KB / indicators

Health Score (19-component composite)

Last verified 2026-05-20

Composite 0–100 measure of structural market health — not direction. A high score means the plumbing is intact (liquidity flowing, credit calm, institutions buying). Normal markets score 55–60.

Components (19, sum to 100)

Component	Max	What it measures	Metric
Energy	10	Energy regime (WTI, crack spread, XLE)	`energy_regime`
HY OAS	8	High-yield credit spreads	`hy_oas`
Stagflation	8	Weighted breakevens (5Y 40%, 10Y 20%, 5Y5Y 20%, 5Y-10Y inversion 20%)	`stagflation`
DIX	7	Institutional dark-pool buying activity	`dix`
GEX	7	Options dealer gamma exposure	`gex`
Breadth	7	% of S&P 500 above 50-day SMA	`breadth_50d`
Liquidity	7	Fed net liquidity direction	`net_liquidity`
Breakeven	6	5Y breakeven inflation expectations (lower is better)	`breakeven_5y`
Vol Term	5	VIX term structure (contango/backwardation)	`vix_regime`
SPY/ZGL	5	Price vs zero gamma level	`spy_vs_zgl`
0DTE PCR	5	Same-day options put/call ratio	`zero_dte_pcr`
Sectors	5	Sector breadth favorability	`sectors`
Correlations	5	Weighted per-pair health (SPY/VIX, SPY/Oil, SPY/DXY, SPY/TNX)	`correlations`
BTC	3	Bitcoin as liquidity proxy	`btc`
SKEW/VIX	3	Tail-risk hedging activity	`skew_vix`
Real Yield	3	10Y TIPS real yield level	`real_yield`
Copper/Gold	2	Growth expectation proxy ratio	`copper_gold`
USD/JPY	2	Carry trade unwind risk	`usdjpy`
HYG	2	Credit ETF selling pressure confirmation	`hyg_confirmation`

Pure additive sum. Each band-classifier maps the raw value to a per-component points award between 0 and the listed max. Logic lives in app/signals/score.py — pure, FastAPI-free, directly unit-testable.

Score ranges

Score-range readings

Favorable 70+ Conditions support risk-taking. Most supports intact.

Mixed 50 – 70 Some supports intact, others not. Be selective.

Cautionary 30 – 50 Multiple supports weakening. Reduce risk.

Adverse < 30 Structural supports broken. Defensive only.

The 50–60 band is where most days live.

Regime-conditional rebalancing

score_weights.json → regime_overrides lets any regime declare scaling factors per component. Currently configured:

energy_shock (energy_regime in SHOCK/CRISIS/SHOCK_UP/RISING): Correlations doubles 5→10 (cross-asset transmission dominates), BTC drops to 0, Sectors scales 0.6× (5→3). Total stays at 100.
energy_relief (FALLING/SHOCK_DOWN): Correlations 0.6× (5→3), Breadth boosted 1.43× (7→10).

Adding an override: {trigger: {field, in|eq}, scales: {component: factor}}. Factors must sum-offset so the post-override max stays at 100.

Score vs regime — when they disagree

The dashboard’s bridge text (italicized line under the regime name) labels the divergence:

Score	Regime	Label	Meaning
70+	RISK-ON	Favorable	Score and regime agree
70+	Not RISK-ON	Favorable, with caution	Score high but a lagging signal holds the regime back
50-70	Any	Mixed	Some supports intact, others not
50-70	RISK-OFF/PANIC	Mixed, deteriorating	Mid-range score but conditions worsening
30-50	Any	Cautionary	Multiple supports weakening
< 30	PANIC/RISK-OFF	Adverse	Structural supports broken
< 30	Other	Adverse, holding	Score very low but regime hasn’t caught up yet

Scale headroom

A May 2026 regime-quantile audit confirmed the 0-100 scale spans cleanly without saturation:

Regime	n	p5	p50	p95	max
PANIC	18	23	28	39	40
RISK-OFF	289	28	37	49	57
TRANSITIONAL	308	20	47	55	77
CAUTIOUS	467	42	50	58	74
RISK-ON	205	48	55	82	87