Institutional allocators rely on managed futures strategies for diversification and drawdown control, yet often misunderstand how risk is actually taken inside these allocations. They frequently lack clarity on which trend horizons drive performance, how similar managers truly are to one another and to benchmarks, and how differences in horizon mix shape behavior during periods of market stress.
By decomposing CTA managed futures returns into a small set of distinct trend horizons (fast, medium, and slow), this post shows that much of the variation across managers and benchmarks reflects differences in horizon mix rather than fundamentally different strategies. Framing managed futures allocations in this way allows investors to better diagnose overlap, benchmark more precisely, and assess whether their exposure is aligned with its intended role in the portfolio.
The analysis that follows is necessarily technical, introducing a horizon-based framework that decomposes CTA returns into a limited set of systematic building blocks. While the mechanics are described in detail, the objective is practical: to provide a clearer, more transparent way to interpret managed futures behavior and to link observed outcomes to explicit, governable risk choices.
WHAT SITS INSIDE TREND FOLLOWING
Commodity trading advisors (CTAs) and managed futures funds are often described in broad terms as “trend followers.” A closer look shows that CTA allocations can be decomposed along three distinct dimensions that help explain differences in risk, behavior, and outcomes.
Which trend horizons actually drive risk and return, for example, fast 20‑day versus very slow 500‑day signals.
How similar different managers are to each other and to benchmark indices in terms of those horizons.
How horizon mix interacts with realized performance, especially in periods of market stress.
The research underlying this post constructs a library of five mono‑horizon trend‑following strategies (20, 60, 125, 250, and 500 trading days) and uses them as building blocks to decompose both the SG CTA Trend Index, a widely followed CTA benchmark, and seven anonymized CTA programs.
This “horizon fingerprint” perspective turns a black‑box allocation into a more transparent set of style and risk exposures, which can be explicitly managed via SMAs or AI‑driven replication mandates.
A HORIZON-BASED VIEW OF CTA RISK
From Trend to Trend Horizons
Most CTA replication work proceeds along one of two paths:
Bottom‑up, starting from futures and reconstructing positions market by market, or
Top‑down, modelling returns with generic trend and carry factors.
The mono‑horizon approach sits between these. It keeps a realistic futures universe and cost structure but organizes trend exposure by a horizon look‑back straddle [1]window, used as a generic way to replicate managed futures, rather than by an individual contract or generic factor.
Conceptually, the framework asks:
“How much of this manager’s risk comes from fast, medium, and slow trend signals, and at what overall risk intensity?”
For allocators, this intermediate level of detail is often the most useful: it is rich enough to distinguish strategies, but simple enough to support clear portfolio investment decisions.
The Mono-Horizon Library
The analysis is built on a diversified set of liquid futures across:
Equity indices,
Government bond and short‑rate futures,
Major G10 currency futures versus the US dollar, and
Key commodity contracts (energy and metals).
Each mono‑horizon sleeve:
Uses the same universe and volatility target,
Faces the same assumptions for transaction costs, roll costs and a 50 basis points (bps) management fee, and
Differs only by the look‑back window used to construct its trend signal (20, 60, 125, 250, or 500 days).
The signal itself can be interpreted as the delta of a look‑back straddle: it is long near recent highs, short near recent lows, and close to flat in trading ranges. Positions are bounded and combined with risk‑parity weights so that each sleeve is an investable, volatility‑controlled portfolio.
The five sleeves therefore span:
Fast trend (20 to 60 days),
Medium‑term trend (around 125 days), and
Slow trend (250 to 500 days).
Together, they form a basis of horizon factors that can be used to explain and replicate CTA behavior.
WHAT IS INSIDE THE SG CTA TREND INDEX?
Regression on Mono-Horizon Factors
We begin by applying the framework to the SG CTA Trend Index. The index’s daily excess returns over the past five years are regressed on the five mono-horizon sleeves, with statistically non-significant horizons sequentially removed via a standard backward-elimination procedure.
The resulting model is both simple and instructive:
The intercept is small and statistically insignificant, suggesting limited residual “alpha” once horizon styles are accounted for.
The index is well explained by a positive combination of three horizons:
20‑day (fast),
125‑day (medium‑term), and
500‑day (very slow).
The sum of the three betas is approximately 1.06, implying that the index behaves much like a fully invested multi‑horizon trend portfolio.
Roughly two‑thirds of the exposure lies in the mid/slow block (125d + 500d); about one‑third in the fast 20‑day sleeve.
From a style standpoint, SG CTA Trend can therefore be viewed as a mid‑ and slow‑trend strategy with a structurally embedded fast overlay.
Table 1: SG CTA Trend index: horizon decomposition (last 5Y).
Correlation Is Not the Whole Story
At first glance, you might expect the regression to select the sleeve that is most correlated with the index.
The correlation matrix, however, tells a different story:
The 125‑day and 250‑day sleeves have the highest correlations with the index (around 82%).
The 20‑day sleeve is the least correlated, with a correlation of about 66%.
Despite this, the regression retains 20‑day and 500‑day, and drops 250‑day. This highlights an important point for practitioners: the best multi‑factor representation is not necessarily built from the individually “closest” factors.
Fast and slow horizons contribute complementary information:
Fast trend helps capture sharp reversals and shorter‑lived regimes.
Slow trend anchors the portfolio to longer‑term drifts and tends to stabilize drawdown behavior.
Used together, they can deliver a more robust payoff pattern than any single medium‑term sleeve, even one with higher standalone correlation.
Table 2: Correlation Matrix of mono-horizon sleeves and CTA Index (monthly, in%).
PT 20d/60d/125d/250d/500d = CTA Pure Trend N d Decoding; CTA Idx = NEIXCTAT Index.
MANAGER-LEVEL HORIZON FINGERPRINTS
The same methodology is applied to seven anonymized CTA programs (CTA 1–CTA 7) that are, or have been, constituents of the SG CTA Trend index. For each manager, a regression on the five mono‑horizon factors is estimated over the last five years, with non‑significant horizons iteratively removed.
Common Structure Across the Cross-Section
Across managers, several consistent patterns emerge:
Trend factors explain most of the variation: Coefficients on retained horizons are positive and highly statistically significant; intercepts are generally small. The mono‑horizon library appears to capture the dominant systematic component of returns.
Every manager combines fast and slow sleeves: Each program has material exposure to at least one short horizon (20d or 60d) and at least one long horizon (250d or 500d). A slow sleeve — most often 500 days — acts as a recurring backbone.
The mid band is the main style dial: Exposure to the 60–125‑day range varies widely: some CTAs are mid‑heavy, others use it sparingly. This region is therefore a primary source of differentiation in horizon style.
Overall trend intensity is “around one,” but not fixed: The sum of horizon betas per manager ranges from roughly 0.75 to 1.20. Some programs resemble fully invested multi‑horizon trend portfolios; others operate at somewhat lower or higher trend beta levels.
Interpreted through this lens, many CTAs look less like fundamentally distinct return streams and more like different convex combinations of shared fast, mid, and slow building blocks.
Horizon Shares and Examples
Rebasing the horizon betas to 100% yields a horizon share for each program. For example:
The index itself is roughly 31% 20‑day, 36% 125‑day and 33% 500‑day.
CTA 1 is dominated by slow trend, with around 63% in 500‑day and 37% in 60‑day.
CTA 5 combines 20‑day, 60‑day and 250‑day sleeves but has negligible exposure to 125‑day and 500‑day.
CTA 7 closely mirrors the index, with an approximately one‑third fast, one‑third mid, one‑third slow composition.
These stylized numbers provide an immediate, quantitative sense of how each strategy differs from the benchmark and from its peers.
Table 3: Horizon shares (in %) for the index SG CTA Trend and the 7 CTAs.
(5Y regressions on mono-horizon trend factors, coefficients rebased to 100%).
HORIZON MIX AND REALIZED PERFORMANCE
The analysis further relates these horizon fingerprints to 5‑year risk‑adjusted performance metrics (Sharpe ratio and Return/Maximum Drawdown).
While the sample is limited and the results should be interpreted cautiously, three observations are noteworthy:
A strong slow‑trend backbone is associated with better drawdown efficiency: CTA 1, whose horizon mix is tilted heavily to the 500‑day sleeve, exhibits the highest Sharpe ratio (0.75) and the best Return/Max Drawdown ratio (0.84), substantially above the index (0.38 and 0.35, respectively). This aligns with earlier findings that very slow horizons can improve drawdown profiles by emphasizing persistent moves over noise.
Index‑like horizon mixes deliver index‑like outcomes: CTA 7, whose fast/mid/slow split closely matches SG CTA Trend, displays risk‑adjusted performance that is very similar to the index itself. In effect, it offers an efficient, slightly de‑levered implementation of the benchmark’s horizon structure.
Concentrated fast or mid‑band exposures can weaken risk‑adjusted returns: CTAs 2, 4 and 6, which lean more aggressively into fast or mid‑band risk, show weaker Sharpe ratios and lower Return/Max Drawdown, despite all having some slow exposure. CTA 5, with an idiosyncratic mix that omits the 125‑ and 500‑day sleeves, occupies a middle ground in performance terms.
These patterns do not imply that slow trend is universally superior or that fast trend should be avoided. Rather, they suggest that:
Slow trend often plays a performance stabilizing role,
Fast trend adds reactivity and convexity, and
Large bets in the mid band or highly concentrated fast exposures, without a dominant slow core, may be more fragile in the sample examined.
IMPLICATIONS FOR ALLOCATORS AND MANDATE DESIGN
The mono‑horizon framework lends itself directly to both diagnostics and implementation.
A Practical Diagnostic Checklist
For each CTA or index allocation, allocators can seek to answer the following:
Horizon mix: What percentage of trend risk is fast (20–60 days), medium‑term (around 125 days) and slow (250 to 500 days)?
Trend intensity: Is the overall trend beta closer to 0.7, 1.0 or 1.2 relative to the mono‑horizon basis?
Stability over time: Is the horizon composition relatively stable, or is the manager actively timing horizons?
Benchmark comparison: How does the horizon fingerprint compare with SG CTA Trend? Does the allocation meaningfully diversify the index?
Crisis behavior: Did the strategy’s realized behavior in stress periods align with what its horizon mix would suggest?
Even approximate answers provide a more structured basis for portfolio and risk‑budget discussions than generic labels such as “faster” or “more tactical.”
Using AI-Driven or SMA Mandates to Adjust Horizon Exposure
Growing demand for AI‑driven replication and customized SMAs reflects a desire not only to reduce fees but also to shape exposures more intentionally.
A horizon‑based view offers a natural design space for such mandates:
Adding a slow‑trend core: For portfolios dominated by medium‑term CTAs, a mandate can be specified to emphasize 250‑ and 500‑day sleeves at a defined risk budget, providing a more robust backbone to the overall allocation.
Introducing a controlled fast overlay: For investors with substantial exposure to slow CTAs or macro‑oriented systematic strategies, a carefully sized fast overlay (20 to 60‑day horizons) can improve responsiveness to regime shifts while keeping turnover and costs within acceptable bounds.
De‑crowding the mid band: If diagnostic work reveals that the aggregate CTA book is heavily concentrated around 60 to 125 days, an SMA or replication mandate can deliberately underweight this region, reallocating risk toward fast and slow sleeves to improve diversification.
In each case, AI‑enabled tools can assist in parameter selection, execution, and risk management, but the overarching horizon mix remains a governable choice of the investment committee, grounded in a transparent factor interpretation.
CONCLUSION
Mono-horizon trend decomposition provides a clearer and more interpretable way to understand CTA risk. The analysis shows that both benchmarks and individual CTAs can be explained as combinations of a limited set of shared trend horizons, rather than as opaque strategies.
At the index level, the SG CTA Trend benchmark emerges as a convex combination of fast, medium, and very slow horizons, with a structural tilt toward mid and slow trend and a meaningful fast overlay.
At the manager level, much of the apparent diversity across CTA programs reflects different allocations across the same horizon building blocks rather than fundamentally distinct sources of return.
From a portfolio perspective, slow horizons tend to underpin drawdown resilience, fast horizons contribute reactivity and convexity, and the mid band acts as a style lever that meaningfully differentiates strategies.
For allocators, reframing managed futures exposures in terms of horizon mix enables clearer benchmarking, better overlap diagnostics, and more intentional mandate design.
Framing CTA allocations as explicit horizon-based exposures allows investors and fiduciaries to move beyond generic classifications and toward governable, portfolio-relevant risk decisions, whether implemented through traditional SMAs or AI-supported replication approaches.
Backtested or simulated results referenced in this discussion are hypothetical, subject to model risk and to the assumptions on costs and capacity described in the underlying research. Past performance is not indicative of future results.
Reference
[1] William Fung and David A. Hsieh, “The Risk in Hedge Fund Strategies: Theory and Evidence from Trend Followers,” Review of Financial Studies, 14(2), 313–341, 2001.
