Paper: arXiv 2507.15876

Authors: Eric Benhamou, Jean-Jacques Ohana, Alban Etienne, Béatrice Guez, Ethan Setrouk, Thomas Jacquot

Abstract

Commodity Trading Advisors (CTAs) have historically relied on trend-following rules that operate on vastly different horizons from long-term breakouts that capture major directional moves to short-term momentum signals that thrive in fast-moving markets. Despite a large body of work on trend following, the relative merits and interactions of short-versus long-term trend systems remain controversial. This paper adds to the debate by (i) dynamically decomposing CTA returns into short-term trend, long-term trend and market beta factors using a Bayesian graphical model, and (ii) showing how the blend of horizons shapes the strategy’s risk-adjusted performance.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced Bayesian graphical models and detailed mathematical proofs of option-pricing theory, indicating high math complexity. It also presents extensive empirical backtesting on CTA replication over 2010-2025 with transaction cost analysis, showing strong empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Disentangle Short vs. Long-Term<br>Trend Factors in CTA Returns"] --> B["Methodology<br>Bayesian Graphical Model"]
  
  B --> C["Data/Inputs<br>Historical CTA Return Indices<br>& Market Data"]
  
  C --> D["Computational Process<br>Dynamic Decomposition<br>of Returns into 3 Factors:"]
  
  D --> D1["Short-Term Trend<br>(Momentum)"]
  D --> D2["Long-Term Trend<br>(Breakout)"]
  D --> D3["Market Beta<br>(General Exposure)"]
  
  D1 & D2 & D3 --> E["Outcome<br>Risk-Adjusted Performance<br>Analysis by Horizon Blend"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper introduces a Bayesian graphical model that simultaneously tracks short-, long-term and market‑beta exposures at the contract level and shows a short‑term‑trend plus raw‑beta sleeve outperforms classic multi‑month break‑outs.

Source sentence
We introduce a Bayesian graphical model that simultaneously tracks short-, long-term and market-beta exposures at the contract level.

Assumes 4

  • Log‑price follows a geometric Brownian motion with constant drift and volatility. model
    Source sentence
    Assume that the log-price follow a geometric Brownian motion dXt = µ dt + σ dWt, X0 = x
  • Pricing of the look‑back‑straddle is done under the risk‑neutral measure (µ = 0). model
    Source sentence
    Throughout this subsection we work under the risk-neutral measure (µ = 0) when pricing, and return to the physical measure when we interpret the result as a drift filter.
  • Short‑ and long‑horizon trend factors have a correlation ρ. statistical
    Source sentence
    Let F ST and F LT be the short– and long–horizon factors and assume that there are correlated: ρ = corr F ST, F LT.
  • Transaction costs are modeled as round‑turn execution expense including bid‑ask, brokerage, exchange and clearing fees plus a small slippage buffer. execution
    Source sentence
    We model three layers of implementation cost: (a) Transaction cost (Tx. Cost). Round-turn execution expense that bundles bid–ask, brokerage, exchange and clearing fees plus a small slippage buffer; see Table 1 for contract-level figures.

Data

Backtest sample: 2010 – 2025 · 24 liquid futures contracts · futures · daily

  • 24 exchange‑traded futures market-prices unknown
    Futures contracts across equity, bond, currency and commodity
    Source sentence
    Table 1 lists the 24 exchange‑traded futures used in the study and the representative basis‑point costs we assume for each.

Costs

Modeled Costs are part of the reported headline results.

Three layers: transaction cost (bid‑ask, brokerage, exchange, clearing fees + slippage), roll cost (front‑to‑next spread), and 0.5% per‑annum management fee; total 1.38–1.68% p.a.

Source sentence
We model three layers of implementation cost: (a) Transaction cost (Tx. Cost)

Finds 5

Reported performance (MKT+STT (Markets + Short-Term Trend) sleeve): Sharpe 0.49 · annual return 7.1% · total return 99.3% · max drawdown 14.9%

  • The short‑term‑trend plus market‑beta (MKT+STT) sleeve achieves the highest Sharpe (0.49) and best Return/MaxDD (0.48) with the shallowest max drawdown (14.9%). performance
    Source sentence
    Integrating raw market returns with the short-term trend factor yields the strongest per‑formance: highest Sharpe (0.49), best Return/MaxDD (0.48), and shallowest max drawdown (14.9%).
  • Combining short‑ and long‑term trend factors (STT+LTT) raises Sharpe/MaxDD efficiency to 2.37, showing the benefit of mixing fast and slow signals. performance
    Source sentence
    Combining STT and LTT raises the Sharpe/MaxDD efficiency to 2.37, reflecting the asymmetric payoff benefits of mixing fast and slow trend signals.
  • The long‑term trend (LTT) factor improves risk‑adjusted returns (Sharpe = 0.39) relative to STT but still lags mixed‑horizon approaches. performance
    Source sentence
    The LTT factor improves risk‑adjusted returns (Sharpe = 0.39) relative to STT but still lags mixed‑horizon approaches.
  • Short‑term trend alone underperforms (Sharpe = 0.20; Return/MaxDD = 0.26) and mainly adds diversification rather than standalone alpha. performance
    Source sentence
    On its own, the STT factor underperforms (Sharpe = 0.20; Return/MaxDD = 0.26). Frequent whipsaws lead to “false starts,” so STT primarily adds diversification rather than standalone alpha.
  • Across extended samples (2020‑2025 and 2004‑2025) the MKT+STT sleeve consistently leads, achieving Sharpe/MaxDD of 3.05 and 4.94 respectively. performance
    Source sentence
    The same sleeve maintains leadership with a Sharpe/MaxDD of 4.94 and a Return/MaxDD of 0.64, demonstrating that the short‑horizon/market‑beta blend is not a recent artefact but a persistent source of convex carry.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 24 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 2 papers in the archive

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (31)

  1. Two Centuries of Trend Following (2014) — link
  2. Dissecting Investment Strategies in the Cross‑Section and Time Series (2015) — link
  3. Do momentum-based trading strategies work in the commodity futures markets? (2015)
  4. Basis-momentum (2019)
  5. In Pursuit of Trend-Following Beta: The Promise and Pitfalls of Replication (2024)
  6. Momentum Strategies in Futures Markets and Trend‑Following Funds (2013) — link
  7. Beyond Trend Following: Deep Learning for Market Trend Prediction — → 2407.13685
  8. Demystifying time-series momentum strategies: Volatility estimators, trading rules and pairwise correlations (2020)
and 23 more
  1. Autoencoder asset pricing models (2021)
  2. High-order Hidden Markov Model for trend prediction in financial time series (2019)
  3. Machine Learning: A Probabilistic Perspective (2012)
  4. Time Series Momentum (2012) — link
  5. Futures Markets, Bayesian Forecasting and Risk Modeling (2010)
  6. Implementing Models in Quantitative Finance: Methods and Cases (2008)
  7. XI. The asymmetrical probability-curve (1896)
  8. Following the trend: diversified managed futures trading (2023)
  9. Forecasting, structural time series models and the Kalman filter (1990)
  10. Trend without hiccups: a kalman filter approach (2018) — link
  11. Kalman filter demystified: from intuition to probabilistic graphical model to real case in financial markets (2018) — link
  12. 10 structural time series models (1993)
  13. The horseshoe estimator for sparse signals (2010)
  14. Momentum trading, disposition effects and prediction of future share prices: An experimental study of multiple reference points in responses to short- and long-run return trends (2009) — link
  15. Short-term reversals and longer-term momentum around the world: Theory and evidence (2022)
  16. The risk in hedge fund strategies: Theory and evidence from trend followers (2001)
  17. Path dependent options: The case of lookback options (1991)
  18. Path dependent options:" buy at the low, sell at the high" (1979)
  19. Trend Following Struggles to Return to Vogue (2024) — link
  20. Should Trend Followers Lower Their Horizons? (2024) — link
  21. Novel Risk-Off CTA Strategy Passes Tariff Test (2025) — link
  22. Cliff Asness: AI is ‘Still Just Statistics’. The AQR Head on Passive Investing, Private Equity, Tail Insurance, Trend-Following and Much More (2024) — link
  23. Designing robust trend-following system: Behind the scenes of trend-following (2018)

1 of 31 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.