Beyond the standard four (trend, carry, mean-reversion, valuation), organized by evidence strength and decay concerns.

Tier 1: High Evidence, Manageable Decay

Variance Risk Premium

Implied minus realized volatility. Compensation for bearing volatility risk.

• Sharpe: 0.5-1.5 across asset classes
• Horizon: 1-3 months
• Implementation: VIX - HAR-RV forecast; contango in VIX futures
• Risk: Concave payoff, tail events

Open Interest Growth

Hong & Yogo (2012): OI predicts commodity returns better than hedging pressure.

• Effect: +0.73%/month per 1σ increase in commodity market interest
• Rationale: Hedging demand meeting limited risk absorption capacity
• Implementation: Z-score of OI changes, cross-sectional ranking

Term Structure Momentum

Changes in roll yield over 12 months, not static carry levels.

• Sharpe: 0.7-1.2 (roughly 2x static carry)
• Rationale: Captures momentum in supply-demand conditions
• Crowding concern: 0.85-0.95 correlation with carry

Factor Momentum

Ehsani & Linnainmaa (2020): Factors with positive returns over prior year earn significant premiums.

• Implementation: Track factor performance over 12 months, tilt allocations
• Key insight: Partly explains individual stock momentum

Tier 2: Moderate Evidence, Worth Exploring

Risk-Neutral Skewness

Negative relationship with subsequent equity returns. Trading RNS yields ~37 bps/week.

• Bonus: RNS loadings help avoid momentum crashes
• Decay: Within one month - requires weekly rebalancing

Macro Surprises

Savor & Wilson: 11.4 bps average on announcement days vs 1.1 bps other days.

• Strongest: FOMC days (R² of 11% for intraday momentum)
• Implementation: Event-driven overlays
• Caution: Most predictability comes from the surprise’s sign, not magnitude

Implementation Notes

• Combine signals using daily IC weighting framework
• Monitor correlation between signals to avoid overcrowding
• Regular out-of-sample testing essential for decay detection

Signal Evaluation Framework

When evaluating new signals, consider:

1. Academic evidence - Peer-reviewed research with out-of-sample testing
2. Economic rationale - Clear explanation for why the signal should work
3. Decay characteristics - How quickly does the signal lose predictive power?
4. Transaction costs - Is the effect large enough after costs?
5. Correlation - How correlated with existing signals?
6. Capacity - Will the signal work at scale?