Boston Marathon at 3:30:

The targets are 1:31 apart per mile; the effort is the same. See every mile on the Boston pace plan.

Where GAP stops being accurate

• Steep descents. The curve credits a fast pace downhill, but how fast you can actually go is limited by your legs, the footing and nerve, not by energy. Expect to fall short of the downhill numbers on steep or twisting roads.
• Very steep climbs and trails. On steep ground many runners hike, which a running curve does not describe, and rough surfaces cost more than the smooth treadmill the curve was measured on.
• Bad elevation data. GAP is only as good as the gradient it is given. Noisy GPS altitude or elevation that includes buildings and trees invents hills that are not there; the GPX guide shows how to check.
• Everything that isn’t a hill. Heat, wind, altitude and fatigue change your pace too. GAP leaves them out.

Models also disagree with each other. Strava’s model is fitted to heart-rate data from its users rather than lab measurements, and differs most on descents. On road gradients the disagreement is a few seconds per kilometre.