Skip to main content

Risk Scoring Methodology (Deep Dive)

TODO: Complete Mathematical & Technical Risk Model Documentation​

Comprehensive specification of how risk scores are computed.

Required Sections​

  • Mathematical Specification

    • Formal definition of risk score function
    • Component calculations (formulas for each dimension)
    • Weight justification
    • Bounded on [0, 100] — normalization details

  • Component Definitions

    • Slashing History

      • How slashing events are detected & recorded
      • Time windows & decay functions
      • Severity classification (minor vs major)
      • Edge cases (partial vs full slashing)

    • Uptime

      • Beacon chain attestation rate calculation
      • Missed proposal detection
      • Downtime incident definition
      • Minimum observation period requirements

    • AVS Concentration

      • How operator stake allocation across AVSs is tracked
      • Concentration metric (e.g., Herfindahl index?)
      • Correlation detection between AVSs
      • Threshold definitions ("over-concentrated" = ?)

    • Reputation

      • Historical performance tracking
      • Governance participation scoring
      • Cross-protocol data integration (if any)
      • Tier definitions

  • Update Triggers & Frequency

    • When risk score recalculates
    • Minimum change threshold (to avoid noise)
    • Manual override procedures (if any)
    • Version history & deprecation

  • Assumptions & Constraints

    • Stationary / non-stationary market behavior
    • Correlation assumptions between operators
    • Data availability assumptions
    • Time horizon assumptions (what is "recent"?)

  • Sensitivity Analysis

    • How sensitive is risk score to component changes?
    • Weight sensitivity
    • Time decay sensitivity
    • Threshold sensitivity

  • Validation & Backtesting

    • How is accuracy measured?
    • Precision vs recall tradeoffs
    • Backtesting methodology
    • True positive rate (did high-risk operators actually slash?)

  • Future Extensions

    • Planned risk dimensions
    • Machine learning integration (if planned)
    • Market regime detection
    • Operator-specific customization

Examples​

  • Calculate risk score for example operator
  • Risk score sensitivity to uptime change
  • How a slashing event affects score

Status: NOT STARTED — Requires detailed risk model specification from team