How to build a polar for the boat you actually sail

Many offshore crews are routing on numbers computed for a fictitious yacht. Here is where those numbers come from, why they go wrong, and how to replace them with your own.

I acquired a 1970s Ron Holland half tonner, called Silver Shamrock, with a Half Ton Cup win to her name and a set of assumptions built into her that no longer applied. She had been designed and campaigned under IOR, and IOR shaped everything about her: the sail plan with overlapping genoas, the way she was designed for the rule then. I then rebuilt her for doubled-handed racing, went to non-overlapping headsails and added a code zero and asymmetric spinnakers.

Whatever performance model had ever existed for that boat died the day I did that. The hull was the same. Nothing else was. A polar built around a big overlapping genoa tells you nothing useful about a boat now carrying a blade and a code zero because the crossovers have moved, the reaching angles have moved, and the entire shape of the curve between sixty and one hundred and sixty degrees has changed. I was not looking at an inaccurate polar. I was looking at a polar for a different boat.

So I had to build a new one. I took generic upwind and downwind speeds for a boat of roughly that size and type, put them into a spreadsheet by hand, and then went sailing to find out how wrong they were. They were wrong in the way you would expect, optimistic in the light and pessimistic reaching, and I corrected them. Then I let the software do the rest. SeaTrack, a rudimentary routing software, which by modern standards was primitive but whose isochrone routing algorithms are still essentially what everything uses today, had a routine that would sample boat speed at different true wind angles and update the polar when the boat sustained something better than the number already held. It offered several ways of doing that depending on how long a speed had been maintained at a given angle, which is a more intelligent approach than simply grabbing the maximum, because a surf down one wave is not a polar entry.

Over a season or two it produced a decent set of numbers. Not a beautiful curve. A set of numbers I trusted, which is a different and better thing. We used it to work out the offset for a Channel crossing and to estimate when we would arrive at a tidal gate, which in 2016 was about the limit of what the routing was really good for. It was enough to win the Myth of Malham in class that year.

What I did not appreciate at the time is that I was doing, by hand and rather slowly, precisely what the industry now sells in a box.

Silver Shamrock after conversion. The hull was unchanged. Everything that determined her performance was not, and no polar existed that described her

Many yachts racing offshore are routing on a performance model that describes a boat, but not their actual boat and the tools they are paying for are not going to fix that for them. Building it yourself is not a refinement for the well organised. It is the only way it happens. On the top offshore boats this is the main task of the tactician/coach and the focus for a lot of the training - model the boat accurately and keep improving.

Everything downstream of your polar carries its errors: every routing run, every layline, every target speed on the mast display, every decision about whether the extra miles are worth it. The forecast gets all the attention and all the money. The input that converts the forecast into a decision can be overlooked.

What a VPP actually is, and why yours is mostly not about your boat

A Velocity Prediction Program is not a record of anything. It is physics plus maths, and knowing its two halves tells you exactly where it can be wrong.

The VPP solution algorithm is looking for balance. A sailing boat is in balance when two things are true at once. The sails are pushing the boat forward exactly as hard as the water and the air are holding it back, so the speed is steady. And the wind is heeling the boat over exactly as hard as the hull and its ballast are standing it back up, so the angle of heel is steady. Get both right at the same moment and the boat has settled into a state it can hold. That is what the program is hunting for.

It does this by guessing. It picks a speed, an angle of heel and a state of trim, works out whether the forces actually balance, discovers they do not, adjusts and tries again, and keeps going until it finds the fastest speed at which both conditions are satisfied at that wind angle. Then it repeats the exercise for the next angle, and the next wind strength, until the table is full. So every number on your polar is the answer to a calculation about balance.

Note: Your designer almost certainly did not write that solver. It is generic mathematics and it knows nothing about your boat, so the industry buys it off the shelf. The standard commercial package is WinDesign, developed by the Wolfson Unit at the University of Southampton and now in its fourth version, which costs a little under three thousand pounds and has been in use by designers, model testers, handicappers and rule-makers for over twenty years. Larger offices run proprietary solvers in-house, and the rating authorities keep their own. What the naval architect actually supplies is the second half, the boat model, which is where all the real judgement and all the real proprietary knowledge sits. The first VPP came out of MIT in the early 1970s, funded by Commodore H. Irving Pratt to predict the performance of a yacht from knowledge of its hull, rig and sailplan geometry, and everything since is a refinement of that idea.

The boat model supplies the physics, and this is where it stops being about your boat.

Hydrodynamic resistance splits into viscous drag, essentially skin friction, which scales with wetted surface, and residuary resistance, which covers everything else on an upright bare hull in calm water. Predicting residuary resistance for an arbitrary hull from first principles is hard and expensive, so the industry leans on systematic series. The Delft Systematic Yacht Hull Series began at TU Delft in 1971 and now runs to around seventy systematically derived models, all towed under a consistent procedure, from which polynomial regression equations were derived. A companion Delft Systematic Keel Series does the same for appendages. When a designer runs an early-stage VPP, those regressions are usually doing the hydrodynamic work.

The regression describes a family of hulls. Your hull is being estimated by its resemblance to that family. A boat sitting near the centre of the sampled parameter space gets a reasonable estimate. A boat at the edges is being extrapolated. A boat outside them is being guessed at politely.

Every number on your polar is the answer to a physics question, not a record of anything your boat has done.

Modern practice adds towing tank work on the actual design, wind tunnel testing for sail force coefficients, and increasingly RANSE CFD that can model hull, sails and rudder together rather than decoupling the two problems. That is better physics and considerably more expensive, which is exactly why it is standard on a Grand Prix programme and absent from the production cruiser-racer most of us are actually racing.

How ORC does it

ORC maintains its own VPP as the engine of its rating system and publishes full documentation of the method, which is worth downloading if you want the equations rather than my summary. Its solution algorithm balances sail driving force against hull and aerodynamic drag, and heeling moment against righting moment, with righting moment coming from an actual stability measurement, reduced for stability loss due to forward speed and increased for crew shifting. The aerodynamic model accounts for hull, mast, rigging and crew windage as well as the sails.

The output grid is worth committing to memory, because it is the grid every ORC-derived polar is defined on: true wind speeds of 6, 8, 10, 12, 14, 16 and 20 knots, and true wind angles from the optimum beat through 52, 60, 75, 90, 110, 120, 135,150,165 degrees to the optimum run. Everything between those points is interpolation, including quite a lot of the reaching range where offshore races are actually decided.

ORC’s International Technical Committee, composed of designers and aero and hydrodynamics specialists, revises the VPP annually using tank testing, wind tunnel studies, CFD and analysis of the season’s race results which has a consequence most owners never think about every time you get a new certificate.

How it all plays out in reality

I wrote recently about whether there is a battle between IRC and ORC. That piece was about the health of the sport. This one is about what happens on your chart table on a Tuesday, and there is a practical consequence of the two rules that I did not draw out at the time.

An ORC certificate carries a complete set of predicted boat speeds across that grid, with beat and run VMGs and their angles, explicitly intended for comparison against real performance on board. You can buy more of it: a Speed Guide gives the full set of polar diagrams in graphic, tabular and digital formats, including the format Expedition reads, and you can buy one whether or not you race ORC. An IRC certificate carries a Time Correction Coefficient and your measurements. No VPP output, no target speeds, no polar, nothing you can load into a routing package or an instrument system.

That is not a criticism of IRC, whose reasons for keeping its mathematics closed are sound and which I have argued elsewhere. The ORC racer is handed a performance model with their certificate, flawed and rule-shaped but real. The IRC racer is handed a number that determines their result and nothing whatsoever to navigate with.

There is one detail from the rating world worth carrying into this article, because it is about polars rather than politics. ORC does not measure the wind in order to score a race. It infers it from the boats elapsed time divided by course length gives an average in seconds per mile, that value is found on the boat’s own performance curve, and the wind speed at the intersection becomes the Implied Wind, with the highest implied wind winning. ORC, in other words, runs the VPP backwards. Under that system your polar is not merely a tactical tool you happen to own; it is the instrument by which your result is determined. Under IRC it is nothing of the sort, because your result comes from a formula your polar has no part in.

Which produces a conclusion most IRC owners have never had put to them. Under ORC an inaccurate polar costs you results directly. Under IRC an inaccurate polar costs you nothing on the score sheet and everything on the race track, because it is still the input to every routing decision you make, and nobody is going to hand you a better one.

What are you actually doing about it?

In rough order of rigour, IRC crews are doing one of five things.

Buying an ORC-derived polar without racing ORC. The route most IRC owners do not know exists. You can obtain a Speed Guide or Target Speeds from ORC Sailor Services for a modest fee whether or not you ever enter an ORC event, and it arrives in formats routing software reads. For a boat that has never touched ORC, that is a legitimate, cheap, physics-derived starting point, and it is a considerably better base than a library polar from a different boat.

Using the designer’s numbers. These are always optimistic, and the mechanism matters more than the fact. One sailor comparing logged data against his designer’s polars found the designer’s had been run assuming a lightship vessel and flat sea. Not merely optimistic, but modelling a boat carrying nothing, in conditions that do not exist offshore. That is a shape error as well as a level error, and no percentage adjustment will fix it.

Pulling from community databases. The collections at Blur, sailingpolars and the scraped ORC certificate archives are genuinely useful and honestly labelled. Blur’s own page says the polars there are published to play around with and that for serious use you should get a real one for your specific boat. The people curating these archives are more candid about their limits than most of the people downloading from them.

Borrowing a similar boat’s polar and applying a scalar, which is what the routing packages recommend when your boat is not in their library. More on why that fails below.

Building from logged data, which until very recently meant real effort. The clearest documented case is a sailor mid-passage, frustrated that he had no accurate polars and was reduced to guessing and applying scaling factors to somebody else’s numbers, who responded by writing Python scripts to log NMEA0183 from the boat’s B&G system, log the data by TWA and TWS, average it, and output CSV files per sail configuration. That is a competent engineer solving from scratch the problem.

The scale of the disagreement is real. ORC Club certificate, allegedly, may predict speeds more than fifteen per cent faster than the designer’s polars. Experienced navigators responding said this was common, because different VPPs make different assumptions about how the boat is sailed. Fifteen per cent is not a rounding error. Route six hundred miles on a polar that fast and you will arrive at every tidal gate with a plan built for a boat you do not own.

I will put my own experience rather than a statistic. In the boats I coach and sail on, the common position is a designer or library polar carrying an ad hoc correction somebody applied at some point, and very few crews are doing systematic refinement. I offer that as testimony, not as a survey. The gap is not knowledge. It is that the rule provides no starting point, the tools that make refinement easy are recent and cost money, and the tools that make it rigorous require you to write code, which interestingly is becoming a lot easier!

What the automatic tools are really doing

The new generation of polar tools invert the problem entirely. A VPP predicts what a boat should be capable of. These record what your boat did.

Take PredictWind’s implementation as the worked example rather than an endorsement, because it is the most visible. A DataHub plugs into the NMEA network and uploads performance data alongside GPS tracking. The generation process deliberately uses GPS speed and course over ground rather than speed through the water, precisely because that requires no calibration of a paddlewheel, and corrects for tide using their own coastal current model. It then combines that with significant wave height, angle and period from ECMWF, producing what the company calls a five-dimensional polar: wind speed, wind angle, and the sea state you were actually in.

Two details are genuinely clever. The system sorts incoming data into what its documentation describes as hundreds of separate groups representing different wave conditions, refining within each, and weights sailing data over motoring data where both exist for similar conditions. It also models night sailing separately, on the reasonable assumption that crews sail more conservatively in the dark. No designer polar knows what time it is.

The honest limitation was stated plainly by PredictWind’s Nick Olson, asked by Sailing World in 2024 whether the system would notice a headsail blown out mid-race. He said it would “adjust over time, but not in the short term.” That is the fundamental trade of the observed-data approach. It converges on the truth about your boat slowly, and it cannot distinguish a boat that has changed from a crew having a bad night.

Other implementations use the same idea with different statistics. Analytics platforms typically sort logged points into cells by wind speed and angle, then take a high percentile within each cell, commonly the ninetieth, rather than the mean. The logic is that the mean describes an average moment, including the ones where you were reefing, eating or asleep, whereas a high percentile approximates what the boat achieved when it was being sailed properly. Some seed from a VPP polar and progressively replace it with observed data wherever the real numbers exceed the prediction, which is exactly what SeaTrack was doing for me a decade ago with far less compute and considerably more patience required.

The distinction to hold on to: a VPP polar is theoretical; a P90 logged polar is evidential. The first describes what the model says the boat could achieve. The second records the upper edge of what your boat has actually demonstrated in the conditions you have sampled. It excludes the worst watch by design, so it is not an operational forecast. For routing, apply separate allowances for crew performance, darkness and sea state. Keep both models, and know which question each one answers.

Why a PredictWind polar can be rubbish, and how to tell

The AI polar is the good end of PredictWind’s offering. The problem is that most users never get near it, because it requires a DataHub and a Professional subscription, and what they use instead deserves more scepticism than it generally receives.

There are four tiers, and they are wildly different in quality.

Predefined polars are a library of specific designs, and the library is sourced rather than computed. PredictWind’s own guidance says that if you can obtain polar data from the designer, manufacturer or an owners’ forum, they will upload it, provided it arrives as a text or .pol file rather than a PDF. So a predefined polar is not a PredictWind calculation at all. It is whatever number the designer’s marketing department was comfortable with, or whatever an enthusiast posted on a forum, passed along in good faith. PredictWind is a distributor of that number, not an auditor of it.

Sail polars are the tier that should worry you. For a boat not in the library, you enter your boat speed at 50, 90 and 160 degrees true wind angle in fifteen knots of breeze. Their documentation then states the assumption plainly: that this represents the vessel’s maximum speed, with boat speed decreasing as the wind drops.

Think about what has just happened. Three numbers, guessed by the owner, at one wind strength, have been extrapolated by assumption across the entire wind range and every angle in between. There is no physics in it. There is no hull, no sail area, no displacement, no stability. It is a shape fitted to somebody’s optimism. If the owner is generous with those three figures, and owners generally are, then every route that boat ever runs is generous by the same margin, and the software will present the result with a clean line on a chart and an ETA to the minute.

The polar speed adjustment is the fallback when your boat is not in the library, and it is recommended for exactly that case: pick a similar design and scale it. The limitation is arithmetic. A scalar moves the whole curve. It can correct a level error and it can never correct a shape error, which is precisely the error I introduced on Silver Shamrock when I went to non-overlapping headsails and added a sprit. My boat did not become uniformly slower or faster. She became differently shaped between sixty and one hundred and twenty degrees, and no percentage in the world fixes that.

It is a ladder of rigour, and nothing in the interface tells you which rung you are standing on.

None of this is dishonesty on PredictWind’s part. Every one of those limitations is documented on their own help pages, in plain language, by people who clearly understand the problem. It is a usability trap rather than a marketing one: the tiers look like a menu of options when they are actually a ladder of rigour, and nothing in the interface tells you that the tier you have chosen has reduced your boat to three numbers.

The test is simple enough. Ask where every number in your routing software came from. If the answer is a library entry for a boat that is not yours, or three figures somebody typed in once, you do not have a polar. You have a placeholder that the software is treating as fact.

Where the work actually happens, and how it gets back to the team on deck

It is worth being clear about the division of labour, because people assume their instruments are doing more than they are.

Your instrument system holds a polar and reports against it. It generates target boat speed and target wind angle from the table you loaded, which is exactly what you want on deck at four in the morning. It does not revise that table. The revision happens in your routing and analysis software: Expedition includes functions to output, analyse and create or modify a yacht’s performance polars, and Adrena in its Pro form advertises creation and adjustment of speed polars alongside management of multiple polars.

So the working loop for most well-run boats is this. Log at sea, analyse ashore in the routing package, revise the polar there, export it, and load the revised table into the instruments so the numbers on deck match the numbers the navigator is using.

Three practical points about that loop.

It is not automatic, and the software is not doing your thinking. Expedition gives you the workbench, not the judgement. You still choose which intervals of data are representative, and decide what was a genuine performance rather than one good wave. That is precisely why the software provides annotation for when you were sailing deliberately for data collection, so you can extract those samples later rather than averaging your whole season.

The formats do not match, so budget for a conversion every time. The instrument grid and the routing grid are different animals. An H5000 expects its own fixed set of angle increments and a defined number of wind speed rows, which is why an ORC polar will not drop straight in and why the standard route is to build the instrument table in Expedition. Do this once and it is fifteen minutes. Forget it exists and you will lose an evening.

Your router can hold several polars. Your instrument holds one. If you are running polars split by sail combination or by day and night, the number the driver sees is a flattened version of what the navigator is routing on, and the two drift apart the moment you revise one and forget the other. Put it on the pre-race checklist: which polar is in the box, and does it match the one downstairs.

Note:

There is a neater alternative worth knowing. Expedition can export computed values directly into the external channels of your instrument system for display on deck. That lets you keep the polar in the nav computer at full resolution, with all its sail-specific variants intact, and send the resulting target speed out to the mast display as a channel. One source of truth, no conversion. The trade is that your targets now depend on a laptop staying alive on a wet night and always using power.

And a coaching point that matters more than any of the mechanics. Think hard about which polar you put into the instruments. If you load your observed polar, the targets on deck now describe what you have historically achieved rather than what the boat is capable of. If those logs contain a season of mediocre trim, the display will cheerfully tell the driver he is on target while the boat is slow, and you have built a machine for confirming your own bad habits. My preference is the other way round: keep the aspirational number in front of the crew, and keep the honest, autobiographical one in the router where it belongs, doing the job of predicting where you will actually be in six hours.

How wrong the standard numbers can be

Wrong enough to lose races, and wrong in patterned rather than random ways, which is worse, because patterned error survives averaging and quietly biases every decision in the same direction.

Designer polars are computed for a stated condition, and nobody tells you what it was. That matters more than the familiar complaint that they run optimistic. A designer’s polar is a legitimate calculation for a specified displacement and sea state, and that specification is usually light ship and flat water. If the assumptions do not travel with the numbers, you cannot correct for them, and you end up treating a calculation for an empty boat in a millpond as a prediction for a loaded boat in the Channel. Ask what was assumed before you ask whether the numbers are right.

Rating polars carry the rule’s biases. An ORC certificate polar is a genuine VPP output, but the VPP exists to rate boats fairly against each other, not to tell you personally how fast you will go. The substantive criticism raised by owners of older designs is that the stability model favours boats with high initial stability and penalises narrower, shallower-keeled older boats. Independent analysis by owners scraping ORC certificate data across configurations has suggested the VPP overestimates stability and how high a boat can be sailed with a code sail, while underestimating speed differences in light air.

The model moves underneath you. Because the ITC revises the VPP annually, the same boat in the same configuration produces measurably different polars from one certificate year to the next. If you are comparing this season’s performance against a polar from three seasons ago, part of the difference you are chasing is the model moving rather than the boat.

Flat water is the default assumption. PredictWind’s Jon Bilger put it in a line to Blue Water Sailing in December 2025, that polars are basically created for flat water while boat speed changes dramatically in a seaway. Inshore in the Solent that is a modest error. On the North Sea leg of a long offshore race it is enormous, and it is the single largest source of divergence between what your polar promises and what you experience at three in the morning.

And then there is your boat specifically. The polar does not know about the tired leech on your old number two, your actual crew weight, your bottom condition, whether the prop folds, or whether the instruments have been calibrated this season. I have put realistic offshore performance at somewhere between ninety and ninety-eight per cent of theoretical targets in typical racing conditions. That range is not crew failure. Most of it is the gap between a modelled boat and a real one.

The consequence for routing is severe and asymmetric. A routing engine does not degrade gracefully when the polar is wrong. It makes confident decisions on false premises. If your polar overstates light air speed, the router will cheerfully send you into a soft patch it believes you can sail through. If it overstates reaching speed, it will choose an offshore route costing you miles you cannot recover. The output looks identical either way: a clean line on a chart with an ETA attached.

Building and correcting your own VPP data

Fix the inputs before you touch the polar

There is no point refining a performance model on top of bad data. If your true wind angle is wrong then your polars are worthless, because TWA is the reference system polar diagrams are written in.

1. Calibrate boat speed against a known distance in both directions on a still day, to cancel tide.

2. Apply heel and linearity correction if your system supports it. Boat speed error varies with heel, and offshore you are rarely upright.

3. Verify masthead unit alignment. A two degree offset produces a polar systematically wrong on one tack, and it will look exactly like a helming problem.

4. Confirm heading and leeway settings, because TWA is computed from them.

1. Get a starting polar, and know exactly what it is

You need a baseline, and the source matters less than your honesty about its provenance. In order of preference: an ORC Speed Guide or certificate polar for your actual boat, purchasable whether or not you race ORC; a designer polar with a discount applied and its assumptions interrogated; a class or production polar for the same model; or a similar-performance boat from a routing library. Record which you used and its date. A polar of unknown provenance cannot be corrected systematically, only fiddled with.

2. Log everything, then sort it

The method the analytics platforms use is entirely reproducible in a spreadsheet, which is essentially what I did on Silver Shamrock.

1. Log at a fixed interval, ideally one second, capturing time, boat speed, SOG, COG, TWS, TWA, heel and sail combination. Note sail changes by hand if nothing does it automatically.

2. Discard anything within a few minutes of a manoeuvre, and anything where the boat was not being raced properly.

3. Sort each remaining point into the group that matches its wind speed and angle. You are filling in the same grid your polar table already uses. Five degree steps of angle and one or two knot steps of wind speed are a sensible starting resolution.

4. Within each group, take a high percentile rather than the mean. The ninetieth is the common default.

5. Count the points in every group before you trust it. This is where amateur polars fail. Finer steps produce a smoother-looking curve and noisier numbers, and a group holding eleven points from one windy afternoon is not evidence of anything.

6. Smooth the result, and be suspicious of any point requiring a kink.

A stylised infographic of the process as an example only: Take a high percentile, not the mean, and never trust a very small group.

The honest outcome of a first season is a polar well supported in the conditions you sail most and thin everywhere else. That is fine, and far more useful than a smooth curve you cannot justify. Mark the sparse regions and keep your baseline numbers there.

3. Run more than one polar

Navigators who do this seriously maintain several sets rather than seeking one true polar: the designer’s, the rating polar, and one or more observed sets. The rating polar answers whether you are sailing to your certificate, which is a different and equally useful question from whether you are sailing fast. Where you have the data, split observed polars by sail combination and by day against night.

4. Get it into the systems, and mind the formats

This is where a good afternoon’s work commonly dies.

Formats are not interchangeable. The common table format has wind speeds across the top and angles down the side. Expedition uses a different structure, each row beginning with a wind speed followed by angle and speed pairs. A near-universal rule is that angles must increase from left to right along each row, and files are rejected outright if they do not, which is a maddening error to diagnose.

The specific trap worth naming: ORC certificate polars will not load directly into a B&G H5000, because the instrument expects its own fixed set of angle increments rather than the ORC angles. You will need to interpolate onto the instrument’s grid, which Expedition will do, or rebuild the table by hand.

5. Close the loop, every season

The bottom fouls, the sails age, the crew changes, and if you hold an ORC certificate the model itself moves annually. Polar work is a programme, not a project.

Testing the argument: is the work worth it?

“This only matters if you route offshore, and most IRC racing is inshore or short coastal.”

Partly fair. If you race round the cans in the Solent you can win a great deal on tide, starts and boat handling without ever opening a polar. But target boat speed and target wind angle are polar outputs, and if you sail upwind to a number on a display, you are already using a polar whether you built it or inherited it. The question is only whether you know where the number came from.

“The gains are marginal compared with sailing the boat better.” Very often true, and I have argued this myself at length: races are won by hundreds of slightly better decisions, most of which are about trim, crew and timing rather than software. But the polar is not competing with those things. It is the benchmark by which you find out whether you are doing them well. Without a trustworthy target speed you cannot tell a slow leg caused by bad trim from one caused by a wrong route, which means you cannot learn from it.

“IRC boats have raced perfectly well for thirty years without any of this.” They have, and boats of all ages win under IRC precisely because the rule refuses to be optimised. But the fleet has changed underneath that argument. When routing software was crude, a rough polar was adequate because you were mostly using it to estimate a tidal gate. Now that most of the fleet is running sophisticated optimisation on high-resolution forecasts, the polar has become the binding constraint on the whole system, and a rough one puts you in the position of executing confident decisions from bad premises.

“You are really just arguing for ORC.” No, and I want this on the record. IRC’s non-publication is a serious answer to a serious problem, and I lived through the alternative. A published rule gets optimised, optimisation shapes the boats, and the boats it shaped in the 1970s are part of why I ended up retiring to Falmouth in the 1979 RORC Fastnet race. The argument is narrower and I think harder to dismiss: the rule you race under has made a choice, that choice has a cost, the cost lands on you rather than on the rating office, and nobody is going to tell you about it.

“The AI tools solve this now, so the DIY method is obsolete.” They solve a good deal of it, and I would use one. But they need a base polar to start from, they converge slowly, they cannot distinguish a changed boat from a tired crew, and they describe how you sail rather than how the boat could be sailed. If your crew work is poor, an observed polar faithfully records poor crew work and then routes you as though that were a law of physics.

What this actually buys you

Not glamour. A corrected polar will never appear in a race report or win an argument in the bar. What it buys is that every downstream decision, every routing run, every layline, every call about whether to reef now or hold on, is made against a benchmark describing your boat rather than a family of hulls towed in Delft in the 1970s.

Under ORC that benchmark is handed to you, imperfect and rule-shaped, and then used to decide your result. Under IRC nobody hands you anything, which means the spreadsheet is not the amateur option. It is the only option.

When everything comes together your data enables the sailmaker to create the perfect solution but remember - it’s a circular routine and new sails mean more testing and changes to the VPP which then means better routing - this is why we love yacht racing - always something to improve ( Image from (C) North Sails LLC with thanks)

Polar Rules

• Calibrate the instruments before refining the model. Bad TWA makes the whole exercise worthless.

• Know the provenance and date of every polar you hold, and what displacement and sea state it assumed.

• Treat designer polars as optimistic and rating polars as rule-shaped.

• Under IRC, assume nobody is going to give you a performance model. Buy an ORC Speed Guide or build your own.

• Sort logged data into cells by wind speed and angle, take a high percentile, and count the points in every cell before you trust it.

• Keep your baseline numbers where your own data is thin, and mark where that is.

• Run several polars at once. They answer different questions.

• Re-derive every season. The boat changes, the sails change, and the rating model changes too.

Pretty polars make good pictures. Honest polars make better decisions.

A request, and an admission

I have made a fairly confident argument here on the basis of published documentation, manufacturers’ own technical material, and forty years of doing this on my own boats. That is not the same as certainty, and there are three places in particular where I would welcome being corrected.

If you know of an instrument system that genuinely revises its polar table from sailed data on board, rather than accepting one you loaded, I would like to hear about it.

If you have measured your own boat against a designer polar, a rating polar and a set built from your own logs, I would be interested in the size and the shape of the differences you found, and particularly whether the divergence was worst where I would expect it, which is light air and reaching.

And if you work at a sailmaker or a rating office and any of the above misrepresents how the work is actually done, say so and I will publish the correction. I would rather be usefully wrong in public than comfortably vague.

Stuart Greenfield
FirstShift — Yacht Racing Intelligence
From first shift to first place.

Offshore racing is never just about the boat, the forecast, or the routing software.

It is about decisions made under pressure, in changing conditions, by tired people trying to keep the boat moving fast and safely in the right direction.

That is where races are won and lost.

At FirstShift, my aim is simple: to help owners, skippers and crews race offshore with more clarity, more confidence, and better preparation whether that is a RORC Channel race, a Fastnet campaign, or your first serious step into offshore racing.

If this article helped you think a little more clearly about the next race, please consider subscribing.

Free subscribers receive regular offshore racing notes, tactical insights, safety thinking, and race-preparation ideas.

Paid subscribers help support the work and receive deeper analysis, including more detailed race briefings, tactical notes, routing thoughts, and practical offshore performance intelligence.

Thanks for reading.