Depeg Duration Resolver Changelog
Full version history of DDR methodology decisions, from v1.0 to v5.2.
What this controls
Depeg Duration Resolver Changelog version history
Use this changelog to trace how Pharos changed the related scoring, data-source, threshold, or interpretation rules over time. Entries record the contract active at each release; archived rules are retained so historical charts and citations can be read against the rules active at the time.
Primary references
Latest version
Clocked forecast and native supply evidence
Held and expired comparisons cannot establish live deviation or forecast readiness. Supply factors use comparable token counts when recorded. Earlier history uses the change in reported USD supply.
Impact snapshot
- Training distinguishes quote provenance from assessment-time landmark eligibility. Review dispositions and admitted event coverage govern eligible analytics.
- Previously sealed forecast contents and identifiers retain their original inputs. Exposure and measured dollar-flow denominators remain USD-valued.
v5.2Nov 19, 2026Clocked forecast and native supply evidence
Held and expired comparisons cannot establish live deviation or forecast readiness. Supply factors use comparable token counts when recorded. Earlier history uses the change in reported USD supply.
- Training distinguishes quote provenance from assessment-time landmark eligibility. Review dispositions and admitted event coverage govern eligible analytics.
- Previously sealed forecast contents and identifiers retain their original inputs. Exposure and measured dollar-flow denominators remain USD-valued.
DetailsHide details
Clocked forecast and native supply evidence
Held and expired comparisons cannot establish live deviation or forecast readiness. Supply factors use comparable token counts when recorded. Earlier history uses the change in reported USD supply.
- Training distinguishes quote provenance from assessment-time landmark eligibility. Review dispositions and admitted event coverage govern eligible analytics.
- Previously sealed forecast contents and identifiers retain their original inputs. Exposure and measured dollar-flow denominators remain USD-valued.
Impact notes
- Training distinguishes quote provenance from assessment-time landmark eligibility. Review dispositions and admitted event coverage govern eligible analytics.
- Previously sealed forecast contents and identifiers retain their original inputs. Exposure and measured dollar-flow denominators remain USD-valued.
v5.1Oct 31, 2026Stage-1 contradiction fixes for supply coverage and live direction
DDR v5.1 fixes two Stage 1 contradictions that could publish a No break signals verdict from unavailable supply evidence or a stale event direction. Stage 2, the sub-component versions and sealed predictions are unchanged.
- K1 supply weaponization and K4 reflexive death-spiral use the seven-day supply-change proxy only when mint-surge coverage is explicitly `supply-history-proxy` or null (legacy unmarked); an explicit `unavailable` coverage marker stays unavailable and cannot stand in for the hourly proxy.
- An `above`-direction event whose live signed deviation is itself a below-peg break deeper than 1,000 bps is evaluated by the below-peg kill signals and tier mapping. A stale stored direction can no longer publish No break signals for a live deep break that its own depth readout already uses.
- The resolution rubric, duration model, incident grouping, support rules and reviewer versions are unchanged; these fixes move only the affected new assessments and forecasts, and sealed predictions keep their original content, hashes and version stamps.
- Activation is 2026-10-31 00:00 UTC (1793404800); no Stage 1 replay is claimed.
DetailsHide details
Stage-1 contradiction fixes for supply coverage and live direction
DDR v5.1 fixes two Stage 1 contradictions that could publish a No break signals verdict from unavailable supply evidence or a stale event direction. Stage 2, the sub-component versions and sealed predictions are unchanged.
- K1 supply weaponization and K4 reflexive death-spiral use the seven-day supply-change proxy only when mint-surge coverage is explicitly `supply-history-proxy` or null (legacy unmarked); an explicit `unavailable` coverage marker stays unavailable and cannot stand in for the hourly proxy.
- An `above`-direction event whose live signed deviation is itself a below-peg break deeper than 1,000 bps is evaluated by the below-peg kill signals and tier mapping. A stale stored direction can no longer publish No break signals for a live deep break that its own depth readout already uses.
- The resolution rubric, duration model, incident grouping, support rules and reviewer versions are unchanged; these fixes move only the affected new assessments and forecasts, and sealed predictions keep their original content, hashes and version stamps.
- Activation is 2026-10-31 00:00 UTC (1793404800); no Stage 1 replay is claimed.
Impact notes
- K1 supply weaponization and K4 reflexive death-spiral use the seven-day supply-change proxy only when mint-surge coverage is explicitly `supply-history-proxy` or null (legacy unmarked); an explicit `unavailable` coverage marker stays unavailable and cannot stand in for the hourly proxy.
- An `above`-direction event whose live signed deviation is itself a below-peg break deeper than 1,000 bps is evaluated by the below-peg kill signals and tier mapping. A stale stored direction can no longer publish No break signals for a live deep break that its own depth readout already uses.
- The resolution rubric, duration model, incident grouping, support rules and reviewer versions are unchanged; these fixes move only the affected new assessments and forecasts, and sealed predictions keep their original content, hashes and version stamps.
- Activation is 2026-10-31 00:00 UTC (1793404800); no Stage 1 replay is claimed.
Earlier versions
Compact summaries retain historical citation targets. Open a version archive for its full evidence and commit provenance.
- v5.0 · Evidence-class verdicts, competing-risk duration odds and reviewer v5
Oct 25, 2026
DDR v5 condemns a coin only on terminal evidence, publishes duration as unconditional recovery odds that keep non-recovery visible, and grades its calls against naive baselines.
- v4.9 · Measured recovery evidence and chronological duration cohorts
Oct 20, 2026
DDR requires measured onset evidence, uses run-local admitted structural posture and groups duration intervals without losing open or overlapping members.
- v4.8 · TVL trend evidence respects stored liquidity measurement epochs
Oct 19, 2026
DDR admits 7-day and 30-day TVL changes only when the historical baseline and stored current liquidity row share a TVL-measurement epoch. A measurement-policy step is unavailable trend evidence, not an observed exit collapse.
- v4.7 · Timed-only duration history and independent reviewer coverage
Oct 10, 2026
DDR duration training now uses only genuinely timed onset and current-deviation evidence rather than backdating final peaks into earlier landmarks. This is a score- and verdict-moving input change for future assessments and forecasts: reduced support can suppress duration bands until timestamp-evidenced recalibration. Sealed and first-published predictions are not rewritten.
- v4.6 · Unknown audit provenance fails closed
Sep 28, 2026
DDR eligibility and repair admission now reject unknown non-null audit verdicts rather than admitting them through a negative exclusion list.
- v4.5 · Derived mint posture drives structural recovery inputs
Sep 12, 2026
DDR now uses the published V9 mint posture for structural fragility and recovery signals, with curated posture fallback when the derived posture is unavailable.
- v4.4 · DDRR coverage accounting counts each incident once
Sep 9, 2026
DDRR state-assignment and finalized-coverage shares are now counted per incident instead of summed across state and operational-cause buckets, so an incident whose assigned state also carries an operational cause is counted once. No prediction, verdict, duration, or accuracy figure changes; only the two coverage-completeness shares move.
- v4.3 · Continuous depeg windows and reason-authoritative recovery labels
Aug 23, 2026
Depeg onset and recovery windows now require consecutive observations, with a 1200-second continuity tolerance deliberately set above the 900-second producer cadence so ordinary scheduler jitter does not read as a coverage break, while DDR training and DDRR review classify recovery from explicit closure reasons with a legacy price fallback only for null-reason rows.
- v4.2 · Mint-posture predicates unified; mint-scoped non-inflatability recognised
Aug 10, 2026
DDR's mint-posture tests now read shared predicates instead of per-engine literal sets, and the mint-scoped `none-resolved-mint` posture earns the weak R1 non-inflatable-supply anchor it previously fell through entirely. No resolution tier, factor weight, duration landmark, or stratum assignment changes.
- v4.1 · Structural inputs read published Safety Score outputs
Aug 8, 2026
DDR's mint-authority structural input now reads the published Safety Score V9 mint posture band instead of the retired standalone Mint Authority band. Prediction weights, factor rules, and exit thresholds are unchanged.
- v4.0 · DDR v4 Methodology Contract
Jul 30, 2026
DDR v4 updates terminality signals, duration landmarks, incident lifecycle grouping, support rules, and reviewer audit metadata for new predictions.
- v3.04 · Rollout Coverage Boundary Review
Jun 29, 2026
DDRR now treats rollout-active incidents whose reliable recovery or terminal evidence predates the DDRv2 public prediction contract as pre-lock coverage outcomes instead of live missed-lock debt.
- v3.03 · DDRR-Calibrated Stage 1 Terminality
Jun 29, 2026
Uses reviewed DDRR outcomes to calibrate Stage 1 terminality: recent compromised/unbacked mint incidents now feed K1, while static very-high reserve concentration becomes severe K2 only when paired with a severe below-peg fingerprint or observed dependency impairment.
- v3.02 · Close-gap Tail Grouping and Superseded Alias Review
Jun 19, 2026
DDR incident grouping now treats live reopens inside the documented close-gap merge window as the same canonical incident even when the original start is days old, and DDRR follows superseded duplicate aliases to the current source event.
- v3.01 · Live Context Input Wiring
Jun 6, 2026
Wired Stage 1's documented live-context inputs into the Worker DDR precompute path.
- v3.0 · Forecast Readiness Contract
Jun 4, 2026
Added the shared DDR forecast-readiness contract for readiness-gated locks with immutable public metadata.
- v2.0 · Sticky Public Prediction Contract
May 27, 2026
Replaced live DDR drift with DDRv2: one immutable public prediction or no-call at the 24h lock landmark, backed by manifest publication and append-only errata.
- v1.1 · Depeg Duration Resolver Reviewer
May 25, 2026
Added the Depeg Duration Resolver Reviewer (DDRR), the audit layer that scores stored DDR assessments against later canonical depeg-event outcomes.
- v1.0 · Initial Depeg Duration Resolver
May 25, 2026
Launched the two-stage Depeg Duration Resolver: a mechanistic Resolution Outlook (terminal vs recoverable) and a stratified empirical duration estimate over recovered historical incidents.