More horizons ≠ more diversification.

This new paper shows that the medium-term trend signal — the supposed “sweet spot” of CTAs — is redundant.

• Once short (20–60d) and long (250–500d) horizons are included, the 125-day layer adds no alpha.

• Removing it boosts Sharpe ratios and cuts drawdowns — while keeping benchmark correlation.

• The optimal structure is a barbell: short-term convexity + long-term persistence.

• Time-scale diversification often hides redundancy, not robustness.