Four stages, one direction: Data → Signal → Allocation → Selection. Each stage reads only the stage before it — never the reverse. Remove any one and the chain breaks. Here is the whole machine.
The question it answers: is the economy a buyer's market or a seller's market?
EMP scores 36 live macro indicators across six categories — Policy, Ultra-leading, Other-leading, Coincident, Valuation, and watch-only Early-Warning/Recovery feeds. Each indicator is scored 0–10 by either an anchor-based piecewise-linear curve calibrated to real historical endpoints, or a hand-tuned shaped formula where the shaped curve outperformed the linear one (for example, an inverted-U on consumer confidence, because euphoria is bearish). The weighted sum produces a single health score out of 760.
The stress penalty: nine crisis tripwires — CPI above 3.5%, VIX above 22, high-yield spreads above 3.5%, the Sahm rule, curve inversion, and others — dock points when they fire. A fragile-but-still-positive economy cannot score like a healthy one.
What makes it distinctive: regime-aware Fed Funds scoring that penalizes rates differently depending on whether inflation is hot; read-only overlay panels for Gold, Energy, Trough-Redeploy, and a Business-Cycle Direction Gate; a fast-turn V-bottom floor that catches sharp recoveries trailing windows would miss; and band thresholds fit on verified history and validated on a hold-out set.
EMP is the only module that looks outward at the economy. Its band is the single instruction every downstream model obeys. Get the regime wrong and everything below it allocates into the wrong weather — which is why it comes first.
The question it answers: given the regime, how much risk vs. cash — and in what shape?
PAM keeps a hand-built base portfolio for each of the 8 bands, sliding from aggressive (semis, mega-cap tech, growth ETFs) at Full Risk-On to nearly all cash-proxy at Sideline. It reads EMP's band and overlay panels, then applies a sequence of tilts and guardrails to produce the day's invested %, cash-proxy %, and a full ETF-level position map.
The overlays it acts on: a Gold panel that rotates a slice of cash into gold when real yields fall and the dollar weakens; an Energy panel that adds energy exposure on inflationary supply shocks (gated against demand destruction); and a Trough-Redeploy tilt that releases cash into recovery leaders only after the EMP direction gate has confirmed the turn — zero false ascendings in stress tests.
Guardrails always on: a 3% position floor and a 25% cap survive every panel, so no single signal can blow up the book. Gold and energy are funded by reducing cash, never by selling existing risk. Every tilt cites multi-cycle backtest results.
PAM is the bridge between economics and a portfolio. It never picks individual stocks — it sets the shape of the book and the rules of engagement the selection models must respect.
The question it answers: what is every stock in the universe actually doing right now?
For each ticker in the 800+ name universe, BSM pulls roughly 300 days of price and volume history and computes 17 technical indicators — MACD, RSI, ADX, %B, moving-average structure, slopes, 52-week position, ATR extension, volume expansion, and more — producing one scored snapshot of the entire market.
Data integrity is enforced, not assumed. Every field carries a freshness timestamp. Data older than one trading day is flagged STALE; failed fetches are flagged BROKEN — and broken tickers in the active book halt the pipeline rather than letting the models score on bad numbers.
Runs once, shared twice. Both selection models score off the identical snapshot, so differences in their picks come from philosophy, not from different data. BSM is the firm's shared sense of reality: with one trusted snapshot, the models argue only about judgment.
Both selection models import every scorer from one shared core. A stock is scored identically no matter which model looks at it: one definition, zero drift.
"Is it moving with conviction?" MACD, ADX, slopes, 3-month return, volume expansion. This pillar measures entry quality — it finds the strong names.
"Is it underpriced vs. its theme cohort?" Relative strength, 52-week base position, theme-relative return. This pillar times the entry and measures asymmetry.
"Constructive, or already stretched?" Confirmation-aware RSI, %B, MA structure, MFI, ATR extension. The early-warning sell signal — designed to fire before a breakdown shows up in price.
The composite is reported on a /100 scale in every brief, alongside a separate downside-support floor score. M finds strength, V times the entry, and Q protects the downside — the firm's mandate compressed into one number.
BAM treats each theme's allocation weight as a budget and deploys its best-scored names proportionally within it, subject to an investment-grade score threshold and a hard cash-proxy lock. It is sector-aware by construction: a name only competes for the budget of the theme it belongs to. The result is a balanced book that expresses the allocation map faithfully — just with sharper instruments than ETFs.
CAM ignores the sector map and buys the highest-conviction names the regime allows, sizing positions by composite-squared so conviction compounds. Its quality floor is EMP-gated: tighter macro raises the score a name must clear, shrinking the book to fewer, better positions in stress. Fewer names, sized by conviction, scored on the same data.
Running both models off one snapshot gives the research both discipline and edge — and when they disagree on a name, that disagreement is itself information subscribers see every day.
Every trading day, the full run — regime, alerts, scores, deep dives — is written up in plain English and delivered by email.