Each metric is scored using a NBER-calibrated Gaussian KDE approach. For each indicator, we compute a "danger signal" (e.g. Sahm unemployment signal, YoY change, QoQ growth, spread level) at every historical date, then label each observation as recession (NBER USREC = 1) or expansion (USREC = 0). The current signal value is compared against this history using a Gaussian kernel to estimate P(recession | current conditions) — the fraction of historically similar readings that coincided with NBER recession months. This produces calibrated scores: a healthy expansion typically scores 0.05–0.15, while signals consistent with active recessions score 0.50–0.85. Category gauges show the simple average of their weighted metrics. The Summary gauge averages all four categories.

Employment Unemployment Rate uses a Sahm-rule signal (3-month average minus trailing 12-month minimum). Job Openings and Quits use year-over-year percent change. Initial and Continued Claims use the raw level.
National Income & Product Quarterly series (GDP, GDI, Imports, GPDI) use QoQ annualized growth. Tax Receipts use year-over-year change. GNP and GNI are weightless (context only, excluded from the gauge).
Production & Business Industrial Production and Real Sales use YoY change. Inventory/Sales Ratio uses a rolling 36-month z-score to account for secular trends.
Interest Rates Treasury Spreads use the raw spread level — lower/inverted spreads historically co-occur more with recessions.

All data from FRED (Federal Reserve Bank of St. Louis), including NBER recession dates (USREC), cached locally for 6 hours. Inspired by Into The Cryptoverse.