Trend following in 2025

The following investigation will explore which models are currently demonstrating effectiveness and the ways in which they differ from those that emerged during the previous two decades.

Recent studies from 2023–2025 have demonstrated that contemporary trend following strategies are characterized by a decline in velocity, increased diversification, heightened sensitivity to macroeconomic factors, and a growing reliance on machine learning algorithms.

The initial point of departure is as follows: The following investigation will examine the phenomenon of trend following in the 2000–2025 market cycle.

It has been demonstrated that, over extended time frames, managed futures and CTA trend strategies have yielded returns that are commensurate with those of a global 60/40 portfolio. These strategies exhibit virtually no correlation to traditional stock and bond indices.

A recent long-term analysis, entitled "The Patience Premium," has revealed that the SG Trend Index consistently generates favorable returns over a period exceeding 25 years. However, the study also noted the presence of discernible cycles of overperformance and underperformance, which frequently extend over a duration of several years.

Following a particularly robust period in 2022, numerous CTA trend programs transitioned into a phase of relative weakness beginning in 2023.

This transition is characterized in a "Strategy Spotlight" study as a typical drawdown phase within the strategy cycle.

Concurrently, evaluations of the Nordic Hedge Index trend followers and managed futures funds demonstrate that substantial double-digit returns (approximately 12.2%) were attained on average in 2024, albeit with considerable dispersion among individual managers.

This underscores the notion that trend following has confirmed its role as a hedge against crises and regime changes in the current environment, even though the years 2023–2024 were operationally difficult for many systems.

The following section will provide a synopsis of the findings from recent studies (2023–2025).

A synthesis of extant studies reveals a number of common patterns.

Pure, fast price trend models have demonstrated significantly diminished performance since the 2010s, exhibiting a high degree of regime dependence.

Models that incorporate macro information, cross-asset signals, or robust volatility control have been shown to exhibit superior and more stable Sharpe ratios.

Machine learning-based trend followers have been shown to exhibit high levels of outperformance relative to benchmarks in equity universes. However, this outperformance is observed in backtests and is accompanied by notable transaction cost sensitivity.

Since 2000, aggregated evaluations have demonstrated that signals with slower signal speeds, which have persisted over several months, have generally outperformed very fast signals.

This shift in perspective has significant implications for the identification of winning models, which, as previously defined, have been considered to be a single entity. However, a more accurate and nuanced understanding of this concept suggests that a winning model is, in fact, a combination of three distinct building blocks: robust timing, a broad universe, and additional sources of information (macro, cross-asset, and machine learning).

Trend for 2025: Tried-and-tested principle – new winning models

Many are already writing off trend following after the tough years of 2010–2020 and the difficult phases of 2023/24 – but current studies tell a different story.

The trend premium is intact, but the strategies that really deliver today look different from the simple 12-month price-only models of the last 20 years.

Macro-enhanced trend models in EM FX, which combine price trends with fundamentals such as current account balance, growth, and inflation, achieve significantly higher and more stable Sharpe ratios than pure price models over 2000–2025.

Cross-asset time series momentum (I-XTSM), for example, uses commodity signals to forecast equity returns and reduces the classic “momentum crashes” of old trend-following approaches.

Analyses by Man Group show that slower trend signals (3–12 months) are clearly superior in the long term—high trading frequency eats into returns without significantly increasing the premium.

Deep learning trend followers (CNN-LSTM) on the S&P 500 Index deliver double-digit alphas in backtesting, as long as turnover and transaction costs are strictly controlled.

Long-term studies of managed futures/CTA strategies confirm that broadly diversified, volatility-scaled trend models achieve returns comparable to classic 60/40 portfolios over decades, with significantly better crisis resilience.

The bottom line: trend following works—but price + macro + cross-asset and a longer-term perspective now beat the old, fast price-only models.