Deterministic allocation methodology
AAA is not a black box. Every allocation output is the result of a defined pipeline: portfolio state in, policy constraints applied, decision record out. This page documents the methodology — what it does, what it assumes, and what it explicitly does not do.
Design principles
Determinism before discretion
The same inputs — portfolio state, policy constraints, regime context — always produce the same allocation output. Determinism is the foundation of legibility: decisions can be audited, replayed, and explained without requiring faith in a model.
Policy as executable doctrine
Allocation policy is not a guideline or a suggestion. In AAA, policy is a set of machine-checkable constraints: asset eligibility rules, concentration caps, liquidity requirements, rebalance bounds, and regime-specific behavior. The allocator resolves outputs inside that constraint set.
Regime awareness without abandoning rules
Markets are not stationary. AAA models allocation as regime-aware: conservative postures in drawdown conditions, neutral in stability, calibrated expansion when conditions permit. Regime context adjusts constraint sensitivity — it doesn't override determinism.
Separation of decision and execution
AAA produces decisions. Execution is always separate. This is not an implementation choice — it is a design principle. Mixing decision intelligence with execution authority creates accountability problems that no audit trail can fix.
Governance before automation
Autonomous operation is the last tier, not the default. AAA is structured so that governance, authority qualification, and policy validation precede any form of automated execution. The allocator can be strong without being dangerous.
The allocation pipeline
Every allocation decision in AAA follows this pipeline. There are no shortcuts, no bypasses, and no undocumented steps.
- 1Portfolio state ingestion
Wallet balances are imported from connected chains and normalized into a policy-compatible portfolio model. Each asset is mapped to a risk class and role classification used in constraint evaluation.
- 2Policy and regime configuration
The operator defines or selects a versioned policy set: eligibility rules, exposure constraints, liquidity floors, concentration limits. A market regime is declared — conservative, neutral, or expansion — which adjusts constraint sensitivity.
- 3Constraint evaluation
The allocator evaluates the current portfolio state against all active policy constraints. Violations are flagged. Permitted allocation space is computed. All constraint evaluations are logged.
- 4Allocation resolution
Within the permitted space defined by policy, the allocator resolves a target allocation. Multiple allocator versions can be run and compared before a target is selected. Churn, sensitivity, and regime behavior are all measurable before commitment.
- 5Decision record emission
The output is not just an allocation — it is a decision record: a machine-readable document containing the inputs evaluated, constraints applied, allocator version used, and the resolved output. This record is the basis for review, governance reporting, and audit.
- 6Approval and execution routing
The decision record routes through the approval tier defined by the operator's authority level. At Production and Doctrine tiers, review gates are enforced before outputs proceed to downstream execution infrastructure.
What the methodology explicitly excludes
- Execution. AAA does not initiate, sign, or broadcast transactions. All execution is handled externally.
- Custody. AAA does not hold private keys or request signing permissions. Portfolio import is read-only.
- Signal aggregation. AAA does not aggregate on-chain signals, social sentiment, or external data feeds for alpha. Policy constraints are the signal.
- Discretionary overrides. The allocator does not support manual output overrides that bypass constraint evaluation. If policy needs to change, policy is updated through a versioned update — not through an undocumented runtime override.