Skip to main content

Risk Scoring

The Risk Score is a probability of failure—the likelihood that a token will experience a rug pull, technical scam, or other malicious collapse.

The model was trained on real-world data from historical token failures and is directly supported by the behaviors detected by TestMachine's simulation engine.

How it works​

  • Trained on actual rug pull and scam data across EVM networks
  • Uses detected behaviors as primary input signals
  • Incorporates deployer history, liquidity dynamics, and contract structure
  • Outputs a probability score representing the likelihood of failure

What you get​

  • A per-token probability of failure, grounded in real data
  • Explainable scoring backed by the specific behaviors that contribute to the result
  • Continuous re-scoring as contracts evolve and new behaviors are detected

Why it matters​

The Risk Score gives custody teams a data-driven signal for onboarding and monitoring decisions. Because it's trained on actual outcomes and supported by concrete behavior evidence, it moves beyond heuristics into predictive risk assessment.