What (One-liner)

Canvas Layer 4 top node: regime-driven cross-asset allocation (equity/bond/commodity/cash) + vessel tilts + dynamic risk budget. 5-regime × 4-asset matrix + ±15% tilt cap.

How to use

macro_factor_v1 + strategy_vessel_v1 + institution_composite → meta_strategy
    → weights {equity: 0.58, bond: 0.23, commodity: 0.12, cash: 0.07}
    → risk_budget: 0.95
    → rationale: "expansion regime + 2 bullish vessels; institution IC confirmed"

Core formulas

See frontmatter. weights = normalize(META_REGIME_WEIGHTS[regime] + vessel_tilts + institution_overrides).

Assumptions & applicability

Assumptions: upstream regime derivable + Pangura META_REGIME_WEIGHTS calibration + US business-cycle semantics.

Applies to: cross-asset allocation top layer, meta-scenario analysis, Applications exit.

Does not apply to: numerical optimization, single asset, minute-level.

Known limitations

See frontmatter. Six items: regime weights hard-coded / 15% tilt cap / 4-asset fixed / not CVaR / US-only calibration / no rebalancing logic.

References

• Black-Litterman (1992) FAJ — cross-asset allocation
• Faber (2007) — Tactical allocation
• Pangura META_REGIME_WEIGHTS (2026-03)

Golden Test

Planned — Tier 2 TS golden in backlog (lock matrix + tilt formula + risk multiplier).

Changelog

• 1.0.0 (2026-04-21) — First Active

Strategy identity & interface (C-n v2, 2026-07-30 — absorbed strategy_measures_v1)

Since 2026-07-30 meta_strategy is also the strategy vessel of a graph: it carries the strategy's identity (name / asset domains) and its published readings interface — the numeric measures the rest of Axe reads from this graph. strategy_measures_v1 was retired into this node (already-registered legacy graphs keep executing; the palette no longer offers it).

Interface resolution (deterministic; Tier-1 golden since 2026-08-03):

• explicit measures definitions win when present (invalid entries dropped);
• otherwise built from the config form: measure presets (sharpe, max_drawdown, expected_return, total_return, max_weight) plus custom dot-paths (one per line or |-separated), registered as custom:<path> with a shared risk direction;
• at read time the node scans upstream results (BFS ≤4 hops) in order and takes the first finite number at each path; the observation date travels with it.

One-click monitoring: "Add to Monitor" on this node registers the whole graph as a monitored strategy — a frozen snapshot re-runs nightly, extracting exactly this interface's readings (same extraction code server-side, kept in lockstep by parity tests), feeding the Monitor board, alerts and the SIMULATED NAV.

Provenance: interface layer golden — tests/golden/fixtures/tier1/meta_strategy_interface/ (7 cases, deep-equality vs an independent Python reference).