Skip to main content

Data Limitations & Blind Spots

TODO: Complete Data Limitations Documentation​

What EigenWatch cannot see and where data comes from.

Required Sections​

  • On-Chain Data Only

    • EigenWatch relies on public blockchain data
    • Off-chain operator behavior is not captured
    • Examples: infrastructure quality, team competence, reputation
    • Implications: risk score reflects only on-chain signals

  • Data Latency

    • Time from event to observation (e.g., slashing takes time to confirm)
    • Time from observation to computation
    • Time from computation to API/oracle publication
    • Total latency: events are not real-time

  • Observer Bias

    • Only observing data from data sources we can access
    • Possible missing data from certain operators/AVSs
    • Survivor bias (operators who didn't slash are "successful" by definition)
    • Correlation bias (confusing correlation with causation in risk factors)

  • Beacon Chain Data Quality

    • Data accuracy depends on Ethereum nodes we query
    • Reorg handling assumptions
    • What if beacon chain has bugs?
    • External finality assumptions

  • EigenLayer Data Gaps

    • AVS contract access limitations
    • Partial information about some operators
    • New operators with no history
    • Staking pool complexities

  • Unknown Unknowns

    • Risk factors we haven't identified
    • Tail events that haven't occurred
    • Novel attack vectors
    • Regime changes in operator behavior

  • Confidence Levels

    • Which risk signals are high-confidence?
    • Which are speculative?
    • How confidence changes over time
    • Confidence per operator (established vs new)

  • Edge Cases & Failure Modes

    • New operators (no history)
    • Operators with tiny stakes (data too small?)
    • New AVSs (no historical slashing data)
    • Protocol upgrades (breaking assumptions)

  • Data Corrections & Restatements

    • When risk scores might be revised
    • How corrections are communicated
    • Handling of prior decisions based on incorrect data
    • Versioning strategy

  • Not a Guarantee of Safety

    • High score doesn't mean slashing will happen
    • Low score doesn't mean safety is guaranteed
    • Multiple operators can slash simultaneously
    • "Black swan" risk is unquantifiable

User Guidance​

  • Don't rely solely on EigenWatch for critical decisions
  • Supplement with off-chain research
  • Diversify across multiple operators despite similar scores
  • Monitor external news & incidents
  • Plan for worst-case scenarios

Status: NOT STARTED — Requires domain expertise & transparency