Argos RWA Risk Score — Methodology
The transparency reference for the RWA Risk Score shown on the Risk Ratings panel and inline across the Tokenized RWAs and Yield surfaces. It mirrors the scoring engine one-to-one.
What the score is
Every tokenized real-world asset we cover gets a 0–100 score (higher = lower risk) and a descriptive band. The score is a weighted blend of six independently-scored dimensions, each carrying its own data-confidence flag. Dimensions we cannot live-source are omitted — never given a fabricated middle value.
| Band | Composite score | Reading |
|---|---|---|
| Strong | ≥ 78 | Strong risk profile |
| Moderate | 55 – 78 | Moderate risk profile |
| Elevated | 30 – 55 | Elevated risk characteristics |
| High | < 30 | High risk characteristics |
The six dimensions
Each dimension is scored 0–100 (higher = lower risk). Weights sum to 100% and are re-normalized across whichever dimensions a given asset can actually be assessed on.
1. Peg / NAV stability weight 22%
Whether a value-holding asset is actually holding its intended value.
- Applies only to value-holding types (stablecoins, tokenized treasuries, money-market funds). Equities and commodities are meant to move with their underlying, so peg stability is not applicable to them (shown as n/a).
- If the asset is gated
avoidby the trust registry (a live depeg, an unresolved exploit, or a wind-down) → score 5. - Otherwise, if there is a live deviation-from-$1 reading within 3% (a genuine at-peg stablecoin) →
98 − deviation% × 18(high confidence). - Otherwise → the issuer's trust tier is used as a coarse proxy (low confidence). This is a deliberate guard: yield-bearing treasuries such as USDY and USTB drift above $1 by design, so a large "deviation" is treated as accrual, not a depeg. A genuine depeg is caught by the live depeg gate (which sets the asset to
avoid).
Sources: the trust registry + live depeg auto-gate; live price (/api/prices).
2. Liquidity weight 18%
How deep and exitable the on-chain market is.
- Log-scaled over the best live liquidity figure available (USD): roughly $10k → 20, $100k → 40, $1M → 60, $10M → 80, $100M+ → 100.
- Prefers 24-hour trading volume; for a token that trades on DEXes but has no listed volume, falls back to its on-chain DEX pool liquidity (deepest pool depth / volume).
- Omitted (n/a) when no live liquidity figure is available — including primary mint/redeem instruments with little secondary trading, where redeemability (below) is the relevant exit metric instead.
Sources: live 24h volume (/api/prices); on-chain DEX pool liquidity (/api/dex-equities, GeckoTerminal + Jupiter).
3. Adoption / decentralization weight 15%
Whether the holder base is broad (distributed) or concentrated.
- Log-scaled over unique holder count: roughly 100 → 20, 1,000 → 47, 10,000 → 73, 100,000+ → 100.
- Tempered by whale concentration: when the top-10 wallets hold more than half the supply, the score is reduced (down to ×0.6 at 100%) — a large count with a few dominant holders is less decentralized than it looks.
- Omitted (n/a) when we don't have a live holder count for the asset.
Sources: live on-chain holder counts + top-holder concentration across Ethereum and Solana (CoinGecko on-chain), with the Ethplorer Ethereum feed as a fallback.
4. Issuer quality weight 20%
The reputation, collateral model, and track record of the issuer.
| Trust tier | Score |
|---|---|
| Trusted (BlackRock, Ondo, Franklin Templeton, Superstate, Circle, …) | 90 |
| Caution (novel/exotic model, thin record, offshore/unregistered) | 55 |
| Avoid (live depeg, unresolved exploit, wind-down) | 15 |
| Unrated (not yet classified) | 45 |
Deliberately not keyed on size/TVL — a new but legitimate issuer isn't punished for being small.
Source: the curated trust registry (functions/_lib/trust.js), which also absorbs the live depeg auto-gate.
5. Backing quality weight 13%
The safety of what the token is backed by — distinct from the issuer's reputation and from short-term price volatility. Two tokens from the same issuer can carry very different backing risk (a T-bill fund vs. a private-credit pool).
| Backing class | Score |
|---|---|
| Government treasuries / T-bills / money-market funds | 90 |
| Physical commodities (gold, silver, metals) | 80 |
| 1:1 share-backed equities / ETFs / index / pre-IPO | 75 |
| Reserve-backed stablecoins / cash / deposits | 70 |
| Private credit (default risk) | 55 |
| Crypto collateral (ETH/BTC — volatile) | 45 |
Derived from the asset's collateral type; an unknown backing class is omitted (n/a) rather than guessed.
Source: the tokenized-asset catalog collateral type.
6. Redeemability weight 12%
How readily a holder can exit at par through primary redemption — the difference between "redeem any business day" and "you can only sell on a secondary market."
| Redemption model | Score |
|---|---|
| Continuous / T+0 (instant, any time) | 95 |
| T+1 | 82 |
| T+2 | 72 |
| Periodic / monthly / epoch (locked window) | 45 – 55 |
| Secondary only (no primary redemption) | 22 |
From the catalog's structured redemptionModel field; an asset with no documented model is omitted (n/a).
Source: the tokenized-asset catalog redemption terms.
From dimensions to a composite
- Weighted average of only the dimensions that are actually assessable (confidence ≠ "none"), with the weights re-normalized across them. So an equity token (no peg dimension) is scored fairly on the dimensions that do apply — liquidity, adoption, issuer, and backing.
- Minimum coverage: at least 2 assessable dimensions are required. Fewer than that → "Insufficient data" (a null score) rather than a made-up number.
- Impaired-asset cap: any asset gated
avoidis capped at 24 (into the High band) regardless of how deep its liquidity is — an actively-impaired asset is high-risk, full stop, and a healthy-looking liquidity leg can't average that away.
Principles
- Strict-data. Every input is live-sourced or the dimension is omitted. There are no hand-curated or fabricated risk values anywhere in the score.
- Transparency by construction. Each dimension exposes its own confidence; a coarse trust-tier proxy is flagged "low confidence." The whole method is published (this page).
- Assessment, not advice. The score describes risk characteristics. It is never a recommendation. This is Argos's publisher posture — we build the assessment, not a managed product; execution and custody are out of scope by design.
- Tunable + versioned. Dimension weights and thresholds live in one file (
functions/_lib/rwa-risk-score.js) and can be adjusted as the methodology evolves.
Limitations (in the interest of honesty)
- Coverage is bounded to assets with at least two live-sourceable dimensions; others show "Insufficient data."
- Holder-count adoption currently covers the marquee on-chain tokens (Ethereum slice); it broadens as on-chain coverage grows.
- Backing quality and redeemability are scored from the asset's documented collateral class and redemption terms — a structural read, not a live verification that the reserves exist or that a redemption would clear on any given day.
- Off-chain factors a pure on-chain feed cannot independently verify — audit status, real-time reserve attestations, legal/regulatory structure — are not yet scored and are a planned extension. Their absence is why we surface a risk profile, not a credit rating.