Methodology & Data Sources

Data Sources

• Provider: Yahoo Finance API (public, free)
• Update: Weekly (Sunday evenings)
• Window: 4-week rolling periods (~20 trading days)
• Access: Same data retail investors can access

Tier 1: Infrastructure Composite

Components & Weights

CRWV: 30%, DLR: 25%, EQIX: 25%, GLW: 20%

Calculation Steps

1. Get current price for each component
2. Get 52-week high for each
3. Calculate: (Current - High) / High × 100
4. Weighted average: (CRWV×0.30) + (DLR×0.25) + (EQIX×0.25) + (GLW×0.20)

Thresholds

• Green: > -30%
• Yellow: -30% to -50%
• Orange: -50% to -70%
• Red: < -70%

Divergence Alert

If CRWV is >20% worse than composite: Flag company-specific vs sector-wide stress

Tier 2: Equipment (SOXX)

Calculation Steps

1. Get SOXX price: Current and 4 weeks ago
2. Calculate SOXX 4-week return
3. Get SPY price: Current and 4 weeks ago
4. Calculate SPY 4-week return
5. Relative: SOXX return - SPY return

Thresholds

• Green: > -7.5%
• Yellow: -7.5% to -10%
• Red: < -10% for 4 CONSECUTIVE weeks

Why 4 consecutive weeks: Filters volatility, confirms sustained weakness

Tier 3: Mag 7 Composite

Components (Equal Weight)

AAPL, MSFT, GOOGL, AMZN, NVDA, META, TSLA (14.3% each)

Calculation Steps

1. Calculate 4-week return for each stock
2. Simple average of 7 returns
3. Compare to SPY 4-week return

Thresholds

• Green: > -5%
• Yellow: -5% to -10%
• Red: < -10% for 4 consecutive weeks

Alert Level Logic

• Level 0: All tiers green
• Level 1: Tier 1 yellow/orange/red, others green (6-12 month window)
• Level 2: Tier 1 orange/red + Tier 2 yellow
• Level 3: Tier 2 red (equipment confirmed)
• Level 4: Tier 2 + Tier 3 red (full cascade)

Known Limitations

1. Sample Size: Only 2 successes (2000, 2008) - not statistically significant
2. CRWV History: Only 10 months of data, IPO volatility
3. System Design: For sector bubbles, NOT macro corrections (2022 proof)
4. Thresholds: Based on observation, not rigorous optimization
5. False Signals: Possible during volatility (1998-1999 precedent)
6. Lag Time: 4-week windows create delay
7. No Guarantee: Past ≠ future, experimental framework

Backtest Results

Period	Exit Date	SPY% at Signal	DD Avoided	False Pos
2000-2002	Apr 2000	-11%	-37%	3-5
2007-2009	May 2007	-7%	-51%	0
2022	Did not work	N/A	N/A	Multiple