Trend following is one of the oldest and paradoxically, one of the hardest quant strategies to stick with. Its premise is elegant: markets tend to underreact to new information, and those reactions gradually build into trends. The strategy seeks to identify and ride those moves – up or down – until they exhaust themselves.

But while trend following has delivered robust long-term returns across asset classes and centuries, its short-term profile is less elegant. In many regimes, it’s a strategy of many small losses, waiting patiently (and painfully) for the next big winner.

This “death by a thousand cuts” dynamic isn’t a bug, but it’s the price of admission.

Why Trend Following Works

Markets don’t digest information instantly. Investors underreact due to cognitive bias, institutional inertia, or limits to arbitrage. Trend followers exploit this lag by entering positions after momentum becomes statistically significant.

• • Moskowitz et al. (2012) showed that time-series momentum exists across asset classes, countries, and time periods.
  • Hurst, Ooi & Pedersen (2017) extended this to over a century of data, concluding that trend following remains robust even through wars, crises, and shifting market structures.
  • Lempérière et al. (2014) pointed to behavioural and risk-driven explanations – herding, feedback loops, and volatility timing – as key drivers of persistent trends.
  • And in practice, Covel (2007) chronicled the now-legendary Turtle Traders who used simple price rules to deliver strong returns for decades.

When it works, trend is beautiful. But most of the time, it’s just… difficult.

The Pain Between the Peaks

Trend systems don’t fail in sharp up or down markets, they fail in noisy, sideways ones. These periods generate false signals: trades are entered and exited quickly, each one contributing a small loss.

This whipsaw effect is especially prevalent in post-crisis, low-volatility, or highly correlated environments. And because trend systems tend to be slow to reverse, their drawdowns can be protracted and emotionally taxing.

Yet this fragility is part of the design: trend following gives up many battles in order to win the occasional war.

Trend as Crisis Alpha

Trend following is one of the few strategies that can perform when everything else fails.

During market crises, trends often accelerate: equity sell-offs, bond rallies, commodity crashes. Trend systems capture these moves precisely because they are agnostic to direction.

• • In 2008, the Société Générale Trend Index gained more than 20% while global equities collapsed.
  • Similar patterns were observed during the COVID-19 market shock in early 2020 and during major inflation regime shifts.

This “crisis alpha” has made trend following a favorite diversifier in multi-strategy and risk-parity portfolios. But that doesn’t make the dry spells any less painful.

Research Spotlight

• • Hurst, Ooi & Pedersen (2017): Demonstrated that time-series momentum strategies have been profitable across asset classes for over 100 years, including periods of geopolitical turmoil and market regime changes.
  • Moskowitz, Ooi & Pedersen (2012): Provided strong empirical evidence of momentum in futures markets across time and geographies, validating the core premise of trend-following models.
  • Lempérière et al. (2014): Analysed behavioural and structural drivers of trend persistence, such as underreaction, feedback trading, and volatility dynamics.
  • Covel, M. (2007): In The Complete TurtleTrader, chronicled how a group of novices, trained to follow simple price-based trend rules, consistently outperformed the market.
  • Lefèvre, E. (1923): In Reminiscences of a Stock Operator, depicted the trading philosophy of Jesse Livermore, whose patient trend-riding approach remains a psychological template for modern quant systems.

Omphalos: Trend as a Component, Not a Conviction

At Omphalos, trend signals are part of the mosaic, but not the masterpiece.

The AI agents in our system do not “believe” in trend. They evaluate it contextually. Trend-related features are used only when:

• • The signal-to-noise ratio is strong enough.
  • Cross-validated results indicate persistent edge.
  • The broader market regime supports trend behavior.

Our agents are diversified across horizons, assets, methods and strategies. Some may lean into trends, others may explicitly fade them. All strategies are dynamically volatility-scaled and monitored in real time.

Crucially, no single signal drives allocation. Trend is one dimension—never the whole thesis.

A Case in Point: The Enduring Wisdom of Jesse Livermore

In the classic “Reminiscences of a Stock Operator” (1923), the semi-biographical story of Jesse Livermore reveals a foundational truth of trend following:

“It never was my thinking that made the big money for me. It always was my sitting.”

Livermore’s principle – wait for the trend, ride it, and don’t flinch – has outlived every quant model that tried to improve on it. While Omphalos uses advanced AI to process thousands of signals, the underlying lesson remains timeless: discipline beats prediction.

Livermore didn’t have GPUs or backtests but he understood trend persistence better than most data scientists today. His success, and later failure, still serve as a cautionary tale for overconfidence in any one model or market.

Design for Robustness. Train for Resilience.

The biggest myth in trend following is that it’s easy. It’s not. It’s emotionally brutal, operationally fragile, and often counterintuitive.

But when embedded into a diversified, AI-driven system, trend following can still serve as a valuable layer of resilience – especially in macro dislocations.

At Omphalos, we don’t treat trend as a prediction. We treat it as a probability. And we prepare to be wrong, many times, before we’re right.

👉 In the next chapter, we’ll turn to the natural counterpoint to trend: ‘Mean Reversion – Always Profitable (Until It Isn’t)’. We’ll explore why reverting signals are tempting, how they can fail spectacularly, and how modern quants are rethinking short-term contrarian strategies in the age of AI.