How we measure a boat’s worth.
Every figure on Hull Gauge is an asking price observed on a public listing, counted once per boat, and published only when enough boats support it. These are the rules, how the numbers are worked out, and how accurately we read the listings.
The ten rules
Each rule has its own link. Cite it as “Hull Gauge methodology §4”.
Rule 1. We measure asking prices, and we say so.
Every figure is what sellers are asking on public listings, not what boats sold for. Sold prices are held privately by brokers and portals, and we have not yet measured how far below asking boats sell, so we never adjust for it silently.
Labelled “asking, not sold” wherever a price appears.
Days listed is how long we have seen a boat advertised so far, never a time to sale: we do not yet see when a boat sells.
Sold-price calibration will be published by segment, only from at least 10 sales reported by 3 or more firms, once brokers contribute sold prices.
Rule 2. A boat counts once, however many sites list it.
The same boat often appears with three brokers and two portals, sometimes at different prices. Counting each would overweight the boats that are hardest to sell. We recognise the same boat across sites and follow its price over time.
Matched on hull identification number where one is published, otherwise on model, build year and length. A hull number shared by unrelated boats (placeholders such as “N/A” or “0000”) is ignored, so two identical boats without one may count as one.
A boat whose length is more than 15% from its model’s is not assigned to it. Each boat counts at its latest asking price.
A boat has cut its price when its ask is below an earlier ask on the same site, in the same currency, at least 14 days before; a drop of half or more is held back as a probable error. It is listed now when an advert for it, on a site we read live, was seen in the last 7 days.
Rule 3. Every price is placed against the right model, year and currency.
A price is only useful against boats of the same model and generation. “Oceanis 38” and “Oceanis 38.1” are different boats to a buyer, so they are different models here. Prices are converted at the rate on the day they were observed.
Model names resolved against a curated list of builders, models and aliases; build-year ranges routed to a separate generation where one exists. Daily reference exchange rates.
Rule 4. Below five boats, we publish nothing.
A median of three boats is an anecdote. When too few boats support a figure, we print the reason instead of a number, and point to the nearest figure we can stand behind.
Model figure: 5 boats; its middle half: 10. Year bands and market splits: 10. Gauge Estimate and its likely range: 5 comparable boats. Year pages: 5 built in the year and 8 within 3 years either side.
Depreciation curve: 20 boats across 6 build years, with a fit whose typical miss is under 42%. Price drivers: 5 boats on each side. Market-wide cut share and days listed: 200 boats, once the live reading is 90 days old. Gauge Report: 8 comparable boats.
Tiers: Firm 25+, Indicative 8–24, Provisional 5–7.
Rule 5. Every figure carries its basis.
A number without its sample size, window and date cannot be checked. Every figure we publish carries all three, and the pages that cite them repeat them.
Shown with each figure: boats counted, years and markets included, and the date it was recomputed. Estimates also state the window used, as the worked example does: “built 1987–1993”.
Published prices are rounded to the nearest £500 at every size, running costs to the nearest fifty pounds and daily figures to the nearest ten. A single boat’s ask is printed as asked, in its own currency.
A specification is printed only with its source: a cited page, or “from listings” with the boats counted (a median of at least 5). A value without a source is not printed.
Rule 6. Over time, we compare like with like.
A market median rises when more new boats list, even if no boat is worth more. Our price level compares the same models at the same age a year apart, so a change means prices moved, not the mix.
Older asks are restated in today’s prices by a yearly factor for their market, from the same model and build year against the last twelve months. The price level is the same-age comparison, published by market and year. Both are in the table below.
The monthly index is withheld until twelve months of continuous observation exist: September 2027.
Rule 7. We never republish a listing.
We store facts derived from listings, not the listings. No photographs, descriptions, seller names or links to the advert appear on Hull Gauge, and comparable boats are shown by month, not day.
Public, per boat: model, year, asking price (in today’s prices and as asked), length, beam, draught, hull material, engine, hours, VAT status, market, country where the listing states it, the month observed and the month first seen, days listed as a band, the size of any cut and whether it is listed now. Not public: the advert, the site it was on, the seller, the exact date, text or images.
The public key the site uses reads only published views and reference tables; the observations themselves cannot be read with it.
Rule 8. We ask permission to read.
We read only sites whose robots rules allow it, identify ourselves, and keep a list of sites we will not read even where we technically could.
Crawler identifies itself as HullGaugeBot with a link to its policy; follows robots.txt, including groups addressed to AI crawlers, and records any Content-Signal header; asks again only for pages that have changed where a site supports it, at most one request a second per site. Full policy on the bot page.
Rule 9. We publish our errors.
Accuracy is measured, not claimed. We check samples of what we read against the original page, publish the result, say what failed, and date every change to the method.
Scorecard below. Changes to the method are logged at the foot of this page. Corrections: corrections@hullgauge.com.
Rule 10. Nobody pays to change a number.
Sellers, brokers and advertisers cannot buy a placement, a correction or a better figure. Paid reports add detail to the same numbers everyone else sees.
What we charge for: buyers’ reports and broker tools. Neither can alter a published figure; brokers who contribute sold prices do so privately and never see another broker’s data.
How a figure is worked out
Worked through for a 1990 Grand Banks 42 Classic, our example boat (chosen from the Firm year pages, and kept while it stays one), with the statistics of 5 November 2026. It is the same calculation its year page makes.
- 1Collect the model’s boats
Every Grand Banks 42 Classic we have observed since Aug 2020, each counted once at its latest asking price and restated in today’s prices. Two have no stated build year and can’t be placed by age.
141 boatsAug 2020 – Oct 2026 - 2Take comparable years
Boats built within three years of 1990, so the estimate rests on more than the six built that year.
46 boatsbuilt 1987–1993 - 3Weigh them
Closer build years count more, and boats last seen more than 24 months ago count less, halving for every further 24. All markets count equally here; given engine hours, boats in the same hours band would count more.
UK · Europe · US · elsewhere27 of 46 down-weighted - 4Take the weighted middle
Half the weight of comparable boats sits above this figure and half below. It is rounded to the nearest £500 and published as the Gauge Estimate.
£131,000Gauge Estimate - 5Set the likely range
The middle half of comparable boats, by the same weights: a quarter of the weight sits below the range and a quarter above.
£108,000 – £162,000likely range · Firm - 6What a year of age does
A robust straight line through log price against age (Theil–Sen), so one mispriced boat can’t bend it: each year of age takes about 4.2% off. It describes the model; it does not move the estimate. The year page uses it to look ahead.
4.2% a yearlost to each year of age, fitted on 139 boats
A rule of thumb, not a measurement: on a year page, an asking price below 0.45 times the Gauge Estimate or above twice it is read as far outside anything we have seen. We suggest checking its currency, model and year, and don’t place it on the ruler.
Confidence tiers, shown beside every figure
The most boats behind it. Safe to quote, with its date.
A sound guide; it can move as more boats are seen.
A first reading. Check the boats behind it before relying on it.
No figure. We say why and point to the nearest one we can stand behind.
Earlier asks, in today’s pricestwo yearly measures
Each earlier asking price is divided by a yearly factor for its market and the year we saw it, so asks from different years sit on one scale. The factor compares the same model and build year a year apart, so it mixes the market’s movement with a year of ageing for each year back. The price level compares boats of the same age, so it shows whether the market itself moved. We restate with the first and publish the second.
The factor an earlier ask is divided by
| Seen | UK | Europe | US |
|---|---|---|---|
| Nov 2024 – Nov 2025 | 1.10 | 1.09 | 1.08 |
| Nov 2023 – Nov 2024 | 1.15 | 1.13 | 1.15 |
| Nov 2022 – Nov 2023 | 1.27 | 1.17 | 1.18 |
| Nov 2021 – Nov 2022 | 1.27 | 1.08 | 1.15 |
| Nov 2020 – Nov 2021 | 1.12 | 1.02 | 1.11 |
| Nov 2019 – Nov 2020 | 1.05 | 1.00 | 1.08 |
Same model, same build year, against the last twelve months, each market in its own currency. Above 1, the same model and build year asked more then than now.
The price level, same age (last twelve months = 100)
| Seen | UK | Europe | US | All |
|---|---|---|---|---|
| Nov 2024 – Nov 2025 | 100.0 | 102.3 | 102.8 | 102.5 |
| Nov 2023 – Nov 2024 | 101.6 | 98.9 | 100.1 | 100.0 |
| Nov 2022 – Nov 2023 | 98.8 | 100.6 | 99.8 | 99.9 |
| Nov 2021 – Nov 2022 | — withheld | 88.5 | 90.4 | 90.4 |
| Nov 2020 – Nov 2021 | 87.2 | 77.9 | 82.0 | 80.4 |
| Nov 2019 – Nov 2020 | — withheld | 72.3 | 71.4 | 72.0 |
The same models at the same age, each earlier year against the last twelve months: below 100, boats of that age asked less then than they do now. All takes in the UK, Europe and the US together. A dash is withheld: fewer than 30 matched boats.
Measured accuracy
How often we read a listing correctly: 201 listings, each checked field by field against the page as it was captured. The reviewers were four AI agents, not people, who split the sample between them, on 22 September 2026. A human spot-check of their verdicts is next.
201 listings · a field counts only where the reviewer could verify it · extractors as of 22 Sep 2026, before the two fixes listed here · includes 1 listing from a third site, too few to show on its own
100 listings · a field counts only where the reviewer could verify it · extractors as of 22 Sep 2026, before the two fixes listed here
100 listings · a field counts only where the reviewer could verify it · extractors as of 22 Sep 2026, before the two fixes listed here
What went wrong, and what we did
Every build-year miss was a classic from 1882–1897 that fell below a 1900 floor. The floor is now 1800.
The listing record carried VAT status in a field we weren’t reading. Now read. One boat stating VAT only in its description is still missed; we don’t read description text.
Each page gave both a nominal length and a length overall; we took length overall, the reviewer compared against nominal. Corrected, that set’s length accuracy is at least 98%. The table keeps the figure as measured.
The copy of the page we checked appears to be a neighbouring capture of the same advert. We will check each capture’s own timestamp when we read it.
The page gives only a county (“Lying · Lancashire”). Revisited when that site is sampled properly.
Every night
Every figure is rebuilt from the observations in this order. Each step runs on its own: if one fails, its figures from the night before stay up and the failure is logged. If the first step or the latest prices fail, nothing after them runs; if the model statistics fail, the steps built on them wait for the next night. Nothing half-computed is published.
- 03:30 UTCRebuild the boats
Every observation re-read; boats matched across sites; prices placed on models.
- thenEarlier prices restated
Each earlier year’s asks compared with the last twelve months, same model and build year, by market: the factors that restate them in today’s prices.
- thenPrice level
The same models at the same age, each earlier year against the last twelve months, by market.
- thenLatest prices
Each boat’s current asking price, in pounds, euros and dollars, with its cuts and days listed so far.
- thenCuts and new boats
Each price cut recorded as an event; the pooled chance of a cut by how far over its estimate a boat is asked; newly listed boats matched to saved searches.
- thenModel statistics
Medians, middle halves and counts for every model, market and build-year band, over seven years and the last twelve months; sheet numbers for newly published models; the quarter-end Gauge Values watch letters compare with; the brokers’ stock readings.
- thenAge curves
Depreciation fitted for every model with 20 boats across six build years, then pooled for each builder.
- thenBuilders, types and pairs
Builder and type figures, the model pairs worth comparing, and the UK against Europe gap.
- thenReference figures
Typical values for the guides, the example boat on this page, and the site’s counts.
- thenWeekly edition
Once a week, the market figures are kept as printed, so a citation can point at them.
- thenHousekeeping
Searches that found nothing are deleted after 180 days; account, order and broker records past their retention period are removed.
- thenPages and emails
Every page re-renders from tonight’s figures on its next visit, then the morning’s alerts and emails go out.
- 07:00 UTCWatchdog
If no clean run has finished in 26 hours, the miss is logged and the pages keep yesterday’s figures.
The figures on the site now were computed 5 Nov 2026, 03:30 UTC.
What we can’t tell you yet
How close estimates are to sold prices
We have no sold prices yet, so we can’t publish estimate error against sales. It will be published by segment once brokers contribute sold data.
How long boats take to sell
We see when an advert appears, not when a boat sells, so we print days listed so far and never a time to sale. It will be measured once we record when adverts end, over at least ninety days of complete reading.
What VAT status is worth
Few listings state VAT status. We publish a VAT difference only where a model has five boats on each side; everywhere else we say it isn’t measured, and no estimate is adjusted for it.
Accuracy on the live sites
The scorecard covers listings read before the live crawl began. The nine live broker and marketplace sites we read get their own checked sample after a month of reading.
A monthly price index
Needs twelve months of continuous observation. Until then, the yearly level is the honest reading.
Anything about one boat’s condition
We see what the listing says, not the boat. Survey findings, rig age and interior wear are for a surveyor.
Changes to this method
Newest first. We never edit a published figure quietly; when the method changes, it’s recorded here.
- Prices rounded to £500 at every size
Published prices now round to the nearest £500; below £100,000 they rounded to the nearest £100, so many model figures moved by up to £200. Running costs round to the nearest fifty pounds and daily figures to the nearest ten.
- Price cuts counted like for like
A boat has cut its price when its ask is below an earlier ask on the same site, in the same currency, at least 14 days before; drops of half or more are held back as probable errors. This replaces a count of any price change, which included currency movements and rises.
- Days listed so far
How long a boat has been advertised is the days between the first and last time we saw it, over boats seen on at least two dates. It is labelled days listed so far and is never presented as a time to sale.
- Middle half from ten boats
The middle half of observed asking prices, for a model, a build year or a band, is printed from ten boats everywhere; below that only the median is shown. Nothing is published from fewer than five.
- Specifications cited or measured
A specification value is printed only with a cited source, or as measured from the boats behind the page (a median of at least five, with the count). Values without a source are no longer printed.
- Same-age price level
The published price level compares the same models at the same age a year apart, by market. The yearly factor that restates older asks (same model, same build year) is unchanged, and both are shown on this page.
- Sheet numbers
Every published model has a permanent sheet number, assigned once when it is first published, so a citation still points at the right model if its name changes.
- Listed now
A boat is listed now when an advert for it, on a site we read live, was seen in the last seven days. If our reading stops, the count falls rather than holding.
- No raw observations are public
The public key the site uses now reads only the published views and reference tables. The observations themselves, their exact dates and the sites they came from cannot be read with it.
- Method described as the code runs
This page described a Gauge Estimate as a reading of the age curve. It is, and was, a weighted median of comparable boats within three build years, with recent boats counting more. The worked example now shows the real calculation, and the scorecard says who checked it: AI agents.
- Model generations split by build year
Where a builder reused a name for a new generation, those build years are now a separate model, with the reason recorded.
- Older asks restated in today’s prices
A yearly factor by market, from the same model and build year against the last twelve months, restates each earlier ask; statistics and curves are now stated in today’s prices.
- New boats excluded; qualifier changes tracked
Listings for new, unregistered boats make no boat. A change from “asking” to “offers over” now counts as a price event.
- Defective observations set aside, never deleted
Pages that were not a single boat’s listing, such as search and category pages, had been read as boats. They are excluded from every figure; the rows stay in the ledger with the reason.
- Build years from 1800; VAT status read from the listing record
Both fixes from the first scorecard. Existing observations are re-read on the next capture of each listing.
- Year pages admitted at the estimator’s floor
Five boats in the year and eight within three years, instead of ten in the year, which would have admitted eight pages in total.
- Valuation, running costs and the market gauge published
The valuation tool, a running-cost estimator from cited tables and the market gauge went live; the first accuracy measurement was published; statistics are recomputed nightly.
- Confidence tiers lowered to match real samples
Firm from 25 boats rather than 30: three-year windows rarely reach 30 even for the best-observed models.
- Method first published
Observations from March 2020 onward; nothing published below five boats, and ten for year bands and market splits.
Found something wrong? Write to corrections@hullgauge.com. Brokers and owners can also ask us to remove a boat from the observations.