How accurate is your phone?
There are levels to it, and we would rather tell you exactly where your phone sits than promise centimetres to everyone. Below is what we actually measured across 25 real phones on the water — and an app that tells you what yours can do in two minutes.
What decides accuracy: the chips, not the brand.
Two hardware questions decide everything: does the phone expose raw GNSS measurements, and does it receive the second L5 frequency band with usable carrier phase? Neither shows up on the box.
Tier 1 — dual-band (L5) with clean carrier phase
The full pipeline: we post-process the raw satellite measurements against the national reference network (in Lithuania, LitPOS) and the track lands around ±30 cm horizontal — tight enough to score a tack in metres and to trust a mark rounding. About a quarter of the phones we tested sit here. Height is looser — two to three times the horizontal error, because every satellite sits above the horizon and nothing pins the up-axis; that's geometry, true of every GNSS receiver on earth.
needs: raw GNSS ✓ · L5 ✓ · carrier phase ✓ — e.g. Galaxy Note 10+, Pixel 7a in our bench
The honest catch: a carrier-phase solver can lock the wrong answer and still report high confidence — a "fixed" position that's metres off while looking perfect. We caught exactly this in a grid test where identical phones, sitting still centimetres apart, disagreed by several metres in height. So a fix has to earn the ±30 cm label: we cross-check every solution against a second reference station and a height-stability gate, and a confident-but-wrong fix is rejected, not reported. The ±30 cm is what survives that check — not the raw solver output.
Tier 2 — raw measurements, single band
No L5, but the raw data is there, so post-processing still pulls the track well under boat-length accuracy. The debrief is fully usable: tack shapes, lines, distances — you just stop arguing about centimetres.
needs: raw GNSS ✓ · L5 ✗ — a phone without L5 is still ok
Tier 3 — raw output present, but not usable for correction
Some phones report raw data with no usable carrier phase (in our bench the Galaxy S20 FE and S21 FE emit exactly 0% of it). Post-processing can't help those — you get good smoothed GPS. Still fine for the live map, speeds and coaching lines; not a centimetre story. And phones surprise us: one frozen-clock S20 still recalced to ±0.33 m.
raw GNSS ✓ · carrier phase ✗ — fine to sail with, we just won't over-claim
Tier 4 — no raw measurements
Older or heavily locked-down phones give the plain position fix only. The live map, replay and basic debrief all work — accuracy is ordinary phone GPS.
raw GNSS ✗ — everything runs, with honest error bars
What our phones are actually doing right now.
Measured from every fix the fleet recorded in the last 30 days, against the 30 days before it, so a model that has drifted shows it. This is raw reported accuracy in ordinary use, not the post-processed result in the study table below — expect metres here and centimetres there. Full per-unit history: phone models → · phone history →
Loading the latest measurements…
A model appears once it has at least 200 fixes in the window — a phone seen for ninety seconds should not sit beside one with a hundred thousand fixes as though the evidence were comparable. “Improving” / “worse” means median accuracy moved by more than 15%, wide enough to ignore the difference between a clear day and a cloudy one.
Phones we've measured, ranked for precision.
From our on-water study, ordered by how well each phone supports carrier-phase correction — the single thing that decides cm-class accuracy. "PPK track" is our post-processed result; "Live RTK" is whether it could also do real-time decimetre positioning. Interactive, sortable & filterable version with the latest field data: phone-models →
| Phone | Chip · band | Carrier phase | PPK track | Live RTK | Best role |
|---|---|---|---|---|---|
| Pixel 7 top pick | Tensor G2 · L5 | 88% | ±0.16 m | ✓ best candidate | Precision · start mark |
| Galaxy S20 / S10 / Note 10 | Exynos 990/9825 · L5 | clean, 59%+ | ±0.11–0.20 m | ✓ | Precision · marks |
| Pixel 10 Pro | Tensor G5 · L5 | 75%, gappy | needs tuning | ✓ with duty-cycle handling | Precision, once tuned |
| Galaxy A33 | Exynos 1280 · L1 | 69% | ±0.63 m | float only | Sub-metre · live dot |
| Galaxy S22 | Snapdragon · L1 | mid | ±0.41 m | float only | Sub-metre |
| Galaxy S20 FE / S21 FE | Exynos 980 / Snap 888 | 0% | not solvable | ✗ | Live dot only — not on marks |
| Huawei P30 Pro | Kirin 980 · L5* | 0% (code-only) | not solvable | ✗ | The "has L5" trap — avoid for precision |
*The P30 Pro reports L5, but only on code, not carrier phase — a naïve "does it have L5?" check passes it, then it can never solve. The real gate is carrier-phase satellites per epoch (P30 Pro = 0, S20 = 8, Pixel 7 = 22), not the L5 badge. Ironically the best live-Fix phones (S20 FE, S21 FE) emit zero carrier phase, so they're the worst for correction — one reason "it feels accurate" isn't the same as "it's correctable."
Don't guess — measure. In two minutes.
The tack.coach app reads your phone's GNSS hardware directly and shows exactly what it emits — satellites, signal, whether L5 is live, and (the one that matters) whether carrier phase is usable. You'll know your tier before you're off the dock, and every session you record adds your phone to our per-model database.
- Install the app (Android) and step outside with a clear sky.
- Open the GNSS check — it reads your chip live: L5 yes/no, carrier-phase satellites, raw output.
- Record a short session (or add your model below) and we profile it — then tell you where it lands.
Not in the table? Add your phone.
Tell us the model and we'll profile it from the next session it appears in — and email you where it lands.
Which phones to anchor on a mark.
A buoy phone is course infrastructure: it holds the mark's position all session and (with a wind sensor) reports the breeze. It deserves your best GNSS, not your oldest handset.
| Role | Recommendation | Why |
|---|---|---|
| Start / finish mark | Tier 1 phone (L5 + carrier phase) | Line calls and rounding scores inherit the mark's accuracy — put the best phone here. |
| Course marks | Tier 1–2 | Anchored phones average their position over time, which recovers a lot — mid phones do fine. |
| Wind buoy | Any tier + wind sensor | Wind readings don't need centimetre position — a stable anchor and a clear mount matter more. |
| Avoid on marks | Phones with zero usable carrier phase | Fine on a boat, but as a reference point they add avoidable wobble (our bench: S20 FE / S21 FE). |
From our 25-phone study: 7 of 25 received L5; carrier-phase quality — not price or age — was the strongest predictor of final accuracy.
When you need more than a phone.
Full disclosure: dedicated instruments exist for a reason. Systems like Vakaros deliver 1 cm live positioning for umpired starts and world-championship race management — at over €1,000 per boat plus a corrections service. If you need that, buy it; it's good equipment. tack.coach is the serious option for the other 95% of time on the water: coaching-grade evidence — ±30 cm horizontal reprocessed tracks (the fixes we trust, not every fix the solver offers), scored tacks and roundings, measured wind — from the phones your sailors already carry.