Nine gauges are built from public data and scored onto one comparable scale: each component is transformed, normalized against its own history in a window that ends at the date being scored, oriented for direction, clipped at ±3σ and weighted. Weights renormalize over live components, so a dead feed is redistributed rather than counted as zero.

Forces read directionally. Equilibriums read as deviation from balance, where the band is set by magnitude and the sign says which way. Levers report a stance and, separately, how much ammunition is left — a stimulative stance with no room left is not health.

Lead-lag confidence is measured, not asserted: rules are backtested against point-in-time data — first-release values for any series that gets revised — and discounted by sample size.

What this cannot see: anything outside the collected series, and any month more recent than the “hard data through” date shown above, where values are held forward rather than observed. Scores are model output, not fact. No financial advice.