Grade Adjusted Pace (GAP) Explained
Your watch says 6:30/km on that hill, but it felt like a 5:00 effort. GAP is the metric that bridges this gap — here's the science behind it, why platforms disagree, and how to actually use it.
- GAP converts hill running pace to a flat-ground equivalent, letting you compare efforts across courses with different elevation profiles.
- The foundational science comes from Minetti et al. (2002), who discovered that the metabolic cost of running is minimized at roughly -20% downhill grade — not on flat ground. (Many sources cite -10%, but that's actually the walking minimum.)
- In 2017, Strava moved beyond lab-based metabolic models to a heart rate equivalency model trained on 6 million runs from 240,000 athletes, producing more real-world-accurate GAP values.
- Garmin and Strava show different GAP values for the same run because they use different elevation data sources, cost models, and smoothing algorithms.
- GAP is most reliable for moderate uphills (+1% to +10%) but accuracy degrades significantly on steep downhills, where prediction errors can reach 3x.
Table of Contents
What is Grade Adjusted Pace?
When you run uphill, you slow down — not because you're trying less hard, but because fighting gravity demands more energy per meter. Your heart rate climbs, your breathing gets heavy, and your legs burn. But your watch only sees the slower pace. It can't tell whether you're jogging easily on a flat path or grinding up a 6% grade at lactate threshold.
Grade Adjusted Pace (GAP) answers a simple but powerful question: if you had been running on flat ground at the same physiological effort, how fast would you have been going? By accounting for the energy cost of elevation changes, GAP converts your hilly run into a flat-ground equivalent — making it possible to compare a 5:30/km hill repeat with a 4:45/km tempo on flat terrain and see that, physiologically, they were the same effort.
Every major running platform — Garmin, Strava, COROS, TrainingPeaks — computes some version of GAP. But they don't all agree, and understanding why requires looking under the hood at the science, the models, and the data they rely on.
The Science Behind GAP
The scientific foundation for GAP comes from a landmark 2002 study by Alberto Minetti and colleagues, published in the Journal of Applied Physiology. They placed 10 male trail runners on a motorized treadmill and measured oxygen consumption — a direct proxy for metabolic cost — at gradients ranging from -45% to +45%.
The results produced an energy cost curve that surprised many: the minimum metabolic cost of running doesn't occur on flat ground. It occurs at approximately -20% downhill. On flat terrain, the cost is about 3.6 J/kg/m. Near -20%, it drops to roughly 1.8 J/kg/m — half the energy expenditure. Go any steeper, and the cost rises sharply as eccentric braking forces become enormous. (A widely repeated claim puts the minimum at -10%, but that figure comes from the walking data in the same Minetti study — the running minimum is at a steeper downhill.)
On the uphill side, cost rises steeply and predictably. Each 1% increase in gradient adds roughly 3–4% to the energy cost. By +10%, you're spending about 40% more energy per meter than on flat. By +45%, the cost reaches nearly 19 J/kg/m — more than 5× the flat-ground cost.
Minetti Cost Formula
C(g) = 155.4g⁵ − 30.4g⁴ − 43.3g³ + 46.3g² + 19.5g + 3.6
where g = grade as a fraction (0.05 = 5% uphill), C = metabolic cost in J/kg/m
Minetti Cost Curve (Energy Cost vs. Grade)
Grade (%) → Cost (J/kg/m)
This 5th-order polynomial captured the nonlinear relationship between grade and metabolic cost across the full range of tested gradients. It became the basis for GAP calculations across the running tech industry.
How GAP is Calculated
In practice, computing GAP involves three steps applied to every ~1-second GPS record in your activity. First, the algorithm calculates the instantaneous grade — the elevation change divided by the horizontal distance traveled in that interval. Second, it looks up the energy cost adjustment factor for that grade. Third, it multiplies your actual running speed by that adjustment factor to estimate what your flat-ground speed would have been at the same effort.
The math is straightforward. If you're running at 3.0 m/s up a 5% grade, and the cost model says a 5% grade requires 1.17× the energy of flat running, then your GAP speed is 3.0 × 1.17 = 3.51 m/s — equivalent to about 4:45/km on flat, compared to your actual 5:33/km.
Sample GAP Adjustments
| Grade | Cost Factor | Actual Pace | GAP |
|---|---|---|---|
| -10% | 0.72× | 4:00/km | 5:33/km |
| -5% | 0.87× | 4:20/km | 4:59/km |
| 0% | 1.00× | 4:45/km | 4:45/km |
| +3% | 1.11× | 5:10/km | 4:40/km |
| +5% | 1.17× | 5:33/km | 4:45/km |
| +10% | 1.40× | 6:30/km | 4:38/km |
| +15% | 1.68× | 7:40/km | 4:34/km |
In practice, most platforms use a simplified version of the full Minetti polynomial. A common approximation is a quadratic model — a(g) = 1 + c₁·g + c₂·g² — where the linear coefficient captures the net grade effect and the quadratic coefficient captures the nonlinear cost increase on steep terrain. The exact coefficients vary by platform, which is one reason GAP values differ.
Elevation Data: The Hidden Variable
Before any cost model can compute GAP, it needs accurate elevation data — and this is where the biggest source of disagreement between platforms originates. There are three common sources of elevation data, each with distinct strengths and weaknesses.
Most modern running watches (Garmin, COROS, Apple Watch Ultra) include a barometric pressure sensor. It measures air pressure changes to estimate altitude, then calibrates against GPS-derived elevation. Barometric altimeters are responsive (they capture real-time micro-terrain like short hills and bridge overpasses), but they drift with weather pressure changes and can be affected by arm swing, body heat, and temperature shifts during the run.
Strava and some post-processing tools replace or augment device elevation with DEM data — satellite-derived terrain maps that assign an elevation to every point on Earth's surface. DEM data is consistent and doesn't drift with weather. However, DEM resolution (typically 10–30m grid) can miss narrow features like bridges, tunnels, and sharp grade changes. A 50m overpass may be completely invisible to DEM-based elevation.
Pure GPS altitude (without barometric correction) is the least reliable, with errors of ±15–20m common and worse in urban canyons. It's rarely used as the primary source for GAP calculations.
The key takeaway: the same 10km run can have noticeably different elevation profiles depending on whether you're looking at barometric, DEM, or GPS elevation data. These differences cascade directly into GAP calculations, since different elevation profiles produce different per-sample grade values, different cost factors, and ultimately different GAP results.
Why Garmin and Strava Show Different GAP
If you've ever compared the same run on Garmin Connect and Strava, you may have noticed that GAP values don't always match — sometimes differing by 5–10 seconds per kilometer. This isn't a bug. It's a consequence of fundamentally different design choices across four dimensions.
| Feature | Garmin | Strava |
|---|---|---|
| Elevation Source | Barometric altimeter (real-time sensor data) | DEM post-processing (satellite terrain map) |
| Cost Model | Metabolic polynomial (based on Minetti lab data) | Heart rate equivalency (2017 update by Drew Robb) |
| Training Data | Lab measurements (10 subjects) | 6 million runs from 240,000 athletes |
| Processing | Real-time (computed on device) | Post-processing (computed on server) |
| Tendency | Conservative — GAP closer to actual pace | Aggressive — GAP often significantly faster than actual pace |
Strava's 2017 shift was particularly significant. Their engineering team, led by Drew Robb, moved from a pure metabolic cost model (like Minetti's) to a heart rate equivalency approach. Instead of asking "how much more energy does a 5% grade require?", they asked "at what flat-ground pace would this runner maintain the same heart rate?" This is a subtle but important distinction — it factors in real-world running dynamics that lab treadmill tests may not capture.
Neither platform is definitively "more accurate." Garmin's barometric data captures micro-terrain that DEM misses, but is noisier. Strava's massive dataset provides statistical robustness, but their DEM elevation may smooth away real grade changes. For most road runners on moderate terrain, the differences are small — but on very hilly or technical courses, the disagreement can be substantial.
When GAP is Useful
Despite its imperfections, GAP provides genuine insight when used appropriately. Here are the three scenarios where it adds the most value.
You ran 5:15/km on a hilly route and 4:50/km on a flat route. Which was a better performance? Without GAP, it's hard to say. With GAP, you can normalize both runs to a flat-ground equivalent and see that the hilly run (GAP 4:48/km) actually represented a stronger effort.
When running by pace on a hilly route, it's easy to accidentally push too hard on climbs and recover too much on descents. Monitoring GAP (or using it post-run) helps ensure you stayed in the intended effort zone — especially valuable for easy and tempo runs on undulating terrain.
If all your training is on hilly routes, your average pace undersells your fitness. GAP provides a rough translation of what that training pace corresponds to on a flat course — useful for setting realistic goal paces for flat marathons or track workouts.
When GAP Falls Short
GAP is a model, and all models simplify reality. Here are the most important limitations to keep in mind.
Steep Downhill Accuracy
The Minetti model's own authors noted that downhill predictions can err by a factor of 3×. Running downhill requires eccentric braking that is biomechanically complex and varies enormously between individuals. Past -15% grade, GAP values should be treated with deep skepticism.
Trail Terrain Not Modeled
GAP accounts for grade, but not for rocks, roots, mud, sand, or technical footing. A technical trail at 5% grade is far more costly than a paved road at 5%. If you're a trail runner, GAP will consistently overestimate your flat-ground equivalent.
Individual Biomechanical Variation
Some runners are natural climbers — lighter body weight, higher cadence uphill, efficient vertical force production. Others struggle on hills but excel on flats. Standard GAP models use a one-size-fits-all cost curve that ignores these individual differences.
Environmental Factors Ignored
Wind, heat, humidity, and altitude all affect the energy cost of running independently of grade. A 5% climb at sea level in cool weather costs less than the same climb at 2,500m altitude in 35°C heat. GAP doesn't account for any of this.
Short, Steep Pitches May Be Smoothed Away
Both the device's elevation data and the platform's smoothing algorithms can flatten out short steep sections (a 20-meter bridge ramp, a brief set of stairs). If these grade spikes are smoothed away, the corresponding GAP adjustment disappears too.
How to Use GAP in Your Training
Here are four practical ways to incorporate GAP into your training analysis without over-relying on it.
If your easy run target is 5:30/km and you ran a hilly route averaging 5:50/km with a GAP of 5:25/km, you can be confident the effort was appropriate. Without GAP, that 5:50 might look like a slow day.
If you regularly run the same hilly route, track your GAP over weeks and months. Improving GAP with stable heart rate is a strong signal of growing aerobic fitness — one that raw pace might mask on undulating terrain.
On steep descents, GAP may show a pace slower than your actual pace (since the model thinks downhill should be "free"). Don't try to adjust your running to match a target GAP on downhills — focus on comfortable form and controlled effort instead.
The most robust analysis combines GAP with heart rate data. If your GAP and heart rate both suggest the same effort level, you can trust the assessment. If they disagree (e.g., GAP says easy but HR says hard), external factors like heat, fatigue, or cardiac drift may be at play.
How Hashiri.AI Calculates GAP
After studying how Garmin and Strava each approach GAP — and the trade-offs each makes — we built our own implementation. Here's exactly how it works, step by step.
Per-Second Record-Level Calculation
We calculate GAP at every 1-second GPS record, not at the lap level. This captures short uphills and descents that lap-level averaging would smooth away, giving a more accurate picture of the effort within each lap.
Elevation Smoothing (±2 Sample Window)
Raw barometric altitude is noisy. We apply a moving average over ±2 samples before computing the grade at each point. Too little smoothing and GPS jitter creates phantom grades; too much and real terrain features disappear.
Simplified Quadratic Cost Model
Rather than the full Minetti 5th-order polynomial (which can over-correct on steep downhills), we use a quadratic approximation that balances accuracy with stability across real-world conditions.
Tuned Against Real Garmin Data
We calibrated our coefficients by comparing output against Garmin's GAP values on the same activities — matching closely on moderate terrain while avoiding Strava's more aggressive adjustments on steep grades.
Hashiri.AI GAP Formula
adjustment(g) = 1 + 0.033 × g + 0.0025 × g²
where g = grade in percent. GAP speed = actual speed × adjustment factor. Positive adjustment (uphill) → GAP faster than actual; negative adjustment (downhill) → GAP slower than actual.
The result is a GAP that closely matches Garmin on moderate terrain, stays conservative on steep downhills where all models lose accuracy, and gives you a reliable way to compare hilly efforts against flat ones.
Frequently Asked Questions
Is GAP the same as NGP (Normalized Graded Pace)?
They target the same concept — normalizing pace for elevation — but use different names and algorithms. Garmin and Strava use "GAP," while TrainingPeaks uses "NGP" (Normalized Graded Pace). The underlying models and elevation data sources differ, so values won't match exactly across platforms.
Why is my GAP faster than my actual pace even on a net-flat course?
Even "flat" courses have micro-undulations — subtle elevation changes that your barometric altimeter detects. Due to the nonlinear cost model (uphill costs more than downhill saves), any undulation will result in a GAP that's slightly faster than actual pace. This is a mathematical property of the cost curve, not an error.
Can I use GAP for treadmill running with incline?
Yes — in fact, treadmill GAP can be more accurate than outdoor GAP because the grade is precise and constant. A treadmill at 3% incline will produce a GAP that's meaningfully faster than your displayed pace, reflecting the extra effort of climbing.
Why does Strava show a very different GAP than my Garmin?
Two main reasons: different elevation data (Garmin uses barometric; Strava uses DEM correction) and different cost models (Garmin uses metabolic polynomial; Strava uses heart rate equivalency from 6M runs). The differences are most pronounced on routes with frequent short grade changes.
Should I pace my races using GAP?
GAP is useful for pre-race planning (estimating effort across a hilly course profile), but should not be your sole pacing guide during the race. Use it alongside heart rate and perceived effort. On race day, trust your body more than the number — GAP can't account for race-day adrenaline, heat, or cumulative fatigue.
Calculate Your Paces
Use our free tools to explore what your training paces mean across different distances and efforts.
Open Pace Calculator