Tortus Standards
Tortus Adjustment Standards
- Edition
- Edition 0.3 · August 2026
- Window
- 60 months, rolling
- Method
- Matched pairs
- Refresh
- Quarterly
A comparable-sales adjustment table for New York City, measured rather than asserted — every rate derived from closed sales, published as a band with its sample size, and withheld where the data cannot carry it.
Nobody in this industry publishes measured adjustment rates with their sample sizes and their suppressed cells. These standards do — including a public record of what each edition corrected, and why.
The three numbers
What is one more square foot worth?
The centerpiece question of any comparable-sales adjustment — and a quantity with different answers depending on who you ask. The distance between them is why these standards exist.
What appraiser convention implies
≈ 0.5 – 0.75
Every appraisal we reviewed sets one flat dollars-per-SF rate for the whole assignment. Read against those reports' own comparable prices, the convention works out to roughly half to three-quarters of a unit's average price per square foot.
What the market measures
0.60 – 0.82
Matched resale pairs, measured per segment and published as bands — with two cells suppressed and one withheld where the data cannot carry a rate.
Why published rates are bands
Not points
Vary any one defensible method choice — the trim, the size-gap threshold, the time window, the estimator — and the same data moves by more than a decimal's worth. A three-decimal rate would overclaim. The band is the honest envelope.
Appraiser convention and our measurements disagree about the level, and that disagreement — now narrow — is worth arguing about. What neither supports is pricing marginal space at a unit's full average rate. Everything below is the evidence, at the depth we are willing to defend.
The size clause
Marginal floor area, by segment.
Published as a ratio of a unit's own price per square foot — resale sales only, as bands. Where a cell reads withheld or suppressed, that is the publication: the reason is printed beside it, and no number stands in for it.
| Segment | Published band | Pairs | Buildings | Basis & caveat |
|---|---|---|---|---|
| Manhattan · condo | 0.72 – 0.80 | 2,106 | 504 | Revised down from a blended 0.85 – 0.92 — that figure was 77% developer-schedule pairs. See the correction below. |
| Brooklyn · condo | 0.72 – 0.82 | 1,073 | 177 | Stable across the resale / sponsor split — the two arms agree within 0.05. |
| Manhattan · co-op | ≈ 0.65 · range 0.60 – 0.81 | 530 | 206 | Effectively a resale segment already. The band is wide, and part of its width is an instability we have not explained — so the range publishes with the number. |
| Queens · condo | Withheld | 152 | 32 | 91% of this segment's pairs are sponsor sales — a developer price schedule, not a resale market. The resale pairs that remain are too thin to publish. |
| Brooklyn · co-op | Suppressed | 177 | 63 | Insufficient data. The crude 95% interval spans 0.36 – 0.92 — too wide to be a rate. |
| Queens · co-op | Suppressed | 16 | 11 | Insufficient data by any measure. |
Method for every rate above: matched pairs within a single building — room count exact, floor within two to three, sold within twelve months of each other, floor areas at least 75 square feet apart, both sides carrying a reported floor area — trimmed against the building's own median price per square foot to exclude non-arm's-length transfers. Pair counts are shown with the building count, which is the effective sample size.
Why the bands are measured on resales
A sponsor sale is a unit sold off a developer's price schedule — explicitly per square foot, with line and floor factors — in a building where nothing has traded before. Measuring a marginal dollars-per-foot rate there recovers the schedule, not the market's behavior.
Sponsor sales also over-sample badly: they close in bursts inside one building, which is exactly the shape a same-building pair design selects for. Manhattan condominium is 23.4% sponsor by sales and 77.1% by pairs — a factor of three. On resales alone, the Manhattan rate falls from a blended ≈ 0.92 to 0.72 – 0.80, and the three publishable segments converge rather than one standing apart.
A comparable-analysis subject is almost always a resale, so the resale rate is the one that publishes. The sponsor cohort is a real measurement of a real product — and it is not this one.
Read the bands as floors, not point estimates
Reported square footage is itself measured with error, and a size gap is the difference of two noisy numbers — which biases a measured rate downward. Restrict the pairs to larger, cleaner gaps and the rates rise in four of five segments. The true rates sit at or above the published bands.
That cuts both ways on purpose: it strengthens the finding that marginal space trades below a unit's average rate, and it is exactly why these figures publish as bands rather than as the three decimals they were computed with.
Why provenance matters
A fabricated control produces a plausible measurement.
The cleanest result in these standards, and the reason every rate now publishes where its inputs came from.
An earlier draft of these standards led with a striking finding: that a condominium is priced by the square foot and a co-operative is priced by the room, with co-op floor area measuring far below its published range. The finding was coherent, it explained three results at once, and it was wrong.
The cause: a large share of co-operative square footage in property records is not a measurement of the unit. Where no reported figure exists, records carry an imputed one — most often a building average. Restricted to sales carrying a reported floor area from a listing feed or an assessment record, the anomaly disappears entirely.
The mechanism is not the obvious one. Two imputed units almost never get compared to each other — carrying the same building average, they differ by zero square feet and fail the pairing rule. The damage comes from mixed pairs, where one side is real and the other is imputed: there, the measured size difference is not a size difference at all, it is noise, and pooling those pairs with the honest ones drags the whole segment's rate toward zero.
| Pair composition — Manhattan co-op | Pairs | Measured ratio |
|---|---|---|
| Reported + reported — the honest cell | 583 | 0.604 |
| Building average + reported — the dilution | 529 | −0.047 |
| Line median + reported | 348 | 0.455 |
| Building average + building average — the suspected culprit | 6 | −0.122 |
Read the two bolded rows against each other. The cell everyone would have blamed — imputed against imputed — contains six pairs. The cell that produced the false finding contains 529, at a measured ratio of roughly zero: pairs whose size difference carries no information at all, diluting an honest 0.604 down to a number that told a great story.
Restricting to reported floor area is the right correction, and it shrinks the co-operative sample by roughly half — which is why two cells in the size table read suppressed rather than carrying a number. The suppressions are the direct price of the correction, and paying them is the point. It is also why every rate in these standards now publishes the provenance of its inputs, not just its output.
Before trusting any measurement that claims to hold something constant, check where that something came from.
The other clauses
Where each adjustment stands.
Summary depth, same discipline: a clause publishes a measured band, a method, or the words 'no rate' — never a number the data cannot carry. Each status is stated on the clause itself.
Floor level
Measured · form unresolvedpercent of price per floor
One additional floor of elevation measures between +0.4% and +1.5% of price per floor, depending on the segment — condominiums high, Queens co-ops low.
Two qualifications keep this from publishing as a rate. The measure cannot separate elevation from what rises with it — light, view, and the better lines — so it reads as the value of height and its correlates together. And the appraisals we reviewed express floor in flat dollars where we measure a percentage; publishing one form where practice uses the other needs a stated reason, and this edition does not have one.
Monthly carrying cost
Measured · confound knowndollars of price per $1 of monthly charges
One dollar per month of common charges, maintenance, and taxes combined measures between $27 and $98 of price, by segment — co-ops at the high end, condos at the low end.
The confound is structural: a monthly fee is a proxy for size, services, and amenities, each of which practice adjusts on its own line. An appraiser who prices the fee as well counts the doorman twice. Until amenities are held constant in the derivation, the measured figures are an upper bound on the fee's own contribution — and this line is never stacked on an amenity adjustment.
Condition
Publishable after measurementper condition grade
Appraisal practice brackets a one-grade condition difference at roughly 4.4% to 7.2% of the comparable's price, stated as a percentage by one source and as flat dollars by another.
The widely cited 20 – 30% figure describes a gut-renovation gap, not one grade — a different quantity. No Tortus rate publishes here until the per-grade rate is measured, not inherited.
Market conditions over time
Method published, not a figurepercent per month
Every appraisal we reviewed takes zero time adjustment over a period its author judged stable — including comparables up to 22 months old — while published guidance runs as high as 1% per month. That is the widest disagreement between sources anywhere on this page, and it is why no single figure publishes.
What does publish is a method: an active listing entering as a comparable is discounted against its original asking price, on a named line of its own — never silently, and never against a later, reduced ask.
Bedrooms & bathrooms
No rate publisheddollars per room
Prior measurements held floor area 'constant' within an 8% tolerance. At Manhattan prices, that tolerance is worth more than the bedroom premium it was meant to isolate — and the measured premium tracks the tolerance almost exactly, from $48,500 where the size gap is zero to $128,000 where it is widest. The tolerance was a channel, not a control.
No room rate publishes until it is re-measured on a genuine size control. The clean-cell figures that survive do not yet form a pattern across boroughs and ownership types, and we will not average our way to one.
Private outdoor space
Unmeasuredratio to interior price per SF
Practice on this clause is genuinely inconsistent: across the guidance and appraisals we reviewed, the same terrace can be worth $0, $150 per square foot, or a six-figure lump sum. That inconsistency is the argument for measuring it rather than copying anyone.
Measurement is blocked on coverage — reliable outdoor area exists for only a small fraction of units. Presence is far better covered than area, and presence carries the clause until area lands.
Condition
Applied · unmeasuredpercent of price, by grade
A five-step ladder from Excellent to Poor, running ±12% of price, inferred from listing remarks, photographs and feed fields. It fires on roughly two comparables in five and moves more money than any clause on this page except size.
The ladder's steps are conventional. They were not derived here, and this edition publishes them rather than leaving them silent, because a reader is entitled to know which numbers in their valuation are measured and which are inherited convention. Measuring it needs a condition-controlled pair set — condition is inferred, so pairing naively would measure the inference rather than the market.
Building amenities
Applied · unmeasureddollars, by staffing tier
$25,000 for a part-time doorman, $50,000 for full-time, $75,000 for a full-service building. Round numbers, and named as such.
This line is now suppressed whenever the carrying-cost line fires, because a monthly common charge already prices the staff it pays for and charging both counts the same doorman twice. That is also why it cannot yet be measured: until staffing is held constant against the fee that funds it, any premium recovered would be the fee's, not the amenity's.
Parking
Applied · never firespercent of price
Eight percent, used to back a bundled parking space out of a comparable so it prices as a unit alone. In practice it fires on none of them: proving what conveyed with a sale requires the deed's legal description, and those are not available to us today.
It is published here anyway. A coefficient that is configured, is documented nowhere, and silently never runs is worse than one that is wrong, because nothing about the output tells you it is missing.
Small features
Not published · fitted, not measureddollars, grouped
Individually minor features — in-unit laundry, a fireplace, central air — are priced on their own lines and then reported as a single grouped amount, with the composition named. A grouped figure that does not say what is inside it cannot be reviewed, and a separate grid row for each small item buries the clauses that actually move the number.
No rate publishes here. The values in use are starting figures on the scale practice uses for a discrete system, and they are fitted against closed sales rather than asserted — the search is allowed to price a feature at zero, which is the outcome we would rather discover than assume. A fitted value is not a measured premium, and this clause will not claim one until it is derived on a controlled pair set like the size clause.
Method
Matched pairs, in plain language.
The estimator appraisers actually use — which matters for a document that has to be defensible inside a valuation report.
Take two closed sales in the same building: same room count, floors within two or three of each other, sold within twelve months of each other, floor areas at least 75 square feet apart, both carrying a reported floor area. Trim away prices far outside the building's own norm to exclude non-arm's-length transfers. The difference in price, against the difference in size, is one observation of what a square foot is worth in that building — and thousands of such pairs, taken together, are a segment's rate.
The method is published with its numbers rather than after them, because every correction in this edition came from a method choice nobody had written down.
What every published rate carries
The pair count
Every rate names the number of matched pairs behind it. A rate without a sample size is an assertion.
The building count
Pairs cluster — hundreds can come from one building. The building count is the effective sample size, and it is the honest one.
A sensitivity envelope
Each method choice — trim, size-gap threshold, time window, estimator — varied one at a time against the identical pair set. The band published is the envelope those choices produce.
Cohort composition
The resale versus sponsor share, measured at the pair level. Measured at the sale level it understates the sponsor share threefold, and it is why one segment is withheld rather than published.
An unrecorded feature is not an absent feature
The single most consequential rule in this method is also the least visible: when a fact about a property was never recorded, it is unknown, not absent. A comparable whose floor area, lot frontage, or laundry was never entered is still a comparable — it simply earns no points on that axis and no adjustment on that line.
Treated the other way, a gap in coverage becomes a penalty. The comparable is quietly excluded, or marked down for a feature nobody ever checked, and the result reads as a market judgement when it is really a record-keeping artifact. This is the failure we correct for most often, because it never announces itself: the number simply comes back, looking reasonable, computed over the wrong set.
Two consequences follow, and both are load-bearing. A filter never removes a comparable for a value it does not have. An adjustment line skips independently — one unknown feature does not suppress the others alongside it.
The refresh cadence
Rates are measured on a rolling 60-month window of closed sales and refreshed quarterly. A shorter window cannot detect a change smaller than its own noise — it would publish noise as a market trend, on a schedule.
Each edition is retained, not overwritten, and any edition-over-edition move smaller than the published band reports as unchanged. A rate that changes because a segment thinned out is not a market signal, and a standard that cannot tell the two apart is not one.
Corrections of record
What this edition corrected, and why.
A standard that cannot show its own corrections is asking to be taken on faith. Each edition publishes what changed, what was withdrawn, and the mechanism that caused it — the corrections are the feature.
Edition 0.2 published rates the valuation engine was not using
For one day these standards described a measured model and the engine ran the one that preceded it — floor area priced at 1.00 against a measured 0.60 – 0.82, a flat time adjustment on every comparable against a finding that no single rate publishes, and a carrying-cost line stacked on the amenity line that prices the same service. The rates were right and unreached. Edition 0.3 closes all three, and the numbers on this page are now read from the same measured object the engine reads, checked on every build. The failure is worth recording because nothing detected it: both the page and the engine kept working, and both looked correct.
Manhattan condominium size band: 0.85 – 0.92 → 0.72 – 0.80
The higher band was measured on a pair set that was 77% developer-schedule sales — a composition effect, not a market fact. On resales alone the rate is lower, and a comparable-analysis subject is almost always a resale, so the resale band publishes.
“Co-ops are priced by the room”: retracted
The finding rested on co-op floor area that was, in a large share of records, an imputed building average rather than a measurement of the unit. Restricted to reported floor area, the anomaly disappears — and every claim built on it goes with it.
Bedroom and bathroom rates: withdrawn
The 8% size tolerance in prior measurements was worth more than the room premium it was meant to hold constant. The premium was tracking the tolerance. No room rate returns until it is measured on a genuine control.
Point estimates: replaced with bands
Equally defensible estimator and trim choices move the same data by up to 13%. Three-decimal rates are gone from these standards; every published rate is a band with its sample size.
One caveat retired
The concern that assessment-sourced floor area might be building-level rather than unit-level was tested against the data and found groundless — the assessment-sourced cells are among the cleanest we have.
Sources. Rates are measured from closed condominium and co-operative sales in Manhattan, Brooklyn, and Queens over the 60 months to August 2026, restricted to sales carrying a reported floor area from a listing feed or an assessment record. The industry comparison draws on published appraisal-adjustment guidance and reviewed New York appraisal reports. Where a source document could not be independently verified, nothing in these standards relies on it alone.
Status. Edition 0.3, August 2026. A working edition: rates refresh quarterly on a rolling 60 months window, prior editions are retained, and future corrections will be recorded here.