Market Trust
ExperimentalDecide whether a market price is strong enough to cite.
Search the markets already in the rated set to see one of four answers, with the caveat and receipt attached: use it, use it with caveats, discount it, or do not cite it. Most rated prices land at discount or do not cite, and the honest reason is coverage: the average rated market has fewer than two of six evidence pillars measured, and the top tier requires four. That is a statement about how much we have measured, not a judgement about the venue.
Strong enough only with the current evidence trail attached.
Useful, but the caveats should travel with the number.
A weak input for decisions until missing evidence improves.
Too thin, stale, incoherent, or unresolved to rely on.
One scale, strongest warning on the right · 6 markets graded in the canary preview set
No market currently clears the evidence bar to cite. Most prices land at discount or do not cite because the top tier needs four of six evidence pillars measured and the average market has fewer than two. A market moves toward Use as we measure more of it, not only as the market improves.
Market Trust remains a preview until resolved-outcome history and the frozen validation contract are mature enough for v1.
Rated markets
6 graded · experimental| No rated markets in the current preview set. | |||||
More views+
1. Start with use, use with caveats, discount, or do not cite.
2. Carry the caveat with the market price.
3. Open the receipt only when you need the technical proof.
Experimental rating caveat
Market Trust Cards are experimental market-quality diagnostics, not compliance certifications or validated credit ratings.
Market-quality scorecard
One plain-English status plus a pillar-by-pillar explanation of whether a market is currently usable, caveated, discounted, or too weak to rely on.
Visible caveats
v0.2 separates measured inputs from heuristics, so a buyer can see which parts of the rating are mature and which are still evidence collection.
Audit trail
Each card carries source artifacts and a row hash, linking the public card back to the reproducible Convexly research ledger.
Future stable-card shape
A stable card must answer the buyer's next question.
This is still a preview surface. A stable rating is not just a higher score. It is a card a user can cite in a memo because the verdict, evidence, limits, and next action are all visible together.
Verdict
Can I quote this probability?
Use, use with caveats, discount, or do not cite, with the exact methodology version attached.
Evidence
What supports the verdict?
Coherence, depth, participant quality, resolution reliability, integrity-risk screens, and source health.
Limits
What is missing or blocked?
Unknown flow share, stale snapshots, dropped events, data-rights gaps, and heuristic inputs travel with the score.
Next action
What should I do with it?
Cite with caveats, monitor for freshness, request a partner feed, or wait for resolved-outcome evidence.
Current canary
Current candidate cards with visible evidence boundaries.
These cards come from the current Market Trust candidate table. They are useful for internal review and buyer demos, but remain a preview until freshness, coverage, and outcome-history gates stay green over time.
Methodology & validation statusthe work behind the verdicts: maturation lanes + readiness gates
Research preview status
What this preview can and cannot tell you yet.
The canary model should make a market legible to an advisor, journalist, venue operator, or institutional risk team: what supports the rating, what is still weak, and what has to resolve before Convexly can make stronger claims.
Current card issues
These are live card-health issues, not long-term validation requirements.
Before a stable rating
These are evidence requirements for public promotion. They can remain open while the internal canary batch is healthy.
Candidate cards
6
latest audit run
Measured inputs
28%
10/36 pillar inputs
Clean source
6/6
no stale snapshot or dropped events
Sensitivity-stable verdicts
6/6
sensitivity-stable cards
Pending
9
inputs still being collected
Blocked rights
0
data unavailable by policy
Pillar measurement
Participant quality
1/6 cards measured
Public wallet flow is joined into Edge Score cohorts when sample floors are met. The remaining work is coverage breadth, not inventing the pillar.
View wallet cohorts
Order book
Liquidity and depth
1/6 cards measured
Canary cards now distinguish shallow snapshots from measurable depth, so weak markets are downgraded instead of hidden behind a single score.
Open canary proof
Integrity
Integrity-risk screen
1/6 cards measured
Flow concentration and pattern screens are treated as evidence flags. They are not proof of intent, and they should travel with every rating.
Read method
Freshness
Source-health gate
6/6 cards clean
A card should not graduate while its source snapshot is stale or event drops are present. Freshness is part of the rating boundary.
Inspect source health
Robustness
Sensitivity check
6/6 verdicts stable
The score must survive leave-one-pillar-out and weight-perturbation checks before the verdict is safe to show as durable.
Review guardrails
Validation
Resolved-outcome ledger
0/6 cards unblocked
This is the claim-upgrade gate. Market Trust stays experimental until resolved cohorts show whether high-trust cards were actually more useful.
Track outcomes
Method
Formula freeze
0/6 cards unblocked
The formula must be frozen before validation so the backtest cannot silently move the target after seeing outcomes.
View preregs
Production readiness
What we're still proving before this is a stable rating.
Market Trust is usable as a preview diligence packet, not a stable public rating contract. The goal is a packet a buyer can cite because the verdict, caveat, source status, participant evidence, integrity screen, and validation limits are all visible together.
Current contract
Status: canary_preview. A formal internal validation review is required before this moves from preview to a stable public rating.
Resolved outcome ledger
BlockedConvexly cannot claim Market Trust tiers predict cleaner or more useful markets until resolved card-time rows exist.
Now: The preview cards carry validation placeholders and promotion blockers; resolved-outcome evidence is still the hard gate.
Next: Write resolved card-time rows, show the denominator, and publish positive and negative result handling.
Formula freeze
BlockedA buyer should know the score was frozen before validation, not tuned after outcomes were visible.
Now: The promotion contract exists, but the public formula and thresholds are not yet validated as frozen.
Next: Freeze weights, thresholds, caps, and sensitivity rules in code and link the receipt before any stable-rating language.
Source freshness
In progressA current-card verdict must degrade when Polymarket or Kalshi inputs are stale, failing, or dropping events.
Now: Candidate cards already expose source health and freshness gates; sustained clean operation is still required.
Next: Maintain the clean-source window required by the promotion contract and keep card language downgraded during stale windows.
Participant-quality coverage
In progressThe rating should know whether meaningful market flow came from scored public participants or mostly unknown flow.
Now: Participant evidence is joined where public wallet-flow floors are met; coverage is not broad enough for a stable public rating.
Next: Raise scored-notional coverage, keep unknown-flow caveats visible, and separate aggregate-only venue limits.
Integrity-risk screen
In progressIntegrity language must stay a risk screen until reviewed positive and negative cohorts prove detector quality.
Now: The preview screen exposes concentration and pattern flags without turning them into intent claims.
Next: Label review cohorts, measure false positives and false negatives, then publish the screen limits.
Outcome calibration
BlockedThe four verdicts need observed calibration before Convexly can say one tier is empirically safer than another.
Now: The verdict framework is visible, but the resolved cohort needed for calibrated tier language is not mature.
Next: Compare use, use-with-caveats, discount, and do-not-cite cohorts against clean resolution and reviewer labels.
Stable public snapshot
The stable public snapshot remains available while the candidate model earns production history. Search is intentionally simple: market slug, question, category, or condition ID.
No public Market Trust Cards match that filter.
The stable public snapshot exposes a limited daily sample while Convexly collects the evidence needed for broader candidate-model validation.