Boat comparison

• The behaviour scores are the softest layer and about to become the most visible one: 1–10s assigned from listings, forums and reviews, sources biased toward owner enthusiasm. Defences: write a one-line observable-anchored rubric per dimension BEFORE scoring; score post-refit; mark evidence quality per cell and show the dots on the strip chart. Without rubrics the apparatus becomes a rationalisation engine with JSON output.
• expectedResale is the largest single guess and swings the final ordering more than any weight debate: ±20% on a $150k resale is ±$30k on the frontier's x-axis — bigger than most gaps between adjacent boats. Pull actual AU sold-comps per model (the count IS the confidence figure) and redraw the frontier at resale ±20% to see which positions survive. Stress-test the boats whose case rests on resale hardest: Ki, the Seawind.
• The flat 5%+10%+10% sale-cost model is per-boat wrong in ways big enough to reorder the frontier's right half: AUSFTA plausibly makes US-built boats (Catalina, both Hunters) duty-free with the paperwork right (~$5–7k swing); the Seawind is an Australian-built boat coming home (duty treatment + home-market liquidity currently priced at zero); French and Taiwanese builds pay the full stack; GST compounds on the duty-inclusive value so the error propagates; 10% brokerage is the top of the AU range. Per-boat sale-cost lines are an afternoon of work.
• Two of your own numbers disagree about the catamaran: the brief says ~$300k all-in after resale; the stored all-in model's central delta is $389,726 (best $238k, worst $636k). At $300k it is an expensive question; at $390k central it is close to indefensible outside Tradewind Liveaboard. Reconcile before drawing any chart with the Seawind on it. Its beam is likewise double-sourced (20'4" vs 21'4") and its haul-out gate turns on the answer.
• The gates as stated will misfire if run as stated: CSF < 2.0 on published dry displacement fails four production boats (2.01–2.11) while five of ten monohulls sit within 0.06 of the line — inside the noise of the displacement figure. Run stability as a hierarchy (AVS where measured, CE category + design assessment, CSF-loaded as coarse screen) and print margins. Same disease on draft: two boats at exactly 6.50, one at 6.56-by-spec vs 6.5-by-listing. Gates must run on measured values or they manufacture pass/fail theatre at ±1 inch.
• The brief says ten boats; the project holds models for eleven. The Jeanneau reads like a boat half-dropped at the draft gate. Decide its status explicitly — a method whose denominator is fuzzy is not yet a method.
• diySaving is a currency exchange, not a discount: it converts money risk into schedule risk against the one deadline with no whisker. The never-DIY list (rigging, seacocks, LPG, saildrive diaphragms) is insurer-constrained; the one large prize (Noctiluca's deck, ~180–200 h ≈ $11k) would spend half a year of DIY budget on the fleet's second-riskiest boat. Ghost dots on the frontier: visible, never banked.
• weeksToReady gates the wrong end of the calendar: the binding constraint is arrival in Brisbane before ~November 2028 cyclone onset, and boat speed changes how much of the fixed season each boat consumes. Add arrivalMargin per boat and gate on both ends.
• Within-cluster comparisons are diligence contests, not weighting contests: Ki and Noctiluca tie on design by construction and differ by $18k of risk expectation — the price of a thin listing. Do not expect the schemes to separate boats inside a cluster; that is the uncertainty ladder's job.

Each of these was considered and cut, mostly for not discriminating between these ten boats or for double-counting something already measured. Adding them back would add noise, not information.

• askingPrice (as a comparator) — Most visible number, least meaningful: ask ordering and all-in ordering are nearly uncorrelated in this fleet (Mayari $69.5k ask carries ~$69k central required work; Affinity $130k carries $34.9k). Keep only for offer strategy.
• offerWithinBudget — A property of the wallet, not the boat. Constraint input to the budget gate, never a comparison output.
• SA/D — Spread across the monohulls is ~14.1–15.4 — a rounding error. It answered 'is any of these a racer or a caravan' and the answer is no. Retire to the anchor table.
• CE design category (as a compared measure) — Category A or equivalent fleet-wide; zero discrimination. Gate input only.
• australianResale (the 1–10 score) — expectedResale/saleCosts/recoupedValue already hold this idea in dollars; a parallel score invites weighting the same fact in two units. The one residue — liquidity/time-to-sell — belongs as a confidence note on the resale estimate.
• motoringSpeed — Merged with motoringCapacity and speedLoaded into speedAndRange: expected loaded passage speed + range under power in nm. The tank spread (≈35 vs 130 gal ≈ 250 vs 900 nm) survives inside the merged score's rubric.
• motoringCapacity — Merged into speedAndRange (see motoringSpeed).
• payloadTolerance — Same idea as speedLoaded measured at a different load. Fix: define speedAndRange in full cruising trim (~a tonne for two crossing an ocean) and keep displacement-to-payload as an anchor input, not a dimension.