The macro engine inside your terminal. ATLAS reads the macro regime in real time, scores every cross-asset dislocation, and forecasts every Tier-1 print. Six sub-models compute daily from FRED data and refresh on every read.
The economy is always in one of four states defined by the direction of growth and inflation. ATLAS classifies the current state in real time, projects forward 3 and 6 months, and finds the closest historical analogs from 1990 → present. Knowing the regime is half the trade.
Q1
Goldilocks · growth up, inflation down
Risk-on regime. Equities lead, duration neutral, dollar mixed. The friendliest backdrop for long-only allocations.
Q2
Reflation · growth up, inflation up
Pro-cyclical reflation. Commodities, EM, financials lead; long-duration suffers; dollar weakens.
Q3
Stagflation · growth down, inflation up
The hardest regime. Long gold, long volatility, long duration; underweight cyclicals + EM ex-China.
Q4
Deflation · growth down, inflation down
Recession regime. Long Treasuries, long dollar, long quality; short cyclicals + commodities.
02 · The engine
Six models, one regime call
ATLAS isn't a single forecast. It's six sub-models that triangulate the regime from different angles — quad classification, recession probability, liquidity flow, curve dynamics, cross-asset z-scores, and rule-based print nowcasts. Each refreshes daily; the panel composes them.
01 · QUAD CLASSIFIER
Quad Classifier
Level + Δ regime nowcast
Weighted composite of 4 growth series and 4 inflation series from FRED, z-scored versus a 5-year rolling window. Outputs a probability distribution across Q1/Q2/Q3/Q4 plus a forward Markov-chain projection at 3m and 6m.
dataFRED daily
refreshlazy 6h
02 · RECESSION COMPOSITE
Recession Probability
EM + Hamilton smoothed probit
Two academic recession models combined: Estrella-Mishkin (1996) probit using the 10y-3m yield spread, and Chauvet-Piger (2008) Hamilton SMM on coincident indicators. Output is a forward 12-month recession probability with 100% NBER recall in backtest.
recall28/28 NBER
horizon12 months
03 · LIQUIDITY PULSE
Liquidity Pulse
Net liquidity Δ tracker
Tracks net liquidity (WALCL – TGA – RRP) and 90-day flow direction. Liquidity expansion is a leading indicator for risk-on regimes; drains tend to compress multiples 30-60 days ahead. Flagged when Δ crosses ±$200B.
window90 days
lead30-60d
04 · CURVE TRACKER
Curve Tracker
10y-2y inversion path
Monitors the 10y-2y spread in real time, flags inversions, and tracks days inverted. Historically every recession was preceded by an inversion; the un-inversion phase (4-12 months) is the danger window. ATLAS tracks where in the cycle we are.
precedents1947, 1973-74
avg lead~9 months
05 · ANOMALY ENGINE
Anomaly Watch
12 cross-asset z-scores
Twelve cross-asset signals z-scored versus 5-year rolling history: WTI, DXY, copper/oil, breakeven inflation, real yields, BTC/M2, IG/HY spreads, Fed funds vs r*, VIX, 5y/5y forward, 2s10s, gold/oil. Ranked by |σ|; >2σ = HIGH severity.
signals12
tier|σ|>2 = HIGH
06 · PRINT NOWCASTS
Print Nowcasts
Rule-based forecasts · 80% CI
Five Tier-1 economic prints forecast every release: NFP, CPI, Initial Claims, Retail Sales, PPI. Each uses real-time leading indicators (ADP, Mastercard SpendingPulse, etc.) with rule-based composition. Output includes 80% CI from 24-month residual std.
prints5
CI80% bound
03 · How to read it
The terminal panel, top to bottom
ATLAS surfaces 9 distinct outputs in the terminal. Each has a different decision-grade. Glance for the regime headline; sit with the probability map for direction; mine the anomalies for trade ideas; pre-position around prints.
01
Regime Banner
CURRENT REGIME · 12-WEEK PROJECTION
The cinematic top card. Names the modal regime (e.g. Q3 · Stagflation) with confidence percentage and a one-paragraph thesis. The status grid below shows Vol Regime, Liquidity 90d, Curve Invert, Recession 12m with semantic colors.
Glance once per session. The thesis is the editorial frame; everything below is evidence.
02
Regime Probability Map
MODAL · ENTROPY · TRAILING vs FORWARD
The model's actual output: probability across Q1/Q2/Q3/Q4 with Shannon entropy (uncertainty score). When entropy is >1.8 bits, conviction is low. TRANSITION ZONE means trailing-data and forward-nowcast disagree.
High entropy = fade single-regime trades. Transition zone = position for the forward, hedge with the trailing.
03
Quad Plane Chart
CURRENT POINT + 5 HISTORICAL ANALOGS
The current macro state plotted on the growth × inflation plane, with the 5 closest historical analogs overlaid. Distance is L2 in z-score space. Closest analog appears in red; others in gray.
Read the closest analog's forward outcome (3m / 6m / 12m). If the analog is recent and high-confidence, lean into it.
04
Probability Distribution
NOW → 3M → 6M
Three rows of horizontal bars showing how the regime distribution evolves. Built via Markov-chain forward propagation from the current state and the historical transition matrix.
If the distribution is shifting toward Q4 by the 6m row, it's the model whispering "recession soon."
05
Top 5 Paths
JOINT PROBABILITY · 3-MONTH TRAJECTORY
The five most-likely 3-step regime paths (e.g. Q3 → Q3 → Q3 "stagflation persists" at 14.8%). Joint probabilities account for transition friction, not just modal-step.
If two of the top-5 paths share the same destination, that's where to position.
06
Anomaly Watch
CROSS-ASSET DISLOCATIONS · RANKED BY |σ|
Twelve rows ranked by absolute z-score. Each shows current value, baseline, Δ, headline interpretation, and a trade-direction. HIGH severity (>2σ) gets red; MED (>1σ) gets amber.
The top 2-3 rows are your trade ideas of the day. Each row's "→ trade" line is actionable in size.
07
Desk Note · Morning Brief
EDITORIAL THEMES · NET BIAS · CATALYST
The narrative wrapper. 4-6 numbered themes covering what changed and why it matters today. Each theme has a tag (WATCH / TRANSITION / RISK ON WATCH) and an invalidator. Refreshes daily.
If you only have 60 seconds, read this. It's the model output translated into "what to do."
08
Print Nowcasts
5 TIER-1 FORECASTS · CONSENSUS Δ
For each upcoming print: ATLAS forecast, consensus, delta, 80% CI, and a market-reaction matrix (HOT / IN-LINE / COLD scenarios with SPX/10Y/DXY/Gold reactions).
Pre-position the night before based on the delta. If ATLAS is 3σ off consensus, the print is asymmetric.
09
Allocation Tilt
σ DEVIATION VS BENCHMARK · 4-12W HORIZON
Top-down asset class tilt: OW / N / UW with σ deviation versus benchmark. Each row has a horizon (4-8w / 8-12w) and a thesis. This is the model's recommended portfolio shift.
Use as a sanity check on your existing book. If ATLAS says +2.5σ overweight TLT and you're flat, ask why.
04 · The workflow
Daily, weekly, pre-event
ATLAS isn't a "set and forget" model. It's designed to be read on a cadence: a 60-second morning glance, a 5-minute pre-print check, and a deeper weekly regime review. Here's the recommended workflow.
Daily · Morning
The 60-second glance
1
Read the Desk Note subject line. That's today's frame in one sentence.
2
Check the Regime Banner confidence — if it dropped below 40%, transition is active.
3
Skim Anomaly Watch top 3. Anything new at HIGH severity is your priority.
4
Note the Next Print nowcast vs consensus delta. Pre-position size accordingly.
Pre-Print · 24h before
The asymmetric setup
1
Open the relevant prediction card (NFP, CPI, etc.). Read the reasoning bullets.
2
Compare ATLAS forecast to consensus distribution (analyst spread). If ATLAS is at the edge of the cone, the print is asymmetric.
3
Read the Market Reaction Matrix. If ATLAS = COLD scenario, position for it. Size by 80% CI bound.
4
Set an invalidator: a strike level or sub-index threshold that flips the thesis.
Weekly · Sunday
The regime review
1
Look at the Last 8 Quarters ribbon. Count how many quarters in the current regime — extreme persistence is a fade.
2
Compare 3M vs 6M distribution. If they diverge, transition is in flow.
3
Read the closest analog outcome. If you trust the analog (similar policy backdrop), front-run it.
4
Review Allocation Tilt. Adjust your book to within ±1σ of the model's tilt.
05 · Track record
Backtest + methodology
Every number on the ATLAS panel is computed daily from FRED + auditable backtests. Below are the validation stats and the academic citations underlying the recession composite. The data and the math are open.
Monthly anchors classified, 1990 → present. Each anchor has full quad probability.
97.4%
All-recession Q3/Q4 hit rate
38/39 NBER recession months including transition tails (out-of-sample).
6h
Refresh cadence
Lazy-stale pattern: auto-recompute on first request after 6h staleness.
Methodology citations
Estrella, A. & Mishkin, F. S. (1996). "The Yield Curve as a Predictor of U.S. Recessions." FRBNY Current Issues in Economics and Finance, 2(7). Probit model on 10y-3m spread.
Chauvet, M. & Piger, J. (2008). "A Comparison of the Real-Time Performance of Business Cycle Dating Methods." Journal of Business & Economic Statistics, 26(1). Hamilton-style smoothed Markov-switching.
Hamilton, J. D. (1989). "A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle." Econometrica, 57(2). Two-state Markov-switching framework.
Federal Reserve Economic Data (FRED).St. Louis Fed. Source for all input series: WALCL, TGA, RRP, DGS10, DGS2, T10Y3M, NFP, CPI, ICSA, RSAFS, PPIFIS, plus 24+ supporting series.
06 · Common questions
FAQ
ATLAS uses a lazy-stale-refresh pattern. The first request after 6 hours of staleness triggers a recompute (~2 seconds), and that user gets the cached value instantly while a fresh compute runs in the background. Subsequent requests get the new value.
In practice this means data is never more than ~6 hours old, and the panel reflects every Tier-1 print (NFP, CPI, etc.) within minutes of release because those prints update the underlying FRED series directly.
Every output has an explicit invalidator. The Desk Note themes flag the level or sub-indicator that flips the thesis. The Print Nowcasts publish an 80% CI, so 1-in-5 prints lands outside that band by design.
The Regime Probability Map also publishes Shannon entropy. When entropy is above 1.8 bits (95% of max), the model is explicitly telling you "low conviction — fade single-regime bets." That's a feature, not a bug.
Track-record-wise: 100% NBER recall and 97.4% all-recession hit rate are out-of-sample backtest stats. The model is not a coin flip, but it's also not infallible.
HIGH severity = |σ| > 2, meaning the signal is more than 2 standard deviations from its 5-year rolling baseline. That's a 1-in-44 occurrence under normal distribution.
Each anomaly row carries a "→ trade" line that's actionable in size. Example: WTI Crude at +2.79σ flags "long XLE · long inflation breakevens · short cyclicals/airlines."
The trade direction is calibrated to the historical mean-reversion or trend-extension behavior of that signal. Always cross-reference with the regime banner: a HIGH anomaly that contradicts the regime is worth fading; one that confirms is worth pressing.
Each nowcast uses real-time leading indicators that economists historically don't include in their forecasts:
The model composes these via rule-based weighting calibrated on the past 24 months of residuals. That's the 80% CI: residual std × 1.28.
A newsletter is one person's narrative wrapped around the data they noticed that week. ATLAS is the opposite: a deterministic model that reads every input every day, then a thin editorial layer (the Desk Note) that translates the output into narrative.
The numerical output (regime probabilities, anomaly z-scores, print nowcasts, allocation tilts) doesn't depend on the editorial. If you only ever read the dashboard cards, you'd get 90% of the value. The Desk Note is the 10% that connects them into a thesis.
Yes. The Desk Note is written in plain English and the regime names (Goldilocks / Reflation / Stagflation / Deflation) are intuitive. Each anomaly row carries a translated "→ trade" line that even a long-only investor can act on (e.g. "long XLE" = buy the energy ETF).
If you're newer, the recommended path is: read the Desk Note daily (60 seconds), watch the Regime Banner shift over weeks, and let the Allocation Tilt guide your portfolio adjustments. Skip the technical layers (entropy, Markov projections, Sharpe) until you want them.
No. ATLAS is a nowcasting model built into the Vector Ridge terminal — it's an analytical engine, not a competition entry.
The WCTC (World Cup Trading Championships) is a separate live trading competition that Darren O'Neill competes in personally. His 2025 results (4th Annual Forex 168%, 5th Q3 Forex 65.9%, 1st October Monthly 59.35%) are verifiable on worldcupchampionships.com.
ATLAS provides the macro view that informs Darren's trading framework. The terminal is where the two meet: model output + actual positions + audited track record.
YOUR WORKSPACE · LIVE
See what ATLAS thinks right now
The terminal updates every 6 hours from FRED. Whatever the regime is at this moment — that's what's on the panel.
