How every rankingis computed.

1. 1

   Criteria to methodology points

   Each criterion is scored 0–10 from public evidence. Its contribution is weight ÷ 100 × score ÷ 10 × 32, so a category's criteria together control 32 of the 40 objective points. The remaining 8 are preference points: unclaimed in the ChainChoice Score wherever a category is scored on its own criteria, because nothing is stated there, and claimed by the attributes you state in a Fit for you run; they tilt, they do not decide. A criterion without a receipt is not scored — it contributes nothing and is not read as 0.

   Fit for you re-weights the criteria rather than adding a bonus: a criterion you mark required has its published weight doubled and the rest renormalise; preferred multiplies it by 1.4. The sentence "your weighting moved cost from 18% to 33%" is one you can check.

2. 2

   Five dimensions to a weighted total

   The criteria dimension (objectiveFit in the feed) joins availability, the friction dimension (frictionFit), trust bonus and quantitative data under published base weights; the total is the ChainChoice Score. A base weight is how much a dimension could move a score; each category also publishes how much it actually did.

   availability20objectiveFit30frictionFit10trustBonus8quantitative18profileFit0

   Weighted total

   weightedTotal = round((SUM over dimensions of (breakdown[d] / baseWeight[d]) * weight[d]) / (SUM of weight[d]) * 86, 6 decimal places), where every breakdown[d] is a multiple of 0.1

   Absolute score

   absolute = min(99, max(40, round(weightedTotal <= 20 ? 40 + weightedTotal * ((60 + (20 / 86) * 38 - 40) / 20) : 60 + (weightedTotal / 86) * 38)))

   Pool-relative score

   score = round(min(min(99, max(41, leaderScore - (leaderWeighted - weightedTotal) * 2.6)), scoreOfRowAbove))

   These dimensions are computed and published in every breakdown but carry no weight in the total. profileFit left the composite in methodology 2026.09.2: it was constant inside all 114 categories, so its 20 points moved every product in a pool identically and ordered nothing. Sum the breakdown against baseWeights and the unweighted terms are simply absent from the arithmetic — that is the intended reading, not an omission.

   A base weight is how much a dimension COULD move a score, not how much it did. Each category publishes dimensionSpread: the range that dimension actually covered across its pool. A spread of 0 means the dimension was constant there and decided nothing, however large its weight.

   Read verbatim from rankings.json, generated 2026-09-27. Scores are clamped to 41–99.

3. 3

   Pool-relative score, then the Overall rank

   The leader anchors the pool; every other row sits below it by its weighted distance, and no row may print above the row above it. The order itself comes from the chain below, the one the engine applies and the feed publishes as formula.tieBreakChain; each key is read only when every key before it is equal.

   1. availabilitySource, for a reader who states a country only: in perp_dex, perp_house_vaults, options_vaults, prediction_markets, a row whose availability in that country rests on a region tag its record gives no source for (no availability note quoting the provider's terms, and no dated country entry) sorts below every row whose availability is sourced. No row carries it in the orders published in this file, which state no country.
   2. weightedTotal as step 1 states it, higher first. Only rows whose weighted totals are exactly equal reach the keys below.
   3. unroundedWeightedTotal, higher first: step 1 over the dimension points before they are kept to a tenth of a point, published per product. Since methodology v2 a shared weighted total from different unrounded points is ordered by those points, so only rows equal on both reach the keys below.
   4. criteriaAverage, higher first: the row's unrounded weighted average of its published criterion scores (0 to 10) under the weights in use. Applied only when both rows are scored on their category's criteria and the averages differ; otherwise the keys below decide. Since methodology 2026.09.12.
   5. tiebreakDelta, higher first: the sum of three factors. (1) Price: when the category has one comparable pricing basis that may decide rank (priced for at least two of its catalog rows, not an entry-plan figure derived from a price ladder, and quoted in a single currency), a priced row earns 0 to 2 points in proportion to where its price sits between the highest and the lowest in the pool, or 1 point when every published price on that basis is equal; a row with no price on that basis, and every row of a category without such a basis, earns 0. (2) Strength: 1 point when at least three of the row's catalog fitStrength values are 9 or above; those values are its published criterion scores, plus further strength keys on some rows. (3) Dimension: 1 point when the row scores at least 85% of the maximum on the highest-weighted dimension of the weights in use (the first in weights order when two are equal). A row outside the alternate product catalog (every payment_gateway row) carries 0. It orders rows and is never added to a score.
   6. objectiveFit points, higher first.
   7. product name, alphabetical.

   The published per-category tieBreakRules are not yet applied by the engine; where two rows share a weighted total and an unrounded weighted total, the chain above decides. Since methodology 2026.09.12 the first key after the totals is the two rows' unrounded weighted criterion averages, so a tie between rows scored on the published criteria is ordered by those criteria; the generic factors decide only where the averages are equal or a row has no criterion scores. tieBreakDecisions counts only pairs equal on both totals, and againstCriteria lists any of them still ordered against the criteria.

   Determinism is asserted in the test suite: the same inputs always produce the same order, across every live category.

1. v2026.04.0-launch2026-04-01First publicly versioned methodology. 11 categories with scoring weights, 5-dimension framework (fees, trust, support, control, liquidity), 6-question default decision flow per category.
2. v2026.04.02026-04-27Outbound provider links now resolve to region-appropriate URLs at click time. US visitors clicking a US-restricted exchange are routed to the operator's US entity (where one exists) or shown a region-block notice instead of a dead link.
3. v2026.09.02026-09-01Trustpilot and App Store averages no longer contribute to any score. They remain visible as evidence on the provider breakdown, labelled as context rather than a scored dimension. The quantitative dimension keeps its 0-18 range: fee and regulatory scoring are rescaled to span it, so the balance between the six dimensions is unchanged.
4. v2026.09.12026-09-02scoreAvailability compared each of a product's markets against the visitor's country with String.includes. On the un-personalised path — every /best page, every prerendered route and the rankings.json feed itself — no country is stated, so the needle was the empty string and every market name matched. Two products were silently excluded from every published ranking, and a 20-versus-18 step became "does this row have a non-empty primaryMarkets array".
5. v2026.09.22026-09-02profileFit is no longer weighted. It took two values across 934 rows — 14 everywhere except the twelve payment_gateway rows — and was constant inside 114 of 114 categories, so its 20 of 106 points moved every product in a pool by an identical amount and never once decided an order. The published total is now 86 points across five dimensions. profileFit is still computed and still shown in every breakdown; it simply no longer pretends to rank.
6. v2026.09.32026-09-06The eligibility gate now reads the decision layer's country verdict for every row before anything is scored. Until now the engine excluded a row only when the visitor's country was one of the 27 the region table maps; a visitor from any of the other 33 questionnaire countries — India, Nigeria, New Zealand, Hungary among them — was ranked against providers whose own market list omits their region, labelled "verify availability" as if it were unknown. The verdict is the same test on the 27 mapped countries (identical on all 26,217 country-by-row pairs) and resolves the other 33 through the wizard's own region at region resolution, which the verdict says. Only an ineligible verdict excludes; unknown is kept and labelled, exactly as before.

Machine-readable schemas for the decision primitives are published at /methodology/spec/ under CC-BY-4.0.