(internal/legacy) Latest risk metrics snapshot

V3 metric keys as returned inside GET /metrics/{ticker} under the metrics property. Suffixes: _mkt market, _sec sector, _sub subsector, _res residual. Hedge ratios (HR) are dollars of ETF to trade per $1 of stock; explained risk (ER) are variance fractions (0–1). Keys l1_cfr–l3_rr are daily simple returns from returns decomposition:

• *_cfr (combined factor return) is CUMULATIVE through the level: l1_cfr = l1_fr, l2_cfr = l1_fr + l2_fr, l3_cfr = l1_fr + l2_fr + l3_fr.
• *_fr (factor return, INCREMENTAL) is each level's own contribution — l2_fr is what the sector factor adds on top of L1 alone, l3_fr is what the subsector factor adds on top of L1+L2. Use these for stacked-bar decomposition charts or for agent-side attribution ("what drove today's return at each level?").
• *_rr (residual return) is what's left after factors through that level — gross_return ≈ l3_cfr + l3_rr. Keys lstar_rr and lstar_level carry the Lstar-dispatched residual return and the level that was picked, per-(teo, symbol), by ERM3's canonical selector. Use these when you want "the best residual" — the idiosyncratic leg at the same hedge depth as production HR/ER. For the legacy marginal-ER rule at a chosen threshold, hit GET /lstar?threshold=. None of these are ER fields and none come from hedge-weight explainability. Keys l1_mkt_beta, l2_sec_beta, l3_sub_beta are dimensionless hierarchical regression betas (one per level: 1 for L1, 2 for L2, 3 for L3 — though only the L*-introduced new factor is stored at each level). These are NOT the same as hedge ratios — betas are regression coefficients, hedge ratios are dollar-notional ratios.

Canonical L1/L2/L3 hedge snapshots (additive on metrics, decompose, batch success rows). Prefer this block over parsing flat l*_mkt_hr wire keys when comparing cascade levels.