Criterion register
Each dashboard gauge carries pre-registered movement criteria: the specific, checkable evidence that would move it one level worse or one level better. This page holds every criterion the desk has published, current and retired, with the wording it replaced, the date it took effect, the brief it took effect in, and the reason it changed.
It exists because a pre-registered test is only worth something if readers can see when it was rewritten. Before Sept 8, 2026 the desk recorded why a classification moved but not when a criterion changed while the classification held, which meant a stable-looking gauge could be sitting behind a test that had quietly been replaced. Rows dated before Sept 8, 2026 were reconstructed from published brief text. Where the reason for a revision was not recorded at the time, the row says exactly that rather than supplying one after the fact.
Criteria are never edited in place. A revision inserts a new row and stamps the row it replaces. Movement criteria are published for gauges at orange or red and for any gauge with a live trigger. For how classifications and verdicts are defined, see sources and methodology. For the classifications themselves over time, see the history page.
Yen carry trade
2 in force · 4 retired
Worsens to ORANGE ELEVATED on a USD/JPY close above 160, on any single-session move of three yen or more in either direction, or on a Bank of Japan hike larger than 25 basis points on Sept 18.
Why this test: Rewritten for a YELLOW starting level and to put a named scheduled event inside the test. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves to GREEN NORMAL on ten sessions inside a three-yen range after Sept 18, with no widening in credit spreads.
Why this test: New criterion written after the prior improve test fired and the gauge moved to YELLOW. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Worsens on a sharp yen rally with cross-asset deleveraging: crosses breaking lower, with crypto and equities selling together.
Why it was replaced: Reason not recorded at the time. Backfilled Sept 8, 2026. The revision is explained by the gauge improving to YELLOW on Sept 4, which required a worsen test back to ELEVATED rather than to RED, and by the Sept 18 BOJ meeting entering the window.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Worsens to RED UNWIND if USD/JPY crosses 160.00, or on a disorderly unwind.
Why it was replaced: Material revision, reason partially recorded. The Aug 28 brief states the asymmetry explicitly: a weak-yen break of 160 raises intervention risk, not unwind risk, and the unwind trigger is the opposite move. The numeric 160 level was therefore dropped from the worsen test. Backfilled Sept 8, 2026.
Methodology pre-v1.0
Improves on USD/JPY stabilizing below 156 without fresh intervention.
Why it was replaced: Retired by firing, not by revision. The criterion fired Sept 4 with Bank of Japan data showing no intervention, and the gauge moved from ORANGE ELEVATED to YELLOW WATCH. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves on a sustained move below the mid-150s with stable JGBs.
Why it was replaced: Reason not recorded at the time. Backfilled Sept 8, 2026. The revision replaced an imprecise range ("mid-150s") with a checkable level of 156 and swapped the JGB condition for an intervention condition.
Methodology pre-v1.0
Oil shock risk
1 in force · 3 retired
Improves to ORANGE ELEVATED on an ICE Brent front-month settlement below $90.00. Improves to YELLOW WATCH below $80 with verified recovering transit.
Why this test: Two-step improve ladder introduced because the gauge sits at RED and a single test could not distinguish a one-level recovery from a two-level one. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Worsens to RED SHOCK on a Brent close above $95.
Why it was replaced: Retired, not revised. The criterion fired Sept 3 when Brent settled above $95 and the gauge moved to RED, the top of the scale, which carries no worsen step. Backfilled Sept 8, 2026 from published brief text.
Methodology pre-v1.0
Improves below Brent $80 with transit recovery verified across two independent trackers.
Why it was replaced: Reason not recorded at the time. Backfilled Sept 8, 2026. The revision is explained by the gauge reaching RED on Sept 3, which required an intermediate improve step that had not previously existed.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves to YELLOW WATCH if Brent settles under $80.
Why it was replaced: Reason not recorded at the time. Backfilled Sept 8, 2026 from published brief text. The revision added a verified transit-recovery condition to a test that had been price-only.
Methodology pre-v1.0
Hormuz risk
2 in force · 2 retired
No worsen criterion. The gauge is at RED DISRUPTION, the top of the scale.
Why this test: Backfilled Sept 8, 2026 from published brief text.
Methodology pre-v1.0
Improves to ORANGE ELEVATED on five consecutive sessions of commodity transits at or above the ten-day average with no new vessel strike.
Why this test: Made countable so the test can be checked from tracker data rather than judged. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves to ORANGE ELEVATED on a formal reopening, or on verified sustained transit recovery toward prewar volumes across at least two independent trackers.
Why it was replaced: Reason not recorded at the time. Backfilled Sept 8, 2026. The revision replaced a qualitative recovery test with a countable one.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves only on verified reopening of the strait to normal transit.
Why it was replaced: Reason not recorded at the time. Backfilled Sept 8, 2026. The revision added a second, lower path to improvement (sustained transit recovery) alongside formal reopening.
Methodology pre-v1.0
Bond stress
2 in force · 2 retired
Worsens to RED STRESS on a US 30Y close above 5.50.
Why this test: Closing basis specified so intraday spikes cannot fire the gauge. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves to YELLOW WATCH below a US 30Y of 5.00 without official support in the market.
Why this test: Threshold tightened and an official-support exclusion added. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Worsens to RED STRESS if the US 30-year crosses 5.50%.
Why it was replaced: Wording revision only: the threshold was unchanged, but the test was tightened from "crosses" to a closing basis. Reason not recorded at the time. Backfilled Sept 8, 2026.
Methodology pre-v1.0
Improves to YELLOW WATCH below a US 30-year of 5.10%.
Why it was replaced: Material revision, reason not recorded at the time. Backfilled Sept 8, 2026. The threshold moved from 5.10% to 5.00%, making improvement harder, and a condition was added requiring the level to hold without official support after Treasury buybacks failed to hold the long end in the Aug 21 week.
Methodology pre-v1.0
XRP flows
2 in force · 3 retired
Worsens to ORANGE WEAKENING on a verified net weekly outflow. A downgrade does not require two weeks.
Why this test: Asymmetry with the improve test made explicit. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves to GREEN POSITIVE on two consecutive weeks at or above the $39.78M reference, verified across trackers.
Why this test: Two-week requirement restored after the single-print test fired and moved the gauge to YELLOW NEUTRAL. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Worsens to RED NEGATIVE on a net-negative weekly ETF print.
Why it was replaced: Clarification, not a change in substance: the Sept 4 wording states that a downgrade does not require two consecutive weeks, correcting an asymmetry with the improve test that had been left implicit. Backfilled Sept 8, 2026.
Methodology pre-v1.0
Improves on a weekly total verified across at least two trackers at or above the $39.78M bar.
Why it was replaced: Retired by firing. The criterion fired hours after the Aug 28 publication when CoinGlass corroborated the SoSoValue weekly total of roughly $110.5M, and the gauge moved to NEUTRAL. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves toward YELLOW on two consecutive weekly prints materially above the collapsed level. Print one of two arrived in the week ending Aug 21 at $39.78M.
Why it was replaced: Revision replaced the undefined phrase "materially above the collapsed level" with the $39.78M print as a fixed numeric bar and added a two-tracker verification requirement. Reason not recorded at the time. Backfilled Sept 8, 2026.
Methodology pre-v1.0
Macro backdrop
2 in force · 5 retired
Worsens to RED HOSTILE on a September Fed hike delivered alongside Brent above $100.
Why this test: Anchored to two named observables and a dated meeting. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves to YELLOW MIXED on a Sept 16 FOMC decision to hold rates with ICE Brent front month settling below $100 that day.
Why this test: The prior improve test was a one-time check on the August CPI release (core below 0.2% month over month with hike odds under 30%). August core printed 0.3% on Sept 11 and hike odds rose to about 90%, so the test failed and could no longer fire. The replacement mirrors the existing worsen test (a Sept 16 hike with Brent above $100), using the same $100 level already in the register. Numeric and checkable; no classification moved.
Methodology 1.1
Worsens to RED HOSTILE on a Fed hike delivered into a US 30Y above 5.50, with risk-off confirmation.
Why it was replaced: Material revision, reason not recorded at the time. Backfilled Sept 8, 2026. The rate leg was replaced with an oil leg and the September meeting was named explicitly, changing which observable settles the test.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Worsens on bond stress reaching RED, or on Brent reaching $95.
Why it was replaced: Material revision, reason not recorded at the time. Backfilled Sept 8, 2026. The Aug 21 test was a one-of-two disjunction on levels the desk already tracked elsewhere. The replacement required three conditions together and introduced a Fed policy action, which is a materially harder test. This is the revision an outside reviewer identified on Sept 8 as invisible to readers, and it is the reason this register exists.
Methodology pre-v1.0
Improves to YELLOW MIXED on August core CPI below 0.2% month over month with hike odds under 30%.
Why it was replaced: Re-anchored to a dated release (CPI Sept 11) and a published odds level. Backfilled Sept 8, 2026.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves on a Fed hold with stable long rates and broadening risk appetite.
Why it was replaced: Material revision, reason not recorded at the time. Backfilled Sept 8, 2026. Numeric checkability was restored by anchoring to a named data release and a hike-odds level.
Methodology 1.0.x (exact patch level not recorded per criterion; backfilled)
Improves on September hike odds below 20% with Brent under $80.
Why it was replaced: Material revision, reason not recorded at the time. Backfilled Sept 8, 2026. A numeric two-condition test was replaced with a qualitative one, removing the checkability the pre-registration rule is meant to guarantee.
Methodology pre-v1.0
Gauges with no published criteria
Global liquidity, Risk appetite. These gauges have not sat at orange or red in the window the register covers and carried no movement criteria at the time. The gap is stated rather than filled. A criterion appears here the first run one is written.
Methodology changelog
Append-only. Material changes to gauge definitions, verdict rules, source rules or classification logic increment the version.
v1.0 (effective 2026-08-22): Formal methodology baseline. Covers the ruleset as it stands today: five-verdict claim taxonomy, eight-gauge dashboard with pre-registered movement criteria, three brief modes with rolling Friday baseline, source tier hierarchy with Tier 1 preference, XRP spot stamp rule, ETF flow print discipline with two-tracker cross-check and two-print upgrade criterion, source label integrity checks, archival honesty and append-only claim history. Four run conventions live from this date: (1) Evidence verified through line in every brief dashboard section, (2) mandatory adversarial counter-evidence checkpoint before Top 3 finalization, gauge transitions, and XRP interpretation, with receipt in internal run summary, (3) gauge change_reason format: pre-registered criterion, observed value with timestamp, source, resulting transition, (4) brief corrections preserve original text and append a dated CORRECTION note with receipt. Rules predating this baseline (gauge rename, Rules 10 and 11, two-print criterion) are pre-v1.0 history documented in past briefs. Versioning scheme: v1.0.x wording only, v1.x material rule change, v2.0 redesign.
v1.0.1 (effective 2026-08-29): Coverage addition, no rule change. The Watch page carries a live XRPL amendment voting panel (every amendment open for voting, validator support count and percentage, majority and earliest activation dates), fetched from XRPScan and cached hourly, stamped with the fetch time in ET. Amendments become watch items only once they reach 80% support, carrying the earliest activation date, and resolve on activation with the ledger as receipt. Amendments do not move any gauge. An amendment is Verified only once enabled on ledger; before that the desk states vote counts and dates, never timing predictions.
v1.1 (effective 2026-09-08): Two material rule changes, both prompted by an external editorial review of the public site.
1. Gauge criterion register. Pre-registered movement criteria are now recorded in the public gauge_criteria table, append-only, with old rule, replacement, effective date, effective brief and change reason held together. The register is publicly readable in full and is not gated on brief published state, because methodology is not gated. Previously the desk logged why a classification moved but not when a movement criterion was rewritten while the classification held, which meant a reader could not distinguish a stable classification from a changed test. The register is backfilled from published brief text for Aug 21, Aug 28, Sept 4 and Sept 8, 2026 across the six gauges that carried Moves If lines. Two gauges, global_liquidity and global_risk_appetite, have no rows because they were not at orange or red in that window and carried no criteria; that gap is stated rather than filled. Backfilled rows carry the reason for the revision only where it is recoverable from the published text. Where no reason was recorded at the time, the row says so. Going forward, every criterion revision is written to the register in the same run that changes it, with its reason, and no criterion is edited in place.
2. Claim scoping rule. A claim is worded narrowly enough that its verification badge applies to the entire sentence. Unverified specifics inside an otherwise verified event are held in the evidence body, not in the claim line. A primary document that is silent on a claim fails to support it and is graded UNVERIFIED; CONTRADICTED requires reliable evidence against the claim. Where an existing claim mixes scopes, it is split into separate rows rather than rewritten, and the original text is preserved with a dated scope note. Applied on this date to the 21-institution bank stablecoin venture claim, re-graded from CONTRADICTED to UNVERIFIED on scope with the narrow debunk split into its own CONTRADICTED row, and to the Kharg Island tanker claim, which keeps its VERIFIED verdict for the strike, vessel and date while a scope note records that the four-missile count is an uncorroborated Iranian state media figure.