The method
How we rank
Every score on this site comes from the weights below, applied identically to every device. Here they are, so you can check the ranking instead of trusting it.
Last reviewed July 24, 2026
Why this page exists
Every publisher in this category ranks devices for accuracy. Almost none publishes what the ranking is made of. Scores appear out of ten with no stated weights; "most accurate" appears in headlines with no measurement behind it. Those rankings are unfalsifiable by construction — you cannot check them, so you can only believe them or not.
We are a brand-new site with no test lab, no owned hardware and no credentialed reviewer. We cannot compete on authority we have not earned. What we can do is publish a method so specific that a stranger could reproduce our rankings and tell us where we went wrong. That is the entire strategy, and this page is the whole of it.
1. The weights
Five factors, one weighting, applied to every device on the site — a chest strap, a $1,000 multisport watch and a sleep ring are all scored on the same five axes. The overall score is computed from these weights in code, not typed in by hand, so if a sub-score changes the headline number moves with it automatically.
| Factor | Weight | What it measures |
|---|---|---|
| Evidence | 30% | How good the published validation is for this device's headline measurement — peer-reviewed against a gold standard at the top, nothing at all at the bottom. |
| Hardware | 20% | Sensor and GNSS generation, taken from the manufacturer's own specification sheet. |
| Metric coverage | 20% | What the device actually measures, and whether the numbers leave its own app. |
| Endurance | 15% | Manufacturer-stated battery life, labeled as manufacturer-stated, in the mode the figure actually refers to. |
| Cost per capability | 15% | Arithmetic from the live price — dollars per GPS hour, or true three-year cost where a subscription is mandatory. |
Evidence carries the largest single share, at 30%, because the premise of this site is that a ranking should follow the validation record rather than the marketing. It is also why a $105 chest strap can outrank a $750 watch on our most accurate page: the watch scores higher on hardware and coverage, and loses the axis that counts for most.
Every ranked product on this site has an expandable "show how this was calculated" panel listing its five sub-scores and the one-line basis for each. If you disagree with a number, you can see exactly which one it is.
2. The evidence hierarchy
The Evidence sub-score is set by what kind of support exists for a device's headline measurement. In descending order:
A peer-reviewed or hospital validation against a gold standard — polysomnography for sleep, ECG or a validated chest strap for heart rate, a measured course for distance. The strongest thing that can exist.
A named, linked, independently measured test. Not peer reviewed, but the method is published and checkable.
The manufacturer's own published specification sheet or manual. Reliable for battery, weight and hardware. Says nothing about accuracy.
A marketing assertion with no published method behind it.
No validation of any kind exists that we could find. Stated out loud, on the page, next to the device.
The consequence is deliberate: a device with no independent validation cannot take a top accuracy rank on hardware alone. The clearest case on this site is the Oura Ring 5 — newer than the Ring 4, probably better, and ranked below it, because the peer-reviewed result belongs to an earlier generation and a new sensor running a new algorithm is exactly what a validation exists to test.
The same rule costs the Suunto Race 2 and the COROS PACE 4 places they might otherwise take. We think that is correct. If it is published, its score moves.
3. Dating every study
A 2020 validation of a 2019 sensor does not describe a 2026 device, and quietly citing a dated study as though it were current is the most common way this category misleads without technically lying.
So every source on this site carries its publication year, and where a study tested a superseded model we say which model it tested. The best distance-accuracy paper we have is from 2020 and tested single-band watches — we cite it constantly, and we say every time that it describes engineering history rather than any watch you can buy.
4. The disqualifiers
Some things stop a device being recommended regardless of its score:
- No verifiable live price.If our price layer cannot confirm a current price, no number renders — we show "Check price" rather than a remembered one.
- An unresolvable product identity.If we cannot confirm via the retailer's API that a listing is the model we are discussing, it does not get a buy button. A wrong ASIN sells the reader a different product.
- A safety or medical claim we cannot verify. We do not repeat clinical claims we have not read the evidence for, and no device here is presented as a diagnostic instrument.
- Discontinued with no successor listing. A recommendation you cannot act on is not a recommendation.
5. What we did not do
Stated plainly, without apology, because it is the most important part of this page.
- We have not worn any of these devices. Not one. Devices we claim to have tested: 0.
- We do not run a test lab. There is no treadmill, no ECG, no polysomnography, no gear cupboard.
- We accept no free units and no paid placements. No manufacturer sees a page before it publishes.
- We hold no credentials. Nobody here is a coach, a physiologist or a clinician, and you will not find those words attached to a byline on this site.
- We did not read everything. Two directly relevant papers were behind publisher blocks on 24 July 2026. They are listed in our sources as unread, and no figure from either appears anywhere on this site.
The category standard is to claim a lab and twenty tested units. We would rather have a method you can check than a claim you cannot.
6. How to check us
Concretely, this is what "reproducible" means here:
- Open any ranked pageand expand "show how this was calculated" on a pick. You get the five sub-scores, their weights, and the one-line basis for each.
- Follow the source link next to any accuracy claim. It goes to the paper, not to a press release. Where we say no study exists, that is the claim you should try hardest to falsify.
- Redo the arithmetic.Cost per GPS hour is the live price divided by the manufacturer's stated GPS hours, with the mode named. Three-year subscription cost is the listing price plus the remaining months. Both are shown, not asserted.
- Change the weights. If you think endurance should count for more than evidence, apply your own weighting to the same sub-scores. You will get a different order, and that is the point — ours is a defensible choice, not the only one.
Current state of the record on this site: 26 devices ranked, 8 with peer-reviewed evidence behind their headline measurement, 6 ranked while we state that no validation study exists, and 26 carrying a live price verified Jul 24, 2026.
7. When rankings change
Accuracy scoreboards and roundups are reviewed quarterly, and immediately when a flagship launches or a new validation study is published. Comparisons update on a firmware or model change. Prices update automatically and are never edited by hand.
Any change to the weights on this page is dated and logged. A methodology that silently changes is not a methodology, and a rubric that quietly re-weights itself to justify a new favorite is exactly the failure we are positioned against.
Found something wrong? Tell us— the contact form leads with "report an error" for that reason, and corrections are logged visibly on the page rather than edited in silently. See the editorial policy for how that works.
A note on what the tiers are called
For completeness, the five evidence labels used throughout the site are: "Peer-reviewed validation", "Independent measured test", "Manufacturer specification", "Manufacturer claim only", "No validation study exists". You will see one of them next to every accuracy claim we make.