The BRI Formula
The BlackHart Risk Index uses a multiplicative formula so that catastrophic weakness in any single dimension cannot be masked by high scores elsewhere. Every dimension matters — there is no way to “average away” a critical gap.
BRI = 300 + 700 × Π(Dᵢ / 100)^wᵢ
where Di = dimension score (0-100), wi= dimension weight, Σwi = 1.0
Why multiplicative?
An additive formula (weighted average) allows a catastrophic weakness in one area to be hidden by high scores elsewhere. The multiplicative formula means a single dimension near zero drags the entire BRI down, regardless of other scores.
Why 300-1000?
The floor of 300 prevents mathematical artifacts. The 700-point scale maps cleanly to 7 Forge Scale tiers of ~100 points each.
Exponent structure:
Each dimension contributes as (Di/100)wi. The weight controls sensitivity — a 18% weighted dimension has more leverage than a 4% one.
How the BRI is Computed
Extract
We pull the protocol's deployed bytecode and source, decompose it into a contract interaction graph, and map every entry point, state variable, and external dependency.
Enrich
The graph passes through 156 analysis modules — from classical static analysis to topological homology, spectral methods, and adversarial generation — producing 200+ feature dimensions per contract.
Score
Features are aggregated into 12 BRI dimensions using calibrated weights. Each dimension is independently scored 0-100, then combined into the final BRI via the multiplicative formula.
Validate
Human researchers review edge cases, verify adversarial generation results, and confirm that the automated score matches reality. Scores are published only after validation.
Publish
The validated score will be pushed to the BROOracle contract on Base. Evidence will be pinned to IPFS. Once live, scores become queryable on-chain within the same transaction.
Weight Governance
Empirical basis
All dimension weights are derived from regression analysis of 200+ historical DeFi exploits. Weights reflect observed correlation between dimension weakness and exploit occurrence.
Quarterly recalibration
New exploit data is incorporated quarterly. Weight adjustments greater than 2% require published rationale and evidence of improved predictive accuracy.
Version pinning
Every published score records the weight version used to compute it. Scores are never retroactively recalculated — historical comparisons remain valid.
Advisory input
Weight change proposals are reviewed by the advisory board. All proposals require empirical evidence demonstrating improved correlation with real-world exploit outcomes.