How to Actually Read Running Power (Stryd) Data: What Watts Do and Don't Tell You

You're two kilometers into a tempo run and the number on your watch reads 318 watts. Is that good? Your training partner, running next to you at what feels like the same effort, glances at his own device and calls out 280. Neither of you slows down, but now you're both quietly recalculating whether you're overcooking the workout or he's sandbagging it. This exact moment — a number with no obvious ceiling, no obvious sense of what counts as good, and apparently no agreement between two devices measuring the same two bodies running the same pace — is where most runners either give up on power data entirely or start treating it like gospel. Both reactions skip the part that actually matters.

A running power meter isn't a smaller version of a bike power meter, and the number it shows isn't interchangeable between devices the way pace or heart rate roughly is. It's a modeled estimate, it's useful for a narrower set of things than most marketing implies, and it's genuinely valuable for exactly one thing pace and heart rate can't do on their own: telling you how hard you're working when the terrain or the wind is actively lying to your GPS. Here's what the number is built from, what two peer-reviewed studies actually found when they tested it, how to establish your own reference point with a simple two-test protocol, and when to trust watts over everything else on your wrist.

A bicycle power meter usually has a strain gauge bonded to the crank or pedal spindle — it measures the actual torque you apply and multiplies it by cadence. That's a direct physical measurement. A running power meter, whether it's a foot pod like Stryd or a wrist- or chest-based system doing the same math from accelerometer data, has no equivalent force sensor at the ground. There's no strain gauge in your shoe reading how hard your foot pushes off. Instead, the device tracks motion — vertical oscillation, ground contact time, leg spring stiffness, forward acceleration, and on some devices an onboard estimate of wind resistance — and runs that motion data through a biomechanical model to estimate the metabolic or mechanical power output that would produce it.

That distinction matters more than it sounds like it should. A modeled estimate is only as good as the model, and different manufacturers use different models, different sensor placements, and different assumptions about things like running economy and air resistance. Two devices watching the identical stride can legitimately disagree, not because one is broken, but because they're solving a similar problem with different equations and calling the output by the same name.

This is the part most runners never get told plainly: your absolute wattage number is not a portable unit. Cerezuela-Espejo, Hernández-Belmonte, Courel-Ibáñez, Conesa-Ros, Mora-Rodríguez, and Pallarés (2021), writing in European Journal of Sport Science, tested five commercially available running power technologies against each other during identical treadmill and track sessions. Within a single device, repeat trials were reasonably consistent — a good sign for tracking your own trend over time. Between devices, though, the picture fell apart: systems disagreed on absolute power output by a wide enough margin, and in a few pairings the disagreement was inconsistent enough in direction, that the authors concluded the field wasn't yet ready to treat running power as an interchangeable, standardized metric the way cycling power largely is. The limitation worth flagging is that this was a controlled treadmill/track protocol with a specific runner sample — it doesn't tell you the exact size of the gap you'd see between your specific watch and your partner's specific foot pod, only that you shouldn't assume there isn't one.

The practical upshot: a same-device trend is trustworthy. A cross-device comparison — your watts versus a training partner's, or your watts versus a number you saw quoted online for elite marathon pace — usually isn't, unless both of you are on the identical hardware and firmware. Stop comparing your number to his. Start comparing today's number to your own number from three weeks ago, on the same device, at the same effort.

Where running power earns its keep is inside your own data, not across devices. Austin, Hokanson, McGinnis, and Patrick (2018), publishing in Sports, put well-trained distance runners through submaximal treadmill trials while measuring both Stryd-derived power output and oxygen uptake — the gold-standard measure of running economy. They found a strong positive relationship between power output and metabolic cost at a given speed: runners who were metabolically less economical at a given pace also produced higher power numbers at that same pace. In other words, power output at a fixed speed behaves like a proxy for how efficiently you're running, not just how fast. The caveats are real — a sample of twelve runners, treadmill-only conditions with no wind or elevation variable for the algorithm to account for, and results tied to one specific device and firmware version — so treat the relationship as directionally solid rather than as a precise conversion formula you can apply to any runner on any device.

The training use is straightforward even without a lab: log your power output at a fixed, repeatable easy pace — say, your usual recovery-run pace on a flat familiar loop — every couple of weeks. A downward drift in watts at that same pace, same rough conditions, is a decent signal that your running economy is improving even before a race result confirms it. A creeping upward drift at that same pace, especially alongside normal or lower heart rate, is worth treating as an early fatigue flag rather than noise, because pace alone won't show it — GPS pace is externally fixed by the clock and the road, and it has no opinion on how much internal cost you paid to hit it.

The most useful single reference point you can extract from power data is your own critical power (CP) — the highest power output you could theoretically sustain for a long time before fatigue accumulates faster than you can clear it. The concept traces back to Monod and Scherrer's 1965 two-parameter model of muscular work, and Vanhatalo, Jones, and Burnley (2011), reviewing the framework in the International Journal of Sports Physiology and Performance, laid out why it holds up as a robust threshold concept across endurance sports: above CP, a finite, non-renewable reserve of work capacity — called W′ (W-prime) — starts depleting, and once it's gone, pace collapses regardless of how much aerobic fitness is left in the tank.

You don't need a lab to estimate both numbers. A simple two-effort field protocol, run on flat ground with a reliable warm-up beforehand, gets you close enough to train from:

With both efforts recorded, the two-parameter model solves for CP and W′ directly:

CP = (P1 × t1 − P2 × t2) / (t1 − t2)
W′ = t1 × (P1 − CP)

Worked example: a runner posts 305W average over the 12-minute trial (P1, t1 = 720s) and 345W over the 3-minute trial (P2, t2 = 180s). CP = (305×720 − 345×180) / (720−180) = (219,600 − 62,100) / 540 ≈ 292W. W′ = 720 × (305 − 292) ≈ 9,600 joules. That 292W becomes the anchor for every training zone below — and the 9,600J tells you roughly how much above-threshold work capacity you have in the bank before a hard interval session forces a collapse in pace.

Once you have a CP number from your own device, a five-zone framework — the same basic structure power-based cycling training has used for years, adapted to running — turns it into something you can actually train off. These boundary percentages are a common coaching convention rather than a fixed, universally validated standard, so treat them as a sensible starting point to adjust from, not a law:

Retest CP every 6–8 weeks during a normal training block, or sooner if you've just come off a dedicated speed phase or a long injury layoff — CP moves with fitness, and last season's number quietly stops describing this season's body.

Power's real advantage shows up exactly where pace and heart rate get unreliable. On a hill, GPS pace slows on the way up and speeds up on the way down, telling you almost nothing about whether the effort itself is steady — power, by contrast, should hold roughly level through a controlled climb if you're pacing it correctly, and a spike in watts on an uphill segment is a much earlier warning that you've gone too hard than a slowing pace number, which only tells you after the damage is partly done. On a windy day, a headwind will slow your GPS pace at a constant effort while a device with a wind-resistance term should show power holding closer to your target zone; a tailwind does the reverse. Heart rate lags both of these by tens of seconds to minutes, so in short surges it's telling you about effort from a while ago, not right now.

Indoors, remember the wind term drops out entirely — a treadmill run at the same power as an outdoor run in a headwind will read differently even though your legs are doing comparable work, because the algorithm has nothing to correct for.

The single most common mistake is the one that opened this guide: comparing your absolute watts to a training partner's number from a different device and drawing a conclusion about who's fitter. Per Cerezuela-Espejo et al. (2021), that comparison isn't measuring what you think it's measuring — it's comparing two different models, not two bodies.

Second is trusting an internet-sourced power chart as a race predictor without ever running your own CP test. Your zones are only as good as the CP number they're built from, and a generic chart built on someone else's device, someone else's biomechanics, and someone else's algorithm version tells you almost nothing about your own physiology.

Third is ignoring pod placement and calibration drift. A foot pod that's been moved to a new pair of shoes, swapped to the opposite shoe, or left with a low battery for weeks can shift your baseline enough to look like a fitness change when it's really a hardware artifact — if your watts-at-easy-pace jumps sharply overnight with no obvious training explanation, check the hardware before you panic about your fitness.

Fourth is applying the model to efforts it wasn't built for — short, high-acceleration reps like strides, hill sprints, or plyometric-style drills, and technical trail terrain with constant stopping and starting. The underlying accelerometer model assumes a reasonably steady running gait; erratic, start-stop movement patterns produce power numbers with far more noise than signal, and a single strange-looking spike on a technical trail descent is not evidence of a training breakthrough or a problem.