Frozen 2026-09-05
The pre-registered forward test of Market Trust (AsPredicted #295174) returned the registered wording "Market Trust tiers do not separate on resolution cleanliness." A null is a valid pre-registered outcome, not a failure. Market Trust is frozen and this promotion contract was never exercised; this page is kept as dated history and does not describe a current product. The calibration verdict artifact is docs/ops/evidence/market-trust-v1-calibration-verdict/2026-07-30-terminal-analysis-point.json.
The registered label must not be read alone. The higher-trust arm held four markets, so no attainable outcome for that arm could clear the filed threshold, and the contrast is therefore not established. That is not the same as the arms being alike: no equivalence margin was filed, so the frozen result settles the question in neither direction.
This contract defines when Convexly may move Market Trust beyond v0.2 canary-preview language. It is grounded in the lessons from v0.1 static scorecards and v0.2 live packets: separate the verdict from evidence confidence, block current-card claims when sources are stale, and do not publish v1 language until resolved or reviewed outcomes exist.
Current cards
1
Need at least 30
Validation observations
0
Freshness, stability, and coherence windows
Resolved outcomes
1,000
0 human-review rows
Latest receipt
81b4cf2d...535463
market_trust_card_v0_2
Lessons learned
v1 cannot be a rename of today's packet. It must be the result of evidence that survived the problems exposed by the earlier versions.
v0.1
The first Market Trust bundle proved that a market can carry a verdict, caveats, pillar scores, and a receipt. It also showed why static bundles cannot support current-language claims without source freshness and live outcome monitoring.
v0.2
The canary pipeline added source-health gates, measured versus estimated versus pending pillar labels, integrity-risk screens, audit-chain verification, and memo language. The lesson: buyers need the verdict and the caveat, but Convexly must keep the machinery close enough to audit.
v1 gate
A large set of live cards is still not enough. v1 waits for clean source-health windows, measured liquidity/depth, measured participant-flow coverage, resolved or human-reviewed outcomes, and explicit founder approval of the formula freeze before validation.
Promotion gates
30 current cards across 3 categories so the sample is not a one-market artifact.
14 consecutive clean source-health days before current-card language gets stronger.
20 cards with measured liquidity and depth, not just estimated evidence.
10 cards with measured participant-flow coverage, each clearing a measured-share floor of 40% and staying under an unknown-flow-share ceiling of 60%. Coverage counts how much flow could be attributed; it scores no participant.
At least 1 resolved or human-reviewed row must be visible. Current count: 1,000.
The formula, thresholds, confidence caps, hard caps, and validation outcomes must be frozen before the validation result is inspected.
Amendment, 2026-08-27
Amended 2026-08-27 by founder direction. The gate this contract used to state as "measured participant-quality cards" is superseded by the participant-flow-coverage gate above: a measured-share floor and an unknown-flow-share ceiling. The 2026-08-09 decision not to build a participant-quality score is permanent, and the participant pillar renders unscored forever. What the gate preserves is the original intent: v1 does not promote while most of a market's flow is unattributed. The gate counts how much flow could be attributed at all; it does not score, rank, or judge any participant, and no future version of this contract may require that it does.
Gate rationale
v0.1 taught us that static cards can be legible without being current. v0.2 taught us that live cards can be current without being validated. The v1 candidate contract therefore separates operational liveness from statistical proof.
The contract requires 30 current cards so packet, export, route, and ledger behavior are exercised across more than one flagship example.
Boundary: It does not prove score accuracy or tier separation.
The promoted sample cannot be a single-event-family artifact.
Boundary: It does not prove category-general performance.
Current-language cards need a sustained operational window before buyers see stronger wording.
Boundary: It does not prove future uptime or enterprise SLA readiness.
The depth pillar needs a real measured denominator before it can influence stronger claims.
Boundary: It does not prove fillability or capacity.
Participant-flow coverage must be measured on more than one good card before it supports formula-freeze decisions. Coverage means an attribution denominator exists for the card's flow; it scores nobody.
Boundary: It does not score any participant, and it does not prove any wallet's behavior predicts outcomes.
A verdict computed over mostly-unknown flow is a verdict about the visible minority, so each qualifying card must clear a 40% measured-share floor and stay under a 60% unknown-flow-share ceiling.
Boundary: It does not prove full wallet coverage or cross-venue identity.
This proves the post-card ledger path can accept evidence before validation starts.
Boundary: It is a liveness marker only, not statistical validation.
Thresholds, hard caps, and confidence rules must be fixed before outcomes are inspected.
Boundary: It does not prove the frozen formula is good.
Coverage breadth, source-health windows, and measured liquidity/depth are defensible as floor gates. Flow coverage, integrity review, and calibration remain underpowered until the required rows accrue.
Boundary: It does not authorize stable v1, calibrated-rating, enterprise-ready, or manipulation-free language.
Draft formula boundary
This draft defines the candidate formula shape only. It does not authorize public v1-candidate, v1-stable, calibrated, or procurement-grade rating language.
use
80
draft cutpoint
use with caveats
65
draft cutpoint
discount
45
draft cutpoint
do not cite
0
draft cutpoint
Current claim boundary
Market Trust may be described today as a v0.2 canary-preview diligence card with verdict, caveat, source-health state, evidence confidence, and receipt. It may not be described as a calibrated rating, stable v1 system, compliance certification, manipulation finding, or proof that a market probability is correct.
Live blocker state: needs coherence outcome rows, needs freshness outcome rows, needs stability outcome rows.
Review ownership
Owns threshold rationale, power analysis, calibration design, formula-freeze readiness, and blocked quantitative wording.
Owns settlement-rule review, venue mechanics, participant-flow caveats, cross-venue equivalence rules, and data-rights labels.
Owns preregistration drafts, literature/citation hygiene, external-review packaging, and public limitation language.