Claims require receipts.

One decision receipt

Market Compass uses one decision contract: source fact, DoubleDigitX interpretation, user belief, research posture, evidence completeness, freshness and expiry, friction, disagreement, and invalidation. A probability or confidence range appears only when an eligible independent model and calibration contract exist. Preview or unavailable states are never presented as live market claims.

Founding Flight: fixed historical practice

The guest flight uses the first three NIST Randomness Beacon 2.0 Chain 2 pulses strictly after January 1, 2024 at 00:00 UTC. The fixed rule was selected without reference to the outputs. Each question asks whether the first hexadecimal character of a future official pulse will be in 0–7. The information cutoff is the immediately preceding pulse. At runtime DoubleDigitX verifies the pinned dataset manifest, adjacent pulse chain, precommitment opening, status, timestamp, and output shape before serving the governed contract. The fixture records the NIST certificate and signature fingerprints, but this release does not claim an independent runtime RSA signature verification.

Before commitment, the future pulse output, answer, comparison, result-dependent evidence, and later path are absent from the HTML, initial serialized payload, metadata, client bundle, filenames, accessibility labels, and precommit APIs. This is an in-product blind historical exercise, not proof that a participant could not find the already-public historical result elsewhere.

Reference, outcome and proper score

The comparison is the YES frequency across the 60 official pulses immediately before each historical cutoff. That trailing window contains no target or later pulse and is fixed by the methodology, but it is only a base-rate comparator—not a claim that past beacon outputs predict the next pulse. It is not a market price, crowd forecast, live signal, fair value, DDX edge, or trade recommendation. The official NIST pulse record is the resolution authority. The score is the binary Brier Score, (p − o)², where p is the submitted probability from 0 to 1 and o is 1 for YES or 0 for NO. Lower is better.

The Replay aggregates exact integer score units before display rounding and compares only paired eligible rounds. Three practice outcomes are not enough to estimate skill, calibration, category strength, or repeatable edge. These rounds add zero Forecasting Passport evidence observations; the honest output is a provisional session description and what remains unknown.

Compass challenge derivation

Market Compass compares an explicitly entered user probability with the current visible bid and ask. Disagreement is zero inside that book; above the ask it is user belief minus ask, and below the bid it is bid minus user belief. Wide or thin books cannot produce the strongest research posture. Fees, slippage, size and model uncertainty are excluded and named, so this value is not expected value, fair value, edge, or a trade recommendation.

Markov Pulse: path, not outcome

Market Compass classifies only adjacent selected-outcome observations whose timestamps fall inside the versioned 45–75 minute cadence window. Those probability changes become falling, stable, or rising states using a one-percentage-point boundary; transition counts require contiguous intervals and never bridge a missing cadence. It is a market-price path diagnostic, not the probability that the event resolves Yes. A forward state distribution, interval, confidence, or event-outcome probability remains null unless a preregistered domain × horizon × adapter artifact passes out-of-sample calibration, independent-target, freshness, drift, and publication-authority gates.

Domains and honest abstention

A prediction contract always remains in the PREDICTION_MARKET forecast domain even when its topic is crypto, sports, or weather. Direct crypto assets, sports events, and weather events require separate adapters and cannot inherit a model merely from a tag. Source unavailable, stale, insufficient history, uncalibrated, out of domain, and quarantined are first-class publication states. They do not collapse to zero.

Public snapshot boundary

PublicSnapshotV1 is constructed field-by-field from an explicit allowlist. Private sessions, runtime identity, account state, positions, orders, P&L, experiments, raw evidence, internal paths, and trading controls are forbidden. If authority, licensing, or freshness cannot be established, the public value is withheld.

Missing is not zero

Unavailable, stale, insufficient, and quarantined evidence are named states. A numeric zero is shown only when an authoritative producer explicitly observed zero under a complete denominator.

Forecast ledger

Governed PAPER forecasts and abstentions are registered before their resolution window as immutable V2 publication receipts. The release candidate exposes only EXPERIMENTAL receipts because no artifact authority is wired. A revision is a new identity-matched linked record; the original is never overwritten. Settlement corrections are append-only, single-lineage resolution events, and Brier is derived from the immutable parent probability and binary outcome. Calibration groups preserve venue, horizon, feature, state-definition and methodology contracts; n includes scored forecasts only and the headline score is suppressed below 30. No private trading account, position, order, P&L, experiment parameter, feature vector, or runtime path is part of the public projection.

Market-data rights

The Founding Flight does not publish commercial market data. Its pinned source is the official NIST Randomness Beacon 2.0 historical record, with public reuse governed by the NIST public-access and license terms recorded in the dataset manifest. A commercial market comparison remains unavailable until exact display, caching, derivation, paid-use, publication, retention, redistribution, attribution, review, and owner evidence is inspectable. DoubleDigitX does not expose a vendor's raw feed as its own API.

Sustainability

Cost Coverage Ratio equals contribution-margin monthly recurring revenue divided by verified monthly infrastructure cost. It remains unavailable until both numbers cover the same verified monthly period. Waitlist entries and PAPER results are never counted as revenue.