Stryd 5 vs Stryd 4 Accuracy Test: Quantifying the Differences

I’ve been using Stryd 5.0 for a few weeks now. I like the new, smaller design and dual-sided charging pod. However, the reported data does feel different, so I decided to go ahead and do some science, having devised what I thought was a rigorous test to validate the claims, or at least some of the more important ones.

Let’s take a step back to figure out what claims I can validate.

What’s Changed? What are the claims?

Here’s a recap of what has changed from Stryd 4.0 to 5.0.

All the changes are listed. I highlighted the data-related changes in bold and underlined those I thought I could validate.

Test 1 – 12 hill reps plus 9km over varied terrain

I mistakenly began with a hill test on a compacted trail, later realising that methodological issues may have affected the results. It’s a 2-minute hill in Richmond Park. I could reliably hold a steady 300 W for each effort, aiming to keep the power approximately constant despite the uneven changes in gradient. It began with a gentle gradient, then flattened, and finally steepened markedly at the end.

At my disposal were 2 Stryd 5.0 pods (Model 30) and 2 Stryd 4.0 Pods (Models 26, 27; Model 27 is assumed to be from a later production batch but otherwise identical). I chose to compare the Model 30 to the Model 27. I have a small amount of gait asymmetry, so I switched the pods after each repetition. I recorded the Stryd 5 with a Garmin Forerunner 970 and the Stryd 4.0 with a Suunto Race.

I planned to let the Garmin record the elevation gain second by second from the same starting point. Overall, I could determine the average gradient profile across all 12 runs and examine how power responded to that profile across the tests.

The hill was approximately a 9 km round trip from my home, so I also conducted additional tests on various surface types (wet, soft grass, and variously compacted trails) along the route to and from the hill. In hindsight, I’m not entirely sure what I was trying to prove in these sections! Other than the variance in the raw, uncalibrated distance between the pods.

Anyway, here’s the data. After the test, I encountered several unforeseen issues, including watches with subtly different timestamps, potential delays in recording to the FIT file, and similar factors, as I used two different watch models that might have affected the results. e.g., I could simply align the data by the time the power started, but if there were even a small timestamp difference, I could be correcting a real difference in response time between pods.

Master Verification Table

Validation Checks

• ✓ 18 tests extracted
• ✓ Tests 4-15 durations: 120s – 129s (expected 120-130s)
• ✓ Hill elevation gains: 23.4m – 24.2m (very consistent)
• ✓ Power range: 255W – 317W
• ✓ Foot alternation pattern correct (L-R-L-R…)

Hill Profile (from Garmin GPS)

• Distance: ~380m
• Elevation gain: ~24m
• Average gradient: ~6.3%
• Profile: Starts gently (3-5%), steepens mid-hill (8-10%), eases at the top

Preliminary Findings (Controlling for Gait Asymmetry)

* Smaller difference than subsequent test may reflect methodological limitations (different watches, timestamps, and use of only one pod per generation rather than two)

** time from start of effort to reaching 90% of average power for that rep

Test 2 – Validating 4 pods with each other, and as pairs

Back to the drawing board. Were same-generation pods different? Were the recording methods equivalent?

I decided to validate all 4 pods against one another in a controlled environment, thereby revealing unit-level variation within and between generations. I decided to obtain the raw data directly from the pod (via the Stryd app) rather than from four watches. This is possible but not straightforward, requiring four smartphones and a test Stryd account for each. Luckily, I remembered to calibrate them identically, which essentially means assigning the same runner weight and height in each app. The watch calibration factor is for distance, not power.

To cut to the chase, I couldn’t do it, and I’m not the sort of person who gives up easily, but you can only spend so much time with these things.

By one of the most fortuitous flukes ever, my bike fitter, who lives nearby (the best in SW London, BTW), was just so happening to be writing an app to help riders quickly understand the differences between their normal road power meter and their winter indoor trainer. Inevitably, many people must just forget about differences that are often far greater than I might notice when I do bike power meter tests (proper calibration, drive chain loss minimisation, cog choice, gear choice). I persuaded him to increase the number of power meters his app could record from 3 to 4, and I was sorted. He could let me export the data as 4 FIT files to use in Dcrainmaker’s Analyzer for easy visualisation, but, more importantly, they all shared the same time stamps when exported as a single CSV file. With the CSV file, I could then simply have an AI statbot analyse it.

To beta test the Freespeed app, contact Richard.

Q: How can you validate accuracy when there is no gold standard?

A: You can’t. But you can assess precision.

Method: Four pods (2× Stryd5, 2× Stryd4) were worn simultaneously during two 5-minute treadmill runs at a constant speed, with jumps on and off to simulate near-instant changes in power. Pods were swapped between feet between tests to isolate calibration differences from gait asymmetry.

Findings:

Conclusion: Both generations perform well with equivalent responsiveness to start/stop transitions. Within-generation precision is tight (±1%), but there is a systematic ~4% offset between generations (Stryd5 is higher). Stryd4 produces a smoother signal, though it is unclear whether this reflects greater stability or more aggressive smoothing.

Test 3 – Validating how 4 Pods respond to a constantly rising treadmill incline

Method: Four pods (2× Stryd5, 2× Stryd4) were worn simultaneously during six 60-second treadmill intervals at a constant belt speed. Each interval: 20 seconds of stabilisation at flat, then an incline ramped to maximum (approx 15°) over 40-45 seconds. Pods were swapped between feet after interval 3 to control for foot placement effects.

Responsiveness to Incline Change (Treadmill Ramp Test)

Note that the Baseline (flat) power is precise (±1% relative to Test 2); however, the incline power values cannot be validated because Stryd requires treadmill incline calibration. These power levels will differ from outdoor power (which incorporates barometric pressure and wind inputs), but this test concerns only responsiveness.

Conclusion: Stryd5 detected incline-induced biomechanical changes approximately 4 seconds faster than Stryd4, supporting Stryd’s claim of improved responsiveness. Stryd5 also reported approximately twice the power increase for the same gradient change, indicating greater sensitivity. However, absolute power values during incline cannot be validated without a reference standard, and outdoor performance may differ due to the absence of barometric pressure input on a treadmill.

Testing Issues

For clarity and completeness, here are some of the issues with my testing. Happy to add to it if you point out errors in the comments

• My Pod 4 were not bought as a pair. They have different model numbers.
• I used the same pair of shoes for all tests
• There was no wind in the outdoor test
• Pod positions AND cradles were potentially moved laterally as pods swapped feet. They were always through the same laces.
• Pricise Pod position on laces varied.
• Pods were always stacked lengthwise on the foot (as per the image at the top). It’s quite posisble foot position made a difference in power and distance.
• My family’s inexpensive treadmill seems to cause lower-than-expected power output. Meaning a dampening/cushioning effect might be in play
• While my outdoor test was at a reasonable level (300w-ish), the internal test was below reasonable levels (180w – I didn’t know how many I was going to have to do and was slightly injured and tired from a long bike ride)
• Stryd point out that, “The hill responsiveness improvements are only available outdoors, as indoor responsiveness on an incline is determined by the incline input.“

Grand Conclusion

Claim 1: “More accurate & consistent power, pace and distance”

• Accuracy: Stryd has clearly updated the power algorithm in Stryd5—the data is demonstrably different (~4% higher power readings). However, it remains unclear whether this represents improved accuracy, as there is no reference standard for running power.
• Precision: Both generations showed equivalent within-generation precision of ±1%. No improvement demonstrated.
• Signal consistency: Stryd4 produces a smoother signal (~50% less second-to-second variation). Whether Stryd5’s higher variation reflects reduced smoothing (potentially capturing more real biomechanical fluctuations) or simply noisier data cannot be determined from these tests. To fully assess the “more consistent” claim, testing would require multiple conditions (surfaces, gradients, fatigue states) and comparison with a reference standard that does not currently exist for running power.
• Distance: Not rigorously tested, though Test 1 showed Stryd5 reporting ~3% longer distances than Stryd4. Without a known reference distance, accuracy cannot be verified.

Claim 2: “Faster power responsiveness on varying terrain, climbs and descents”

• Supported. Stryd5 detected incline changes ~4 seconds faster and showed ~2x greater sensitivity to gradient changes. Test 1 outdoor data also showed that Stryd5 reached 90% peak power ~2 seconds earlier.

Practical takeaway: Users upgrading from Stryd4 to Stryd5 should expect higher power readings (~4%) and faster response to terrain changes. A less-smooth real-time signal will primarily affect users viewing raw (1-second) power; those displaying 3-second smoothed power or longer moving averages are unlikely to notice any difference in signal stability. Precision remains equivalent between generations. Whether the different power values represent improved accuracy cannot be determined.

Caveat: This testing used n=1 runners and 4 pods (2 per generation). Your pods and biomechanics may produce different results—the ~4% offset and other findings may not generalise to all units or all runners.

A note on accuracy claims: Stryd previously claimed accuracy for Stryd4; now Stryd5 is claimed to be “more accurate”, yet reports ~4% higher power. Without a gold standard for running power, neither generation’s absolute accuracy can be verified—this is a limitation of the field, not a criticism of Stryd specifically. The 5.3% pod-to-pod spread observed across four units (two per generation) is notable, though this testing is limited: one runner, four pods, and a small number of test conditions. Stryd’s internal testing with larger sample sizes may show tighter tolerances. Users should treat Stryd power as a consistent internal metric for training rather than an absolute measure of mechanical work.

Note: An ENTIRELY OPTIONAL subscription is not required by most and costs GBP8.75/mo when paid annually. For racing or training for HYROX, you don’t need a subscription, nor do you need two pods. You don’t even need a Stryd 5.0; a 4.0 is sufficient if you can find one for sale (be mindful of older second-hand units, as the battery may have degraded).

Last Updated on 17 February 2026 by the5krunner

the5krunner

tfk is the founder and author of the5krunner, an independent endurance sports technology publication. With 20 years of hands-on testing of GPS watches and wearables, and competing in triathlons at an international age-group level, tfk provides in-depth expert analysis of fitness technology for serious athletes and endurance sport competitors.