Chain risk
Rates the chain a position lives on: if the chain fails or censors you, no property of the lending market can protect the deposit. Scored from L2Beat's published risk criteria, keyed by chain id.
The five criteria
Each is scored 1-5 independently.
State validation — how the chain proves its state transitions are correct.
| Mechanism | Score |
|---|---|
| Validity proofs, no execution delay | 1 |
| Validity proofs with a delay, or multi-proof systems | 2 |
| Fraud proofs | 3 |
| TEE-based | 4 |
| None | 5 |
Data availability — whether the data needed to reconstruct state is posted on Ethereum, to an external DA layer, to a committee, or self-custodied.
Exit window — how long users have to withdraw before an upgrade takes effect. No window is the worst case; a long or infinite window is the best.
Sequencer failure — what a user can do if the sequencer stops or censors: force transactions through L1, self-sequence, or nothing.
Proposer failure — what a user can do if the proposer stops: self-propose, use an escape hatch, or cannot withdraw at all.
Composition
risk_score = round(mean(applicable criteria)), clamped to 1-5
Criteria the source does not report are dropped from the mean rather than guessed. Ethereum mainnet is pinned to 1 across all five as the reference point.
Both the overall score and the per-criterion breakdown are exposed, so a consumer
can filter on the specific property they care about — a protocol that only cares
about censorship resistance can read sequencerFailure directly instead of the
average.
What this does and does not cover
Rollup risk is well specified and comparable, which is why these five criteria are used. It says nothing about the operational risk of a chain's bridges, validator set economics on non-rollup chains, or RPC and infrastructure centralisation. A chain L2Beat does not cover has no score rather than a flattering default.