Wallet Risk Scoring
Probability-based scoring with drivers: counterparty risk, abnormal flows, cluster proximity, timing anomalies, and more.
What ChainSignal is
Minimal claims, measurable outputs: signals, risk drivers, model versions, and an auditable trail.
On-Chain Performance
Track signals against real chain data: outcomes, confidence bands, and model-version deltas across time.
Open Training Network
Community nodes index, train, and verify. Outputs are versioned and reproducible, not “trust me, bro”.
Reward system
Contribution-based incentives: rewards follow measurable work, not popularity.
Compute & uptime
Trainer throughput, successful epochs, uptime windows, and reproducible artifacts.
Data quality
Clean indexing, low reorg errors, verified dataset chunks, and integrity checks.
Verification
Independent verification runs, cross-checks, and attestations for model versions and outputs.
Node rewards in beta
Token economics and exact reward weights iterate during beta. Target: anti-sybil contribution scoring with transparent accounting.
Team
Lean and execution-focused. Replace placeholders with real names/links.
Founder
Product · Strategy
Vision, roadmap, and accountable delivery.
ML / Research
Models · Evaluation
Scoring logic, training design, and performance measurement.
Infrastructure
Nodes · Data · Security
Indexing, pipelines, reproducibility, and operational hardening.
Community Nodes
Compute as public infrastructure: data ingestion, training, verification.
Node roles
Indexer (collect + normalize), Trainer (train + export artifacts), Verifier (re-run + attest). Roles can be combined.
Security posture
Deterministic data steps where possible, strict versioning, cross-checks, and audit-ready artifacts.
Beta is live
Early access to signals, API endpoints, and model versioning. Outputs are probabilistic and depend on chain context and model version.
FAQ
Short and technical.
A modeled probability under defined criteria (e.g., scam patterns, counterparty risk, abnormal flows). Each score carries drivers, confidence, model version, and a traceable hash.
Multi-stage validation, strict versioning, cross-checks between node roles, and reproducible evaluation runs. Suspicious drift is detectable through model/version deltas.
No. ChainSignal provides data-driven signals and risk indicators. Outputs can be wrong, especially with thin data or adversarial behavior.