> For the complete documentation index, see [llms.txt](https://maxim-speransky.gitbook.io/doc/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://maxim-speransky.gitbook.io/doc/gitbook-risk-framework/gitbook-market-scoring.md).

# Market Scoring

Market Scoring evaluates whether a specific lending-market configuration can price collateral, liquidate debt, and return value to lenders under relevant stress. It begins only after the collateral and loan assets have been identified and assessed.

This public page describes the decision architecture. Detailed component anchors, scenario thresholds, stress-model parameters, and operating tests remain part of the internal methodology.

## Outputs

```
Eligibility: Eligible / Excluded / Unassessable
Market Score: 0–100 if Eligible; otherwise N/A
Market band: derived from Allocation Policy if a score exists
Subordination Class: None / Limited / Material / Strong if Eligible
Confidence: High / Medium / Low if a score exists
```

* **Eligibility** determines whether the market may proceed to scoring and downstream policy.
* **Market Score** describes the weakest market-specific loss vector on an ordinal scale.
* **Market band** is assigned by [Allocation Policy](/doc/gitbook-risk-framework/gitbook-allocation-policy.md) rather than by this scorecard.
* **Subordination Class** describes loss absorption created by the deployed collateralization.
* **Confidence** expresses the strength of the evidence supporting the assessment.

These outputs remain separate. Subordination is not added to Asset Score, and strong evidence does not improve weak market mechanics. `Excluded` and `Unassessable` markets receive neither a Market Score nor a Subordination Class.

## What is assessed

The unit of analysis is the complete market configuration required for pricing, liquidation, lender exit, and settlement:

```
Asset pair → oracle path → LLTV and market parameters
           → liquidation routes → market state → lender exit
```

The assessment identifies the deployment, chain, market identifier, collateral and loan assets, oracle composition, current parameters, liquidation venues, utilization, concentration, and settlement path.

Both sides of the pair matter. A loan asset is not an invisible numeraire: its depeg, freeze, redemption failure, bridge dependency, or transfer restriction can impair lenders even if collateral liquidation succeeds.

## Risk routing

The same fact is assessed once:

* [Asset Scoring](/doc/gitbook-risk-framework/gitbook-asset-scoring.md) owns the intrinsic backing, control, custody, and redemption risks of the collateral and loan assets.
* Market Scoring owns the concrete oracle path, deployed parameters, executable liquidation capacity, utilization, position distribution, lender exit, and settlement consequences.
* Shared Morpho contracts and protocol controls are treated as a framework scope assumption and are not re-scored for each market.

An asset-level feed may support the economic claim, but Market Scoring still tests its suitability for liquidation at the required speed and size.

## Pair-stress profile

Before eligibility or scoring, the analysis defines how relative price movement can create lender loss. Descriptive pair labels do not determine the score.

The profile considers:

* collateral value in loan-asset terms;
* oracle delay and basis to executable liquidation value;
* jump, depeg, maturity, queue, bridge, and market-closure scenarios;
* contraction of available depth during stress;
* correlation between collateral impairment, oracle error, and liquidity loss;
* whether liquidation must clear atomically or can be bounded through progressive execution.

[Liquidation Resilience Assessment](/doc/gitbook-risk-framework/gitbook-market-scoring/gitbook-slcr-methodology.md) and related stress tests may inform Liquidation Capacity when the selected model represents the actual liquidation process. Its `dynamic_slcr` output is coverage headroom rather than the name of the complete assessment. Diagnostic results or unvalidated assumptions cannot be promoted into decision-grade evidence. Detailed profiles, parameters, and operating thresholds remain internal.

## Eligibility before scoring

Eligibility is tested before numeric scoring. The assessment stops when a critical mechanism failure or evidence gap prevents a bounded market-loss analysis.

Typical `Unassessable` outcomes include:

* the current market identity, asset pair, configuration, or critical dependency cannot be established;
* a critical oracle leg, liquidation route, position distribution, or lender-exit state cannot be verified;
* required live evidence is too stale to bound the current risk;
* the available stress model does not represent the market's liquidation process.

Typical `Excluded` outcomes include:

* the oracle can materially diverge from realizable pair value without a bounded recovery path;
* current parameters produce a known negative liquidation margin under an ordinary adverse move;
* no executable route exists from seized collateral to the loan asset;
* a mandatory atomic tranche cannot clear before the next safe state;
* liquidations, withdrawals, or settlement are non-functional for reasons specific to the market.

A mandatory atomic tranche is the protocol-required batch, not the whole stressed market exposure. When the market requires progressive clearance and that path cannot be bounded with decision-grade evidence, Stressed Coverage is `Unassessable`. A diagnostic result cannot support a numeric anchor or be made acceptable by reporting lower Confidence.

A limitation is not automatically a knockout when the resulting loss path remains bounded and assessable. The internal methodology defines the complete tests.

## Scoring model

Only an `Eligible` market receives a Market Score. Four non-compensable domains are assessed:

| Domain                            | Assessment focus                                                                                                   |
| --------------------------------- | ------------------------------------------------------------------------------------------------------------------ |
| **Oracle Integrity**              | Whether the full feed path provides a timely and manipulation-resistant representation of realizable pair value    |
| **Liquidation Capacity**          | Whether liquidators can repay debt and dispose of collateral under stressed volume and shared-liquidity conditions |
| **Parameter Fit**                 | Whether the deployed LLTV, liquidation incentive, close factor, caps, and fees fit the pair and its dependencies   |
| **Lender Exit & Market Function** | Whether suppliers can exit or settle without relying on a concentrated or non-functional path                      |

```
Market Score = MIN(
  Oracle Integrity,
  Liquidation Capacity,
  Parameter Fit,
  Lender Exit & Market Function
)
```

No weighted average is used. A lending market is a dependency chain: deep liquidity cannot compensate for a wrong oracle, and conservative parameters cannot compensate for a failed settlement path.

Oracle Integrity, Liquidation Capacity, and Lender Exit & Market Function each take the weakest component result within that domain. Parameter Fit is a direct scenario-led assessment.

### Oracle Integrity

This domain examines whether the complete feed composition represents realizable collateral value in loan-asset terms, updates faster than the relevant loss path, and fails conservatively.

### Liquidation Capacity

This domain takes the weakest result across Stressed Coverage, Absolute Clearance, Execution Access, and Shared Capacity & Route Dependence. It consumes [Liquidation Resilience Assessment](/doc/gitbook-risk-framework/gitbook-market-scoring/gitbook-slcr-methodology.md) through two distinct tiers:

* **Atomic Tranche Clearance** tests only the maximum protocol-required batch before the next safe state. It supports Absolute Clearance together with vault-attributable stressed-loss tests.
* **Dynamic Clearance Adequacy** tests the complete liquidation-demand path against executable service, backlog and loss. It supports Stressed Coverage.

The dynamic test declares a common `t0`, fixes `H_dynamic` before scaling demand, and recomputes state-dependent arrivals, route service, feedback and loss for every demand scale `alpha`. `dynamic_slcr` is the maximum scale for which `T_clear(alpha) <= H_dynamic` and market bad debt remains zero throughout the path. Terminal aggregate volume is insufficient evidence of clearance.

Tier 1, Tier 2, market loss and vault attribution each retain their own result status. Every materially required output must be `decision-grade`; a validated atomic result cannot upgrade a provisional loss model or an unbounded attribution. Positions below the economic execution threshold are loss candidates, not automatically bad debt; market dust is not compared directly with vault TVL.

In LRA v2.2, progressive Tier 2 implementations remain `diagnostic-only`. An atomic-only reduction is possible only when the complete stress path clears inside `H_atomic`. Quoted depth, asset fungibility or historical survival cannot establish decision-grade progressive coverage.

### Parameter Fit

This domain evaluates the deployed configuration rather than the authority that selected it. The same LLTV can produce different risk depending on oracle delay, liquidation incentive, pair behavior, maturity, depth, caps, and borrower concentration. No parameter receives automatic points outside its market context.

### Lender Exit & Market Function

This domain examines available withdrawal liquidity, utilization and IRM response, supplier and borrower concentration, and loan-asset settlement. Ordinary high utilization is distinguished from a market-specific failure, but a measurable repayment path is required for lender exit to remain assessable.

The `MIN` rule and score anchors are expert policy choices. Market Score is not a probability of default, expected loss estimate, or empirically calibrated capital requirement.

## Subordination Class

Subordination Class describes the incremental lender protection created by collateralization after accounting for adverse price movement, oracle basis, incentives and fees, execution loss, competing flow, and model uncertainty.

* **None** — no reliable incremental loss absorption is established.
* **Limited** — protection exists, but severe-stress headroom is thin or dependent on a material assumption.
* **Material** — positive protection remains under the relevant severe scenario and critical dependencies are bounded.
* **Strong** — substantial protection remains under severe stress and adverse sensitivity without relying on one correlated assumption.

Subordination Class is ordinal, not additive. It never changes the intrinsic Asset Score, rescues an excluded asset or market, or independently authorizes an allocation. [Allocation Policy](/doc/gitbook-risk-framework/gitbook-allocation-policy.md) determines whether and how the class affects the final position decision.

Dynamic clearance contributes only through the stressed-loss waterfall. It cannot also map directly to a Subordination Class or create a second automatic improvement from the same protection. `Low` Asset or Market Confidence prevents Subordination improvement under Allocation Policy.

## Evidence and confidence

Every conclusion must be supported by claim-matched evidence, including deployed contracts, current parameters, on-chain state, executable route measurements, and observed market behavior.

The assessment records an as-of date and the block, query run, deployment, and configuration versions needed to reproduce the observed state. Historical liquidations strengthen Confidence only when they are relevant to the current oracle, parameters, pair behavior, and liquidation engine.

Confidence is reported separately:

* **High** — current primary state and independent measurements cover every critical dependency and scenario;
* **Medium** — critical claims are established, but some evidence is model-dependent, delayed, or only partially independent;
* **Low** — the loss path can be bounded, but material evidence or validation limitations remain.

Overall Confidence is the weakest Confidence among critical domains and dependencies. Missing or contradictory evidence on a critical claim produces `Unassessable`, not Low Confidence.

## Assessment flow

```
Identify the market and asset pair
    → map oracle, parameters, liquidation, exit, and settlement
    → define pair stress
    → test eligibility
    → score eligible markets
    → assign Subordination Class and Confidence
    → pass outputs to Allocation Policy
```

The public assessment presents the result, material drivers, stress assumptions, and evidence basis. It does not reproduce the complete analyst checklist or internal thresholds.

## Policy boundary

Market Scoring does not:

* change the intrinsic Asset Score;
* combine Asset and Market Scores into one number;
* assign a strategy mandate;
* authorize an allocation or set its exposure cap;
* treat other allocators' capital as evidence of safety.

These decisions belong to [Allocation Policy](/doc/gitbook-risk-framework/gitbook-allocation-policy.md). A smaller cap can limit the consequence of an eligible risk, but it cannot make an excluded market acceptable.

## Monitoring and limitations

Market state can decay faster than asset quality. Reassessment is required after material changes to an oracle, parameter, utilization regime, position concentration, liquidation route, available depth, settlement asset, or either underlying Asset assessment.

* The score and Subordination Class depend on expert judgment and scenario selection.
* Normal-market liquidity can disappear during correlated stress.
* Historical survival may not cover the scenario relevant to the current pair, configuration, or position size.
* Stress-model assumptions can dominate the result and must not be presented as calibrated without supporting validation.