03 · Rules before results

The rules behind every receipt.

CryptoWatch publishes what it said, how it was resolved, and the rules used to review it. This is the plain-language guide to the Accountability Ledger.

Publication rules · Deterministic resolution · Public verification Last updated: July 17, 2026 · Version 1.2

01 · Publication

When a row is published

A ledger row is added when one of these public events occurs:

Trigger	Call type	Horizon	Benchmark
Internal setup evidence records a setup or no-trade observation	`desk_setup`	T+1d	BTC spot
Truth Engine verdict changes	`truth_engine_verdict`	Not graded	Hash-chain evidence only
CW Risk band changes	`cw_risk_verdict`	Not graded	Hash-chain evidence only
Historical Manipulation Shield HIGH/CRITICAL transition (producer retired 2026-07-15)	`manipulation_shield_flag`	Not graded	Hash-chain evidence only

Not a callTooltips, educational content, and raw data panels are not calls. Historical Shield rows remain immutable, queryable, and hash-verifiable, but they are not directional bets and no new Shield calls are produced.

02 · Resolution

How outcomes resolve

Only call types registered with the automated resolver are graded. Today that is desk_setup; evidence-only verdicts and historical Shield rows stay outside the hit-rate calculation. Setup-path grading runs continuously, and the ledger overview resolver runs nightly at 00:05 UTC.

Setup track record: LONG/SHORT setups are graded first-touch-wins against their declared stop and take-profit levels. If both touch in one candle, the stop wins. A setup that reaches its validity deadline without either is EXPIRED.
No-trade observations: retained as observations and marked EXPIRED at their validity deadline; they are not reclassified as directional calls.
Ledger overview: the separate alpha resolver currently grades only BTC pins with historical BTC benchmark coverage. It uses alpha versus BTC with a ±1 percentage-point NEUTRAL band. Non-BTC pins remain unresolved until token-specific historical prices are wired.

Conservative by designNeutral counts as a miss for the public hit-rate computation. Only HITs fill the numerator.

03 · Context

How market context is recorded

Every call is stamped with the macro-data regime at the time it was made. The same outcome is read in context rather than treated as if every market day carried the same conditions.

The classifier deterministically scores the US 10Y Yield, DXY, VIX, M2 YoY, and High Yield Spread into three bands: TAILWIND, NEUTRAL, or HOSTILE. The exact inputs are logged into the hash-chain before the call.

04 · Integrity

How the chain stays auditable

Once a call is written to the ledger, it is appended with a cryptographic hash connecting it to the previous row:

row_hash = sha256(canonical_json(row_without_hashes) + prev_row_hash)

Every night at 00:30 UTC, CryptoWatch computes the ultimate chain_tip_hash and broadcasts it as a 0-value memo transaction to the Base blockchain. That gives independent reviewers a durable reference for checking the public record.

Need to check a specific row? Run the verification yourself →

Revision history 2026-07-17: Version 1.2 aligned the methodology with the dedicated Bitcoin State page, five-decision Observatory, and current public verification routes.
2026-07-15: Version 1.1 reconciled the page with the live resolver. Shield/Hunter producers are retired; historical Shield rows remain verifiable but are not graded.
2026-04-23: Version 1.0 published. Initial spec launch.