Blockworks Advisory illustrative proof of concept for the Curve Risk Assessment and Market Monitoring proposal. Figures and tiers support review discussion.
Disclosure
About these numbers
How to read scores, tiers, and charts on this monitor. Posture labels are for triage. Downside estimates (CVaR, structural PD) appear where history supports them.
Scores and tiers
Monitoring posture for triage
L1 to L6 scores and posture labels (In range / Review / Elevated / Priority) are threshold crossings for review. Each market gets six dimension scores (0 to 100, higher is healthier) plus an overall posture. Crossing a util, soft liquidation, or liquidity threshold moves the label. CVaR and structural PD are separate estimates with their own status in the register below.
Thresholds
Where the boundaries come from
Measured quantities (utilization, debt, peg spot) are factual. The boundary that turns them into a posture label is a monitoring prior until it is backtested.
| Threshold | Value | Provenance | Note |
|---|---|---|---|
| Stablecoin stress utilization | 98%of utilization | monitoring prior | Expert prior for stablecoin lending stress. Not yet backtested in this repo. |
| Stablecoin efficient band kink | 90%of utilization | monitoring prior | Typical efficient band kink for stablecoin loan markets. |
| Volatile asset stress utilization | 88%of utilization | monitoring prior | Wider buffer for volatile loan assets. |
| Volatile asset efficient band kink | 75%of utilization | monitoring prior | Efficient band start for volatile loan assets. |
| Volatile asset tight utilization | 95%of utilization | monitoring prior | Volatile loan Review posture threshold. |
| Volatile asset critical utilization | 98%of utilization | monitoring prior | Volatile loan Priority posture when nearly fully lent. |
| Elevated soft liquidation share | 50%of soft liquidation inventory | monitoring prior | Elevated soft liquidation conversion. Mechanical on its own. |
| Critical soft liquidation share | 80%of soft liquidation inventory | monitoring prior | Band saturated conversion. Priority applies only with a second signal. |
| Soft liquidation materiality floor | $20,000USD | monitoring prior | Ignore soft liquidation share noise on tiny inventories. |
| Mint rate adequacy base premium | 2percentage points (plus util and soft liquidation add ons) | monitoring prior | Required premium ≈ 2 pp + utilization add on above 70% + soft liquidation add on. This premium is a monitoring heuristic. |
| Soft liquidation stress intensity exponent | 0.85dimensionless | monitoring prior | Convex conversion intensity (drop/100)^exponent on residual LLAMMA collateral. Progressive band fill proxy, not a fitted elasticity. |
Concentration
HHI scale and sample
DEX concentration uses a house 0 to 100 scale (sum of squared shares × 100). Shares are taken over the top DEX pools returned for the token and renormalized within that sample. The index is a top N concentration estimate inside that sample.
Rate heuristics
Rate adequacy and Secondary curves
Rate coverage and Secondary rate charts are monitoring heuristics. They help compare observed borrow rates to a simple premium floor or policy shape. The mint rate coverage index divides observed borrow APY by a required premium heuristic. Secondary markets plot the policy rate curve when on chain parameters resolve. The AMM rate is backed out from today's borrow APY. If those inputs are incomplete, the card falls back to a headroom callout.
Soft liquidation
Mechanism vs priority
High soft liquidation converted share means collateral price has moved through soft liquidation bands. That is designed behavior. Priority or elevated language is reserved for conversion paired with a second signal (thin free liquidity, ceiling pressure, or oracle/AMM gap). Soft liquidation alone is not treated as a priority breach.
Downside estimates
CVaR, structural PD, portfolio Vasicek
Collateral CVaR uses daily price history (historical simulation, loss-positive clipped). Mint structural PD uses Prices API collateral. LlamaLend structural PD prefers on-chain borrower collateral mark-to-market (controller user_state sum) with a disclosed floor fallback. Drift μ sensitivity is on the card. Portfolio Vasicek prefers horizon-scaled hazard PD when a liquidation event model fits. Util→PD is only the interim fallback. ASRF ρ is a collateral return co-movement proxy (floored/capped for the control), not fitted default correlation.
Reduced form hazard model status is Logistic on chain liquidation logs. Shown only when enough events exist in the lookback.
| Parameter | Status | Basis | Close out |
|---|---|---|---|
| Collateral historical CVaR | monitoring estimate | Nonparametric historical simulation on daily log returns (loss-positive clipped) | From price history label at ≥730d daily observations |
| GARCH(1,1) t CVaR | theoretically motivated | GARCH(1,1) grid MLE with Student t innovations (ν=8). Unit tests cover fit constraints and normal-CVaR diagnostic. | Shown alongside historical CVaR on market detail. Multi-day GARCH still omitted. |
| √h horizon scaling | caution | Standard i.i.d. convention | Use overlapping h day returns or show a caveat. Overlapping windows are implemented. |
| Mint structural PD (Merton d2) | theoretically motivated | amm A, loan discount, and liquidation discount from the Prices API. Drift μ sensitivity shown on card. | Validate against liquidation event panel when n grows |
| LlamaLend structural PD | theoretically motivated | On-chain borrower collateral MtM (user_state sum) when RPC resolves; else LLAMMA + loan-discount floor | Shipped on market detail. Prefer full MtM path; floor is fallback only. |
| Merton drift μ | monitoring estimate | Default zero. Sensitivity at 0%, 3%, 5% annualized on the structural PD card. | Sensitivity displayed. Keep μ=0 as monitoring base until a funding-rate prior is agreed. |
| Reduced form hazard β | monitoring estimate | Logistic on util, soft liq, oracle gap, gas stress. Horizon mapped with 1−(1−p)^(H/L). | Fit when ≥2 liquidation events in lookback. Optional Dune controllers merge into labels. |
| Vasicek ASRF ρ (return co-movement proxy) | monitoring estimate | Collateral return correlation + stress conditional ETH down day subset. Floored/capped for the control only. | UI labels return co-movement separately from recommended ASRF ρ. Not default correlation. |
| crvUSD peg deviation CVaR | monitoring estimate | Historical simulation on crvUSD USD price | Same pipeline as collateral CVaR |
| LlamaLend PD proxy from utilization (interim fallback) | monitoring estimate | Monotone map from utilization only when no hazard model fits (<2 events) | Superseded by horizon-scaled hazard PD for all markets once the model fits |
| Soft liquidation stress exponent | monitoring estimate | Convex intensity (drop/100)^0.85 on residual LLAMMA collateral (progressive band fill proxy) | Keep disclosed as unfitted until a conversion panel exists |
Binding dimension
Live data only
The binding (worst) L1 to L6 code uses only dimensions with status live. Partial fallbacks do not compete for the binding headline.