data/auction-crisis.jsonβ¦The detector scans every settled US Treasury auction against six quantified historical-crisis signatures. Each signature is calibrated to fire at the same numerical thresholds that fired during real crises. Your composite is a weighted aggregate.
| Date | Term | Type | High% | BTC | Indirect% | PD% | Direct% | AAH% | Score |
|---|---|---|---|---|---|---|---|---|---|
| When | Tenor | BTC | Indirect | Tail | Score | Context |
|---|
| Regime | SPY | BTC | HYG | GLD | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1m | 3m | 6m | 1m | 3m | 6m | 1m | 3m | 6m | 1m | 3m | 6m | |
Source. All auction data is pulled from the official
api.fiscaldata.treasury.gov Treasury Securities Auction dataset (1979βpresent). Lambda
justhodl-auction-crisis-detector fires every 15 minutes during the Treasury publishing
window (14:00β22:00 UTC weekdays) plus a 4-hour off-hours backstop, so dashboard latency is at most
15 minutes from official auction publication.
The 6 patterns. Each is calibrated to fire at the *exact numerical threshold* that fired during real historical crises. Calibration anchors:
Composite. Per-auction score is 70% Γ worst-indicator + 30% Γ indicator average, weighted by accepted dollar size, rolled over a 14-day window. Issuance-explosion overlay can add up to 15 points. Regime thresholds: 0β30 CALM, 30β50 WATCH, 50β75 ELEVATED, 75β100 ACUTE_STRESS.
No backtest survivorship. The 2024-10-09 anchor was added LIVE before November 2024 volatility β it's not curve-fit. The detector genuinely caught that signal in real time.