> ## Documentation Index
> Fetch the complete documentation index at: https://docs.alterscope.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Pool Risk Scores

> Independent lending-pool risk scores, each computed with a published peer methodology and surfaced side by side without blending.

There is no single right way to score a lending pool's risk — a bad-debt accountant, a simulation shop, and a machine-learning team will each answer the question differently, and each answer is informative. Rather than collapse them into one opaque number, Alterscope computes **several independent scores using established, published methodologies and presents them side by side**, each with its own units and provenance, so you can see where they agree and where they diverge.

## How scores are surfaced

The pool risk-scores endpoint returns the **latest score from each methodology** for a given pool. Critically:

* **Each score is presented as-is — no blending, no averaging, no normalization.** Different methodologies answer different questions in different units (a dollar figure, a probability, a percentage), so collapsing them would destroy information. You get each firm's number with its own `score_unit`, the block and time it was computed `as_of`, a confidence, the input summary, and a link to the published methodology.
* **Stale scores are withheld, not shown.** Each score is checked against a freshness window (seven days). If a score is too old to trust, the endpoint returns no score for that methodology rather than a stale number — the same freshness-first discipline applied everywhere on the platform.

## The methodologies

Each score is **Alterscope's own implementation, built to follow the firm's publicly documented methodology.** These are independent reconstructions from public research — Alterscope does not call these firms' APIs or resell their proprietary outputs. Each is labeled with the methodology and version it implements.

<CardGroup cols={2}>
  <Card title="RiskDAO-style bad debt">
    Estimates the pool's bad debt — the shortfall where a borrower's debt exceeds the liquidation-adjusted value of their collateral — summed across underwater positions. **Unit: USD.** Follows the RiskDAO bad-debt methodology.
  </Card>

  <Card title="Gauntlet-style agent simulation">
    A Monte-Carlo agent-based simulation that runs many price paths and measures the fraction in which positions cross into insolvency. **Unit: probability of insolvency.** Follows Gauntlet's published risk methodology.
  </Card>

  <Card title="Block Analitica-style vault liquidation">
    A liquidation-probability model that estimates the probability a position is liquidated within 24 hours, aggregated to a pool-level score. **Unit: probability.** Follows Block Analitica's vault-liquidation methodology.
  </Card>
</CardGroup>

Each methodology may run multiple versions in parallel during transition windows; the response always names the exact `methodology_version` behind a number so results are reproducible and comparable over time.

## How to read divergence

Because the scores measure different things, **divergence is signal, not noise**:

* A high RiskDAO-style bad-debt figure tells you losses may *already* be embedded in the pool.
* A high Gauntlet-style insolvency probability tells you the pool is *vulnerable under stress* even if healthy now.
* A high Block Analitica-style liquidation probability tells you *near-term churn* is likely.

Reading them together — rather than as one blended score — is the point. Each links to its source methodology so you can judge the approach yourself.

## What we publish vs. withhold

| Published                                                     | Withheld                                                        |
| ------------------------------------------------------------- | --------------------------------------------------------------- |
| Which methodologies are implemented, and their public source  | Exact internal parameters of each reconstruction                |
| The unit, version, freshness window, and inputs of each score | Internal model parameters (for the Block Analitica-style score) |
| That scores are surfaced independently with no blending       | —                                                               |

## Where this shows up

Pool risk scores are returned by the pool risk-score endpoint in the [API reference](/api-reference/overview), with [freshness and quality metadata](/trust/data/freshness-and-quality) on every response. Coverage varies by pool and methodology — which pools currently carry which scores is documented under [Coverage & gaps](/trust/data/coverage-and-gaps). The boundaries of these reconstructions are on the [Limitations](/trust/methodology/limitations) page.