Overground vs. Treadmill Running Speed: Why Your Numbers Never Quite Match

A masters runner at a club in Denver spent six weeks convinced his aerobic fitness had collapsed. His gym treadmill said he was holding an easy 5:20/km pace at a heart rate that used to feel like tempo effort. Then he ran the same effort on the track behind his kid's school with a stopwatch and a set of cones every 400m: 4:52/km, comfortably. Nothing had changed in his legs or his lungs. What had changed was which number he was trusting.

This mix-up happens constantly, and it's rarely about fitness. A treadmill display and an outdoor pace reading come from two measurement chains never guaranteed to agree, and the gap widens with a machine's age and calibration. Add the biomechanical differences between running on a moving belt and ground that stays still, and comparing the two numbers directly is comparing apples to a fairly specific orange.

Below: what each number represents, why a belt can drift from its own display, what happens biomechanically even on a perfectly calibrated belt, what the research has measured, and a protocol for checking your own machine instead of guessing.

Overground speed is a distance-over-time measurement against a fixed, non-moving reference: a measured track, timing gates, or a GPS/GNSS receiver sampling position relative to satellites at 1 to 10 times per second. Whatever error exists comes from the measurement tool itself - GPS multipath under tree cover, a timing gate mistriggered by an arm swing - not from anything mechanical moving underneath the runner.

Treadmill speed is a different kind of number entirely. The console doesn't measure how fast your body moves through space. It reports how fast the drive motor's encoder believes the belt is spinning, converted using the roller's circumference and a calibration curve set at the factory or last service visit. That number stays fixed on the display regardless of what's actually happening to the belt surface under your feet that session. Comparing an outdoor GPS pace to a treadmill display means comparing a direct measurement of your own displacement to an inferred estimate of a motor's rotational speed - two different objects, much like a radar gun and a bat sensor measure two different points in the same swing.

A drive motor's encoder measures its own shaft rotation with excellent precision. What it can't see directly is belt slip - the small give between the belt surface and the front roller under load, especially at footstrike, when a runner's weight transfers onto the belt hardest. If belt tension has loosened, the deck has worn unevenly, or the drive belt connecting motor to roller has stretched with age, the belt can move measurably slower than the encoder-derived number the console shows, even while the motor spins exactly as commanded.

That's why speed and grade are supposed to be checked against an independent reference on a fixed schedule rather than trusted indefinitely - a practice built into clinical exercise-testing protocols precisely because console displays are known to drift. A freshly serviced unit typically tracks true surface speed within roughly 1 percent. Field checks on treadmills well past a service interval - loose belts, worn lubrication, an aging drive belt - regularly turn up errors in the 3-7 percent range, with unmaintained consumer units sometimes worse at higher speeds. None of this trips an error message; the display keeps confidently reporting motor rotation, not what the belt surface is doing.

Incline compounds it independently: a display reading 1.0% grade can mechanically sit anywhere from roughly 0.4% to 1.6% depending on the incline motor's own calibration, so correcting for outdoor conditions with added incline may stack a correction on top of a number that's already off.

Even a treadmill tracking true belt speed within a fraction of a percent doesn't reproduce overground running mechanically or energetically, for reasons that have nothing to do with calibration. Outdoors, a runner generates forward propulsive force to overcome air resistance and keep driving their own center of mass down the road. On a belt, the surface itself moves backward under the stance foot; the runner's job shifts toward maintaining position over that moving surface rather than continuously overcoming drag, which changes propulsive force output and lower-limb joint loading even at a matched, correctly calibrated speed.

Missing air resistance has a measurable energy cost, not just a biomechanical one. At training and racing paces, air resistance accounts for a meaningful share of the total metabolic cost of running outdoors - a cost a level (0% grade) treadmill doesn't ask the runner to overcome, which is why the same displayed speed can feel physiologically easier indoors even on a machine tracking true belt speed perfectly.

Riley, Dicharry, Franz, Della Croce, Wilder and Kerrigan (2008) put 20 experienced runners through matched-speed trials overground and on an instrumented treadmill, capturing 3D kinematics and kinetics on both surfaces. Form looked broadly similar, but consistent differences showed up: greater peak hip extension and ankle dorsiflexion overground, plus shorter ground contact time and small stride-length shifts on the treadmill at identical displayed speed - a few percent in spatiotemporal variables, several degrees in joint angles. The limitation: a single treadmill model with trained runners at experimenter-set speeds says how one well-maintained lab treadmill compares to overground, not how a given gym machine with its own belt stiffness behaves.

Van Hooren, Fuller, Buckley and colleagues (2020) addressed that gap with a systematic review and meta-analysis pooling numerous crossover studies. Stride length and cadence came out broadly comparable between surfaces, with small pooled effect sizes. But kinetic variables - including impact loading rate and propulsive force patterns - showed small-to-moderate, sometimes inconsistent, differences. The authors flagged the limitation that matters most for any individual runner: heterogeneity between studies was high, driven by treadmill brand, belt stiffness, and familiarization time, so pooled averages can mask a meaningful gap on any specific machine.

Jones and Doust (1996) approached the question from the energy-cost side, measuring oxygen uptake at matched speeds outdoors and on a level treadmill. Running at 0% grade indoors consistently underestimated the metabolic cost of the same outdoor pace, since the treadmill never required overcoming air resistance. Roughly 1% grade closed that gap at the speeds tested - a bounded range of controlled lab speeds without wind, so it's a solid starting point rather than an exact match at every pace or wind condition.

Put four measurement methods on the same effort and each one reports a number built from a different underlying signal, with a different error profile attached to it.

The console-display row and the calibrated row describe the same physical machine reporting two different accuracy pictures - the only difference is whether anyone checked the display against actual belt speed recently. That gap, not the outdoor-versus-indoor comparison that gets all the attention, is where most unexplained pace drift at gyms comes from.

Most panic over a mismatched pace traces to one of these habits, not a fitness change.

• Assuming a treadmill display equals true belt speed. It's a motor's calculated estimate, not a direct measurement of the surface under your feet - and it drifts with service history.
• Comparing treadmill pace to race pace with no incline correction. Without roughly 1% grade, a level treadmill understates the metabolic cost of the same outdoor pace because it removes air resistance entirely.
• Blaming fitness for a gap that's actually a different machine. Two treadmills at the same gym can carry meaningfully different belt tension and calibration age; switching machines between sessions can move your displayed pace at identical effort.
• Trusting GPS pace under tree cover as gospel. Signal quality swings accuracy from roughly 1-3% error to well over 5% in the same run, which can look like a fitness swing session to session.
• Never rechecking calibration after a treadmill is moved or serviced. Relocating a unit or replacing a belt can shift speed and incline calibration even when nothing on the display changes.

A short, repeatable check turns the question into documented data instead of a guess.

1. Measure belt loop length once - the full circumference of the running surface, in meters.
2. Run the belt empty at a fixed low speed (around 4.0 km/h) and time exactly 60 seconds while counting full revolutions past a marked point.
3. Calculate true speed as revolutions times belt length divided by elapsed time, and compare it to the console's displayed speed at that setting.
4. Repeat at two or three higher speeds relevant to your training paces, since drift isn't always uniform across the range.
5. Flag anything over roughly 3% difference for service rather than training off the display as if it were exact.
6. Add about 1% incline when using a level treadmill to simulate outdoor race pace, per the energy-cost correction the research supports.
7. Re-baseline whenever the machine changes - a different unit, gym, or servicing visit resets belt characteristics.
8. Cross-check outdoor sessions with a measured-distance segment periodically, rather than assuming every GPS pace reading carries the same accuracy regardless of route or tree cover.

These ranges reflect typical field-testing experience, not a hard pass/fail line for any one machine.

None of these ranges are a defect by themselves - they're expected behavior under real-world conditions. What's worth flagging is a gap larger than these ranges, or one grown noticeably since the last check, since either points at an actual service or signal issue.