Wallet Graph Module (30%)
Counterparty concentration, wash patterns, insider propagation, and sybil pressure.
Credibility Anchor
How the MESS scoring stack works.
We combine behavioral, structural, and market signals into a versioned scoring system designed to be auditable.
MESS-RPE v0.3
Bi-weekly
Known Rugs + Stable Protocols
Top Signal Attribution
Score breakdown
Composite score from 0-100 across five modules.
Liquidity Stability Module (25%)
Depth resilience, unlock pressure, and volatility stress projections.
Contract Semantics Module (20%)
Privilege paths, pause-mint controls, proxy risks, and hidden behavioral flags.
Tokenomic Simulation Module (15%)
Monte Carlo vesting shock scenarios and treasury concentration risks.
Behavioral Anomaly Module (10%)
Unsupervised detection against historical exploit and rug trajectories.
Versioning and benchmarks
Versioned Releases
Every model update ships with changelog, data window, and validation deltas.
Benchmark Packs
Performance is measured across balanced cohorts: known exploits, false-alarm controls, and live protocols.
Calibration Audits
Quarterly drift checks for precision, recall, and confidence calibration by score tier.
Public Metrics
Precision, false positive rate, and top signal features are exposed in dashboard telemetry.
Public transparency metrics
Precision Rate
Measured accuracy of positive risk classifications.
False Positive Rate
Frequency of risk flags that do not materialize.
Historical Detection Accuracy
Backtested performance across labeled historical events.
Model Version Tracking
Public model versioning with changelog and evaluation deltas.
This methodology is evolving. Stable releases will publish final production weights, thresholds, and validation deltas.