The Boat-Speed Debrief: Turning Races into Lessons

The fastest way to improve is a disciplined debrief after every session — comparing what the boat actually did against its polar targets, quantifying every manoeuvre, agreeing what to change, and recording it. Talent being equal, the crew that debriefs well improves faster, because they compound their learning instead of repeating the same mistakes. This is the method, the maths behind each number, and the failure modes it catches.

Why it works: a logged session is an experiment

A modern racing keelboat is instrumented well enough that a training day is a controlled experiment you have already run. Boat speed (through the water), heel, heading, apparent and true wind, rudder angle and VMG are logged many times a second — typical systems sample at 4 Hz internally and GPS position at 10 Hz. The debrief is where you interrogate that record while the crew still holds the context in their heads, before "that felt slow" decays into a vague impression.

The single most valuable move is to stop arguing about feel and read the percent-of-target channels. Performance software such as Expedition exposes polar boat speed % — if it reads 96 per cent, the boat is making only 96 per cent of the speed the polar says it should for the wind it is seeing. That one number reframes the whole conversation: instead of "were we fast?", the question becomes "we lost four per cent upwind on port in the chop — was that mode, trim, sea state, or a bad polar?". Every honest debrief is the disciplined pursuit of that four per cent.

Trust the numbers before you trust the conclusions

A debrief built on bad data is worse than none, because it launches confident changes in the wrong direction. So the first item is always instrument sanity, and it rests on understanding that the boat computes almost nothing directly — it derives true wind from a triangle.

The masthead unit measures only apparent wind: the vector sum of the true wind and the boat's own motion through the water. True wind speed comes from the law of cosines applied to that triangle:

TWS = √(BSP² + AWS² − 2 · BSP · AWS · cos(AWA))

where BSP is boat speed through the water, AWS apparent wind speed, and AWA apparent wind angle. True wind angle then follows from the same triangle. Crucially the calculation uses speed through the water, not speed over ground, because true wind is defined relative to the water the boat sits in — feed it GPS SOG and any current corrupts the answer. This is why the boat-speed calibration underpins everything: an error in BSP propagates straight into TWS, TWD and every target the polar hands back.

The field test for whether you can trust today's true wind is the tack-to-tack check. Sail a steady beat on one tack, note the true wind direction, tack, and note it again. In steady breeze TWD should read the same on both tacks. If it does not, the fix is a rule of thumb worth knowing: if you are lifted from tack to tack, subtract half the difference from the offset; if headed, add half. Persistent wobble of eight to ten degrees is usually leeway — the boat crabs a few degrees to windward of its heading, the masthead does not know it, and the true-wind solution swings. Sailmon, B&G and similar systems apply a leeway model (leeway ≈ k · heel / BSP²) and a true-wind correction matrix indexed by wind speed and point of sail precisely to kill this. If the tack-to-tack numbers are rubbish, park the speed discussion — you cannot judge height or TWA against a wandering reference.

The second sanity check is the polar itself. A polar is only a model, and the debrief must know where the model is weakest: it is typically built for flat water and a settled breeze, so in a two-to-three-foot chop real upwind VMG commonly falls 10 to 20 per cent below the flat-water target as waves stop the bow, force a wider tacking angle, and impose speed variation through the wave cycle. That is not slow sailing — it is the polar being optimistic for the sea state. Knowing this stops the crew from "fixing" a mode that was already correct.

Straight-line speed and height: the core comparison

With trustworthy data, the heart of the debrief is comparing what the boat did against three target channels, and it matters which one you read for which leg:

• Reaching / straight-line legs: read boat speed vs polar boat speed. Here you are trying to go fast on a fixed angle, so raw speed against the polar is the honest measure.
• Upwind and downwind (VMG legs): read boat speed vs target boat speed and TWA vs target TWA together. On these legs the goal is not maximum speed — it is maximum velocity made good toward the mark, and the two targets trade off against each other.

VMG is the projection of boat speed onto the wind axis: VMG = BSP · cos(TWA). The consequence is the central upwind truth — the fastest boat speed is never the best VMG. Bear away and BSP rises but cos(TWA) shrinks; pinch and cos(TWA) improves but BSP collapses. The optimum sits at the TWA where the product peaks, found geometrically by the horizontal tangent to the top of the polar. For a typical keelboat that lands somewhere around 40–50° TWA, but the exact target moves with wind speed and sea state, which is exactly why the boat carries a target TWA channel rather than a fixed number taped to the bulkhead.

This is where you separate a speed problem from a tactical one — a distinction most crews blur. Two failure signatures:

• On target TWA but under target boat speed: a genuine speed deficit. The angle is right but the boat is not making the speed the polar expects — look to trim, mode, foil state, or crew placement. This is a fast/slow problem to solve on the water.
• On target boat speed but wrong TWA: a mode or steering problem. The boat is quick but pointed wrong — sailing too low (fast but not gaining) or pinched (high but stalling). VMG is being thrown away at the helm and in the mainsheet, not in raw pace.

Read the two channels in isolation and you will misdiagnose constantly; read them as a pair and the loss localises immediately.

To make the comparison honest, average over 30 to 60 second steady-state windows rather than reading instantaneous values. Wave orbital motion, puffs and masthead pitching inject noise that a snapshot will misread; a rolling average over that window filters the noise while staying responsive to a real change in mode or breeze. Compare like with like — same tack, same wind band, same sea state — and never average a number straight through a manoeuvre.

Quantify the manoeuvres: where a race is quietly lost

Straight-line speed gets the attention, but on a windward-leeward course the manoeuvres are where boat lengths leak away unseen, and they are eminently measurable. On a keelboat a tack typically costs one to two boat lengths, and — the key insight — almost none of that loss is in the turn itself. It accrues in the acceleration phase afterwards. Exit straight to a hard, fully-trimmed close-hauled course and the boat bogs into a long, slow rebuild, costing two lengths or more. Exit low with the sails eased, build speed, then squeeze up as the numbers come back, and the cost drops toward one length. The debrief's job is to read which one happened.

The technique is manoeuvre overlay. Align every tack in the log on the instant of head-to-wind and every gybe on stern-to-wind, then plot boat speed, heel, TWA and rudder angle across all of them on one set of axes. This exposes, per manoeuvre and comparably across the session:

• Minimum speed in the manoeuvre and how deep the boat dropped.
• Recovery time — seconds from the low back to target boat speed. A crew that recovers in eight seconds is beating one that takes fourteen every single tack, and over a beat of ten tacks that difference is a leg.
• Rudder trace — a large, sharp, sustained deflection means the rudder was used as a brake through the turn; a smooth, brief trace means the turn was carried on heel and trim with the foil doing steering, not stopping.

Modern tools populate dedicated tack loss / gybe loss channels in time and distance directly, and fleet-aware analysis can express each manoeuvre as time lost versus the fleet median. Either way, once a tack has a number — "1.6 lengths, twelve seconds to rebuild" — it becomes a training target with a pass/fail test, not a matter of opinion. The same overlay makes gennaker sets and drops honest: a set that costs three lengths of VMG because the kite filled late is invisible to feel and obvious on the trace.

The running order, and keeping it constructive

Run the debrief in an order that lets each item build on the last, and cap the output hard:

1. Instrument sanity — TWD agrees tack-to-tack; leeway sane; is the polar trustworthy for today's sea state?
2. Straight-line speed and height — percent-of-polar reaching; boat speed and TWA against target on the VMG legs; speed problem or tactical problem?
3. Manoeuvres — tacks, gybes, sets and drops: minimum speed, recovery time, rudder trace, cost in lengths.
4. Starts — line bias, time-on-distance error, first-cross speed relative to the boats around you.
5. Modes and trim — was it the right mode for the band, and did any speed killers show up in the trace?
6. Gear and reliability — anything to fix, watch, or re-torque.
7. Tactics — the calls that worked and the ones that did not.
8. Actions — three or four specific, testable changes. No more.

Keep it constructive by discipline, not by softness. Debrief the boat and the process — "the exit was slow", not "you steered it badly" — because a number is impersonal and an accusation is not. Cap the change list at a handful because a crew can only genuinely work on a few things at a time; a list of fifteen faults gets none of them fixed. Record every lesson and every setting — rig tension, car positions, sheet loads, target deltas — so the change is acted on and so the next session can confirm whether it worked. And deliberately bank what worked: a fast mode found by accident is as valuable as a slow one you fixed, and if it is not written down it is lost by the next regatta.

A note on this boat's numbers

The published class figures for the Melges 40 — roughly 11.99 m LOA, about 3,250 kg displacement, an electrically actuated canting keel (cant up to 45°, bulb around 1,100 kg), and a sail plan near 72 m² main / 49 m² jib / 200 m² gennaker — set the scale of the loads and the polar, but they are starting points, not debrief inputs. The targets you actually debrief against are the boat's own measured polar and the calibration on the day, and every one of those numbers should be verified against the class rules and this boat's own documentation before it drives a decision — never against a figure remembered from a spec sheet.

The takeaway

The debrief is where races become lessons: trust the instruments first, read percent-of-target rather than feel, split every deficit into a speed problem or a tactical one, put a number in seconds and boat lengths on every manoeuvre, and leave with three or four testable changes that are written down. It is the habit that turns the boat's potential into steady, compounding improvement. Pair it with a working knowledge of the polars and a trained eye for the common speed killers.

This is a general debrief method; adapt the targets, channels and windows to the crew and to the polar and calibration your boat actually uses.