Methodology Changelog
Why We Version Control
Transparency is critical for credibility. We document every change to our methodology so you can:
- Understand why signals may behave differently over time
- Evaluate whether changes improved or degraded performance
- Trust that we're not "moving the goalposts" after the fact
- Learn from our refinements and mistakes
- Hold us accountable
All historical data uses the methodology version that was active at the time. We never retroactively apply new methodology to historical signals.
Tier 1: Infrastructure Demand Composite
| Component | Weight | Metric |
|---|---|---|
| CoreWeave (CRWV) | 30% | % from 52-week high |
| Digital Realty (DLR) | 25% | % from 52-week high |
| Equinix (EQIX) | 25% | % from 52-week high |
| Corning (GLW) | 20% | % from 52-week high |
Thresholds:
- Green: > -30%
- Yellow: -30% to -50%
- Orange: -50% to -70%
- Red: < -70%
CRWV Divergence Alert: Triggered when CRWV is >20% worse than other components average.
Tier 2: Equipment Providers
- Primary: SOXX (Semiconductor ETF)
- Confirmation: NVDA
- Benchmark: SPY
- Metric: 4-week return vs SPY
Thresholds:
- Green: > -7.5%
- Yellow: -7.5% to -10%
- Red: < -10% for 4 consecutive weeks
Tier 3: Narrative Leaders
- Components: AAPL, MSFT, GOOGL, AMZN, NVDA, META, TSLA (equal weight)
- Benchmark: SPY
- Metric: 4-week return vs SPY
Thresholds:
- Green: > -5%
- Yellow: -5% to -10%
- Red: < -10% for 4 consecutive weeks
Alert Levels
- Level 0 (All Clear): All tiers green
- Level 1 (Early Warning): Tier 1 not green, Tiers 2&3 green
- Level 2 (Infrastructure Stress): Tier 1 orange/red, Tier 2 yellow
- Level 3 (Equipment Confirmed): Tier 2 red
- Level 4 (Full Cascade): Tiers 2&3 both red
Data Source
- Yahoo Finance API (public, free)
- Weekly updates recommended
Design Rationale
This methodology is based on analysis of the 2000 Dot-Com and 2008 Financial Crisis cascades. The composite approach for Tier 1 was chosen to:
- Reduce single-stock risk (CRWV IPO volatility)
- Provide broader infrastructure validation
- Enable historical backtesting (REITs have longer history)
Known Limitations at Launch
- CRWV has only ~10 months of trading history
- Sample size of 2 historical examples is small
- System failed in 2022 (macro-driven correction)
- Thresholds are based on limited data
Future Change Process
Any methodology changes will follow this process:
- Proposal: Document proposed change with rationale
- Backtest: Test against historical data when possible
- Announcement: Publish change before implementation
- Version Bump: Increment version number
- Documentation: Full details added to this changelog
We will NEVER retroactively apply methodology changes. Historical signals will always reflect the methodology that was active at that time.
Planned Considerations
Areas we may evaluate for future versions (no guarantees):
- Adjusting CRWV weight as more trading history accumulates
- Refining thresholds based on additional data
- Adding or removing composite components
- Adjusting consecutive week requirements
Any changes will be thoroughly documented before implementation.