How Aegis Works

Three-Layer Risk Architecture (ADR-006)

🎯 Core Philosophy

  • Market Prices > News > Lagging Data: Speed matters for risk reduction
  • Layered defense: Tactical (fast) + Strategic (fundamental) + Narrative (context)
  • Minimize drawdowns: Accept more false positives for faster exits
  • Graduated response: 100% → 70% → 50% → 30% equity allocation tiers
  • Objective scoring: Clear formulas, reproducible calculations

📉 Why Avoiding Losses Beats Chasing Gains

Aegis is built on a fundamental asymmetry in investing: losses hurt more than gains help.

Loss Gain Required to Recover
-10%+11%
-20%+25%
-30%+43%
-40%+67%
-50%+100%

The math favors caution:

  • Missing 10% upside only requires +10% to catch up
  • Taking a 40% loss requires +67% just to break even
  • The cost of a false alarm (single-digit % missed gains) is far less than the cost of a correct alarm ignored (double-digit % drawdown)

Example - COVID-19 (Feb-Mar 2020):

  • WHO alert on Jan 30 → Aegis would signal YELLOW
  • Action: Reduce to 70% equity
  • Feb 19: S&P at all-time high (you "missed" ~3%)
  • Mar 23: S&P bottom (-34% from peak)
  • Your portfolio: ~24% down instead of 34% (30% cash cushion)
  • Plus: Cash available to redeploy at lower prices

Aegis doesn't predict crashes - it positions you slightly defensively when warning signs appear. The goal is avoiding catastrophic drawdowns, not timing every dip.

🏦 How Aegis Compares to Professional Risk Management

None of this is novel. The same indicators and philosophy are used by professional funds:

Type Examples What They Do
Risk ParityBridgewater, AQRSystematic risk allocation
Macro Hedge FundsBrevan Howard, TudorSame indicators, faster
CTAsMan AHL, WintonTrend-following + risk rules
Quant ShopsTwo Sigma, DE ShawAll this + ML + alt data

What they have that Aegis doesn't:

  • Proprietary data (satellite imagery, credit card flows)
  • Speed (milliseconds vs daily updates)
  • Leverage and derivatives for hedging
  • Teams of PhDs and decades of backtesting

Why Aegis still makes sense:

  • Access: Can't invest in Bridgewater's Pure Alpha (min $7.5M, institutional only)
  • Cost: Hedge funds charge 2% + 20% of profits. Aegis is free.
  • Transparency: You see exactly why the score is what it is
  • Control: You decide what to do with the signal
  • Simplicity: Good enough for personal portfolio

Think of it like home security: the pros do it better, but a Ring doorbell + locked doors gives you 80% of the value at 1% of the cost. Aegis isn't trying to beat the market - just avoid catastrophe.

Understanding Scores: What Does 0.0 Mean?

Aegis uses a signal-based scoring system. A score of 0.0 is not an error or missing data - it means:

  • Data is present and the indicator is being actively monitored
  • No warning conditions detected - values are within normal/healthy ranges
  • No risk signal to report - the system only adds points when thresholds are crossed

Examples:

  • Credit = 0.0: HY spreads at 2.9% are below the 5.5% warning threshold - credit markets healthy
  • Positioning = 0.0: VIX at 15.8 is in normal range (15-40) - no complacency or panic signals
  • Trend = 0.0: S&P 500 above both 50 and 200-day moving averages - uptrend intact

Think of it like a medical checkup: 0.0 means "all clear" - your vitals are normal. The score only increases when something needs attention. A patient with no symptoms doesn't get a baseline "1.0" just for existing.

Note: "N/A" indicates data is unavailable or failed to fetch - this is different from 0.0.

📊 Three-Layer Architecture

Layer 1: Tactical (Daily)

Purpose: Fast-moving market signals for rapid response

Components:

  • VIXCBOE Volatility IndexThe "fear gauge" - expected S&P 500 volatility over 30 days. Warning: >20, Danger: >25, Panic: >40.Learn more → regime (level >25 = danger) + term structureVIX Term StructureRatio of VIX3M to VIX. Normal: VIX < VIX3M (contango). Inverted (<1.0): panic mode.Learn more → (inverted = panic)
  • Credit spread velocityHigh-Yield Spread VelocityRate of change in junk bond spreads. +30 bps/5 days = warning, +50 bps/5 days = danger. Credit leads equities in crises.Learn more → (+50 bps/5 days = danger)
  • S&P 500 trendTrend / Moving AveragesPrice position relative to 50-day and 200-day moving averages. Below 200 MA = bear territory. Death cross = 50 MA below 200 MA.Learn more → (below 200 MA + declining = bearish)
  • Cross-assetCross-Asset Risk-OffMultiple assets moving defensively: treasuries up, gold up, USD up, equities down, HY spreads widening. 4/5 signals = coordinated risk-off.Learn more → risk-off (4/5 signals = danger)
  • Volume/Liquidity FilterVolume/Liquidity FilterCompares SPY volume to 20-day average. Low volume periods make VIX unreliable. <60% = caution, <50% = warning, <40% = VIX "lying". Also flags holiday windows.Learn more → (<40% = VIX unreliable, holiday windows flagged)
  • Breadth DivergenceRUT/SPX DivergenceRussell 2000 vs S&P 500 20-day performance. Small caps lagging while large caps rise = narrowing breadth = warning sign. Trigger: RUT underperforming by 2%+.Learn more → (RUT/SPX divergence, small caps lagging = narrowing breadth)
  • AI Bubble (NVDA)NVIDIA Canary SignalNVDA serves as the "canary in the coal mine" for AI infrastructure spending. Monitors price vs 50/200-day MAs and RSI. Below 200 MA = exit AI-dependent trades. Note: Forward-looking signal, no historical AI bubble to backtest.View NVDA → (NVDA below 50 MA = warning, below 200 MA = exit AI trades)
  • Turbulence IndexKritzman Turbulence IndexMahalanobis distance measuring correlation breakdown across 5 assets (SPY, TLT, GLD, HYG, UUP). Detects when diversification fails. >43 = crisis.Learn more → (correlation breakdown, >43 = crisis override)

Weights: VIX 22%, Credit 18%, Trend 18%, Cross-Asset 14%, AI Bubble 12%, Volume 8%, Breadth 8% | Turbulence is an override (not weighted)

See Tactical page for alert tiers and details →

Layer 2: Strategic (Weekly)

Purpose: Fundamental macro regime assessment

5 Dimensions (weights sum to 1.0):

  • Recession Risk (30%): Unemployment claims velocityJobless Claims VelocityYear-over-year % change in weekly jobless claims. Rising claims velocity signals weakening labor market - a leading indicator.Learn more →, ISM PMIPurchasing Managers IndexSurvey of manufacturing activity. Above 50 = expansion, below 50 = contraction. Regime cross is key signal.Learn more → regime, yield curveTreasury Yield Curve10Y-2Y spread. Inversion has predicted every US recession since 1970. Also monitors 3M→18M forward spread.Learn more →, prediction marketsKalshi Prediction MarketsReal-money bets on economic outcomes. Recession probability and unemployment expectations provide forward-looking crowd wisdom.Learn more →
  • Credit Stress (25%): HY spreadHigh-Yield Bond SpreadDifference between junk bond yields and Treasuries. Velocity (70%) + level (30%). Normal: 300-400 bps.Learn more → velocity + level, BBB spreadsBBB Corporate SpreadsLowest investment-grade rating. Widening BBB spreads signal stress at the boundary between investment grade and junk.Learn more →, TED spreadTreasury-Eurodollar SpreadDifference between interbank rates and T-bills. High TED = banks don't trust each other.Learn more →, bank lending standardsSenior Loan Officer SurveyFed survey on credit tightening. Tighter standards precede economic slowdowns.Learn more →
  • Valuation Extremes (20%): CAPEShiller P/E Ratio10-year inflation-adjusted earnings. Above 30 = historically expensive. Doesn't cause crashes but amplifies them.Learn more →, Buffett indicatorMarket Cap / GDPTotal stock market value divided by GDP. Above 150% = overvalued territory.Learn more →, Forward P/EForward Price-to-EarningsPrice divided by expected future earnings from analyst estimates.Learn more →
  • Liquidity Conditions (15%): Fed funds trajectoryFederal Funds RateThe Fed's primary policy tool. Rate of change matters more than level. "Don't fight the Fed."Learn more →, M2 velocityM2 Money SupplyCash + checking + savings. Growing M2 = more liquidity. Velocity measures how fast money circulates.Learn more →, real ratesReal Interest RateFed funds minus inflation. Positive real rates = restrictive policy. Rapid increases cause selloffs (2022).Learn more →
  • Positioning & Speculation (10%): CFTC positioningCommitments of TradersWeekly futures positioning data. Extreme speculator positioning is a contrarian signal.Learn more →, VIX futures complacencyMarket ComplacencyHeavy short VIX positions + low VIX often precede volatility spikes.Learn more →

See Strategic page for alert tiers, warnings, and worked example →

Layer 3: Canary (Daily)

Purpose: Early warning from adjacent markets — "Is stress building elsewhere?"

4 Sub-Signals (composite weighted average):

  • EM Corporate OASEmerging Market OASFRED BAMLEMCBPIOAS. EM credit spreads widen before US credit — stress appears in weaker markets first.Learn more → (55%): EM credit widens before US credit
  • Bank/SPX RelativeXLF vs S&P 50020-day rolling relative return of XLF (financials) vs ^GSPC. Banks underperform 2-4 weeks before broad market stress.Learn more → (20%): Banks underperform 2-4 weeks before stress
  • Copper/Gold RatioCu/Au RatioHG=F / GC=F. Copper = economic growth, Gold = safety. Declining ratio = risk-off shift.Learn more → (15%): Declining ratio = risk-off (copper = growth, gold = safety)
  • CBOE SkewCBOE Skew Index^SKEW measures demand for tail-risk hedges (out-of-the-money puts). Rising skew = increasing crash insurance demand.Learn more → (10%): Rising skew = increasing demand for tail-risk hedges

Scoring: Each signal uses a 60-month rolling Z-score. For declining signals (Cu/Au, Bank/SPX), Z is negated. Z-scores map to 0–10 via piecewise linear bands.

Role: Amplifier, not independent trigger. Fires advisory at score ≥3.0. Adds context to tactical/strategic alerts but does not change tiers alone.

See Canary page for scoring details and thresholds →

Layer 4: Narrative (NexusMind)

Purpose: Provides context when signals fire - does NOT trigger actions

  • Source: NexusMindNexusMind IntegrationAI-powered news analysis system that filters investment-relevant articles and assigns risk tiers (RED/YELLOW/GREEN/BLUE).Learn more → investment_risk filter output
  • Tiers: RED (act now), YELLOW (monitor), GREEN (opportunity), BLUE (educational)
  • Usage: Query when tactical/strategic signals elevate to understand "why"
  • Philosophy: Explains, does not trigger. Data leads, narrative follows.

See Narrative page for tier details →

🔍 Data Sources

  • FREDFederal Reserve Economic DataFree database of 800,000+ economic time series from the St. Louis Fed. Primary source for macro data.Learn more →: Federal Reserve Economic Data (macro indicators)
  • Yahoo Finance: Real-time market data (VIX, S&P 500, ETFs for cross-asset signals)
  • KalshiKalshi Prediction MarketsCFTC-regulated prediction market. Real-money bets on economic events provide forward-looking probabilities.Learn more →: Prediction markets (recession probability, unemployment expectations)
  • Shiller: CAPE ratioShiller CAPEProfessor Robert Shiller's cyclically adjusted P/E ratio from Yale website.Learn more → (monthly update)
  • CFTC: Commitments of TradersCOT ReportWeekly futures positioning data for speculators and hedgers.Learn more → (positioning data)
  • NexusMind: AI-filtered investment news for narrative context

📈 Historical Performance

Backtested: 300 months (2000-2024)

  • Historical Max: 5.55 (April 2020, COVID crash)
  • Major Crises Detected: 3 of 4 (2000-2002, 2008-2009, 2020)
  • False Positive Rate: ~2.4% (acceptable for capital preservation)
  • Lead Time: 0-6 months (coincident indicator, not predictive)

System Limitations:

  • Coincident, not leading - detects crises in progress, doesn't predict them
  • Monthly sampling misses fast crashes (e.g., Feb-Mar 2020)
  • Valuation warnings alone don't trigger alerts (2000 dot-com initially missed)
  • Tactical layer added to catch fast-moving 2020-style crashes

📚 Learn More