Why this exists
A masthead vane mounted a few degrees off skews every number your team debriefs with: pointing angles, target comparisons, wind direction, the shift ladder. The error is small enough that nobody sees it by eye — headers, lifts and gusts are all far larger — and constant enough that it quietly biases a whole season the same way.
Guided calibration finds that error in the sailing you have already recorded, and proposes the correction. You review it, and you decide whether to apply it. The platform never changes your data on its own.
The principle
A vane mounted 3° off reads 3° wrong on every beat, in every breeze — but with opposite effect on the two tacks. Wind shifts and gusts are bigger than that, which is why the offset hides from the naked eye; but they are not attached to a tack, so over enough tacks they average out. What refuses to average out is the constant.
Note the scope of that claim, because the rest of this page depends on it: it holds within one window of sailing. What the constant turns out to be made of, and what changes about it between one sailing day and the next, is section 3 — and it is not only the vane.
1 Finding the usable sailing
Before anything is measured, the analysis cuts your day into steady close-hauled segments — the stretches between maneuvers where the boat is simply sailing. Tacks and gybes are detected from wind angle, heading and boat speed, with deliberately conservative rules:
- Close-hauled and running are recognized inside guarded wind-angle bands, with dead zones around head-to-wind and dead downwind — so wobbling through an axis can never be mistaken for a maneuver.
- A state must hold for ten seconds to count, and a boat doing under two knots is drifting, not producing evidence.
- Every segment is trimmed at both ends: the speed sag and rebuild around a maneuver are not steady sailing, and they don’t get a vote.
2 The tack-symmetry measurement
Write the vane’s error as a constant b. If your boat’s true close-hauled angle is A, the instruments read A + b on one tack and A − b on the other. Take the median wind angle of each steady segment, then the median of those per tack — medians, so one long beat can’t outvote five short ones, and a single odd stretch can’t drag the answer. Then:
Read as a sentence: a boat that appears to point higher on port tack has a positive offset, and needs a negative correction.
The suggestion is made on both true wind angle and true wind direction with the same number — a constant in the boat’s frame is a constant in the compass frame — and deliberately not on apparent wind, where the wind triangle would turn the same number into a different, wrong correction.
3 What the constant contains
Everything above has called that number a mounting error. It is not only that, and the full version is worth the paragraph.
The wind does not blow the same way at deck level and at the top of your mast. Friction with the water drags on the lowest air and turns it, so the breeze usually sits a little further round at the masthead than in the air your sails are actually working in. That is wind shear — and your vane lives at the top of the mast, in the turned air, while the boat is driven by the mean of what is below it.
So the wind picture your instruments hand you is rotated relative to the wind the boat is sailing in. A rotation is precisely what a crooked vane is. It widens the measured wind angle on one tack and narrows it on the other, by the same amount, with the same signature — and the tack-symmetry measurement cannot tell you which of the two it just measured, because in one window of sailing they are the same thing.
How much shear? Crews who calibrate this way as a matter of routine report a degree or three as ordinary, and considerably more when a cold sea sits under warmer air — a spring morning, early season — where past ten degrees is described rather than theorised. Those figures are practitioner experience, offered here for a sense of scale. Nordic Stars does not measure wind shear, and none of those numbers are ours. What we can tell you is the part that matters for the workflow: shear holds roughly steady through a day, and it moves with the next weather.
Which gives the rule the whole feature hangs on: analyse each sailing day on its own, and give each entry that day’s effective-from.
Do that and the calibration timeline stops being a list of corrections and becomes something more useful. The part that repeats, day after day, is your mounting offset — it does not move until somebody touches the rig. The movement around it is the air: how much the breeze was turning above your deck on each of those days. The timeline is your shear log. Nobody has to build that record; keeping the habit is the record.
Two schools
Some race crews deliberately leave shear uncorrected, and their argument is a good one: if the vane is accurate and the shear is left in, then the asymmetry between the tacks is live information — it tells the afterguard, during the race, that the air aloft is favouring one board.
Our recommendation is to correct it. Laylines, target tables and every polar comparison this platform makes assume a symmetric wind; hand them an asymmetric one and they are quietly wrong on one tack all day, which costs more than the trim signal is worth. And correcting does not throw the signal away — it moves it into the timeline, where the day-by-day constants say the same thing more precisely, and with a date on it.
If your team sails the other school, that is a legitimate choice: leave the wind entries off and read the asymmetry on the water. What does not work is half of each — a partly corrected wind is neither an accurate instrument nor a usable signal.
4 The cross-check
The wind belongs to the sea, not the boat. When you tack, the true wind direction does not change — the boat turned, the wind did not. So any jump your instruments show in wind direction at the moment of a tack was manufactured by the boat, not the wind.
The two kinds of error behave differently at that moment. A constant angle error — the vane, the shear, or the two of them together — shifts your entire wind picture by the same few degrees on both tacks, all day: wrong, but consistently wrong, so nothing jumps and wind direction sails smoothly through the tack. The other kind of error — leeway, current, a vane that leans with heel — pushes the reading one way on starboard and the other way on port, and that flip shows up as a wind-direction jump at every tack.
You may recognize the manual version of this: the old instrument check where you tack the boat and watch whether the wind direction holds. This is that check, done statistically over every tack of the day — measured in short windows either side of each tack so slow wind drift cancels out. Only the second kind of error can make it jump, which is what makes it the perfect cross-check for the first:
| Cross-check step | What it means |
|---|---|
| holds steady | Confirms the finding really is a constant — the error does not flip with tack. Apply with confidence. |
| jumps | Warns that part of the error changes sign with tack — leeway, current, or a heel-dependent sensor — and no single correction number can fix a thing that changes sign. The platform tells you so instead of suggesting one. |
5 The one thing that can fake a constant
Weather cannot fake a constant. A crew can.
Sail one board a shade lower and faster than the other — a trim difference, a helm preference, a habit nobody on board has noticed — and that board carries a wider wind angle on every beat of the day. It is as persistent as a vane three degrees out, it survives exactly the averaging this method relies on, and in the angle alone the two are indistinguishable. The measurement would read the sailing and call it the instrument.
Boat speed is what separates them, and it is already in hand: the fast board is the board sailed low. So every tack-symmetry finding is served with what each board actually held — accepted time, median wind angle and median boat speed, over exactly the segments the offset was taken from — and the difference between the two speed medians. Past a fifth of a knot, which is where practitioner calibration methodology stops reading a tack’s speed difference as the instrument and starts reading it as crew work, the finding is shown with a caution beside it.
It is a caution, not a gate. The suggestion stays, at the confidence it earned, and nothing about applying it changes: a boat can be both out of calibration and sailed unevenly, only the crew knows whether they held target speed on both boards, and hiding a number because it might be contaminated teaches nobody anything. The platform prints the difference and leaves the judgment where it belongs. A boat with no speed log gets a plain line saying so — an absent measurement is not a passed check.
One consequence worth keeping: the cross-check above is largely immune to this. Drive a board low and the heading falls by the same amount the wind angle widens, so TWD ≈ HDG + TWA does not move at all — which is why it is reported beside the suggestion rather than folded into it.
6 The speed ratio
Alongside the wind work, the analysis reports the median ratio of GPS speed to boat speed. A paddlewheel reading 2% slow shows up as a ratio of 1.02 — already the correction you would enter. This one is always reported as informational, never a recommendation: current flows straight into it, and no amount of data separates a slow paddlewheel from a knot of tide.
The safeguards
A number without its conditions is noise wearing advice’s clothes. Every suggestion has to clear all of these, and when one fails, the analysis says which — it never quietly drops a result:
- Enough tacks. At least four tacks and three steady segments per tack. The wind averages out over tacks; your instruments don’t. Too few, and both answers are just the weather.
- Tight enough scatter. If the spread across segments exceeds the size of the effect being measured, the day can’t answer, and the result is shown as measured-but-not-recommended.
- A noise floor. Below half a degree, nothing is suggested. A well-calibrated boat gets the answer it earned: the wind picture looks square.
What applying does — and doesn’t
Applying a suggestion writes an ordinary calibration entry: a correction the platform applies when your data is read, from the moment you choose. Your recorded telemetry is never rewritten. Delete the entry and every chart returns to exactly what it showed before — calibration in Nordic Stars is a lens over your data, not an edit of it.
Because analysis reads through that same lens, the loop converges: apply the correction, run the analysis again, and it now measures what’s left. In validation, an injected 3.0° vane error is recovered as −3.04°; after applying, the re-run measures a residual of −0.02° — and suggests nothing.
Those are recorded measurements, not promises. The test they come from passes anything within half a degree of the injected error; the numbers above are what it actually measured on the day it was written.
Where to find it
In the web app: Boats → your boat → Calibration → Analyze. Pick a window of sailing, run the analysis, and review the cards — each shows its correction, its evidence (segments, tacks, spread, and what each board held), and its cross-check verdict. Apply writes the entries; the timeline below shows every correction in force, and any of them can be deleted at any time.
Run it a day at a time. One sailing day is the window this is built for, and giving each entry that day’s effective-from is what turns the timeline into the record described in section 3.
Guided calibration is a measurement offered as evidence, not a verdict. Your rig, your sailmaker, and your own judgment stay in charge — the platform’s job is to make the constant visible.