Over the past decade, model speed has been a key driver of performance dispersion among systematic trend-following CTAs. Although the largest trend-following managers by AUM remain highly correlated with an average monthly return correlation of around 0.9, their cumulative returns since 2023 have diverged by nearly 50%. This highlights that seemingly similar implementations of the same strategy can behave very differently. Our analysis shows that differences in trend-signal speed alone may explain a substantial share of this dispersion.

We evaluate hypothetical returns of a generic trend model across lookbacks from 10 to 1’000 business days and across both traditional and alternative markets. A 60-70 businessday half-life most closely replicates benchmark CTA behavior, providing a robust empirical definition of “trendfollowing beta”. A similar 75-day half-life best captures the behavior of leading Alternative Markets-focused CTAs.

From 2015–2022, medium-term configurations with halflives of 60-100 business days delivered the strongest and most robust cross-asset performance. These models were sufficiently fast to adapt to regime shifts yet slow enough to avoid excessive turnover and false signals, generating consistent contributions from fixed income, rates, and commodities. Trend-following in equities performed best only at slower speeds, reflecting structural exposure to the long-term equity risk premium rather than shorter-term trend inefficiencies.

Since 2023, this traditional ‘sweet spot’ has become far more fragile. Small changes in model speed have produced unusually large performance differences – over this period, a 50-business-day lookback underperformed a 150- business-day configuration by roughly 35% cumulatively. Faster models struggled with rapid reversals, particularly in fixed income and rates, while slower configurations benefited from persistent equity strength, helping explain the industry’s unusually wide cross-manager dispersion.

An equity-regime analysis confirms that fast models excel in bear markets but pay a cost in normal and bull environments due to higher signal noise and turnover. Very slow models increasingly resemble static equity-riskpremium exposures and fail to respond to major market inflection points. Across the full 2015–2025 period, the 60- 120 business-day speed range delivered the most attractive balance of long-term returns and reliable equity-downside protection.

Overall, our findings reaffirm that medium-term lookback windows remain the most robust foundation for building a diversified trend-following exposure. However, the post2023 regime highlights the limitations of static model configurations and underscores the potential value of adaptive frameworks that adjust model speed to evolving market conditions – an increasingly important consideration for capturing the trend-following risk premium going forward.