Unstable Intelligence
Index Assets Methodology Proof

Daily intelligence report · 08:00 UTC

The Unstable Index

Fear & Greed tells you how the crowd feels. This tells you how fragile the system actually is — measured, not guessed.

Unstable Index —loading
Assets scored —across 9 classes
Risk dimensions 8weighted sub-indices
Early warning 1–3whistorical lead
Last computed —daily 08:00 UTC

The Index

methodology v2.1
— / 100
loading
—
03055 7085100
30-day history —
Fragility
—
Stress
—
Assets stressed
—
Top driver
—

A high score does not predict a crash. It says the system has less slack than usual — the same conditions that preceded Terra, FTX and the March 2023 depeg.

Sub-indices · score / weight higher = more fragile

Why eight and not one

Peg & leverage carry 40% together — they are what actually breaks first. Liquidation & volatility add 30% — how violently a break propagates. Security, flows, narrative 25% — the slower signals that precede it. TradFi 5% — rates and credit stress leaking into crypto.

Every input is on-chain or from a public venue, timestamped and hashed daily. No discretionary overrides, no analyst adjustments. When a source is missing the sub-index returns a neutral value and says so, rather than silently scoring zero.

Assets

—
#AssetClassRisk type FragilityScore

Ranked by fragility score, highest first. The public board shows the six most fragile; the full set of 100+ scored assets, exact scores and per-signal drill-down open with the free trial.

Ranked by fragility score, highest first — the most fragile assets the index is tracking, recomputed daily at 08:00 UTC from public on-chain sources.

How it is built

reproducible by design
01 · Collect

Public sources only

Chain state, Curve and Uniswap pool balances, funding and open interest from major venues, reserve attestations, bridge flows, and 17 macro tickers. Everything a second party could pull themselves.

No private feeds, no paid oracles
02 · Score

Eight dimensions, fixed weights

Each signal maps to one of eight sub-indices on a 0–100 scale. Weights were set once from historical events and have not been tuned since — which is the only way a backtest means anything.

Weights published, never silently changed
03 · Amplify

Correlation matters

Independent stress in three places is noise. Simultaneous stress across peg, leverage and flows is a system with no slack left. A CISS-style amplifier lifts the composite when sub-indices move together.

Same construction the ECB uses for TradFi
04 · Publish

Hashed and frozen daily

One run per day at 08:00 UTC. The result is written once and hashed, so a score can be checked after the fact against what was actually published — not what the model would say today.

No retroactive edits

What it would have caught

backtested, same weights
Critical 91.2

Terra / Luna — May 2022

UST peg deviation plus Luna supply inflation pushed the score to 91.2 six days before the cascade. Curve pool imbalance at 78% UST and $2.4B of Anchor outflows were the leading inputs.

6 days of lead time
Danger 84.7

FTX — November 2022

Exchange fragility: abnormal FTT concentration, a collapsing collateral-to-liability ratio and unusual withdrawal patterns took the exchange sub-index to 84.7 four days before the freeze.

4 days of lead time
Elevated 72.3

USDC depeg — March 2023

Reserve attestation staleness against Circle's $3.3B at Silicon Valley Bank, plus Curve imbalance, flagged USDC at 72.3 18 hours before it traded at $0.877. The macro overlay picked up the FDIC announcement.

18 hours of lead time

Backtests are run with the weights fixed at their original values. They show what the published methodology would have produced, not a model fitted after the fact to events everyone already knows the ending of.