Diversified systematic trend following has a well-documented tendency to perform strongly during periods of market stress. The term “Crisis Alpha,” developed by Alex Greyserman and Kathryn Kaminski in their landmark work Trend Following with Managed Futures: The Search for Crisis Alpha, captures this property precisely: trend following programs have historically outperformed during the periods when most other investment styles and asset classes were failing. This is not a recent phenomenon. It is an enduring one, spanning decades of evidence across multiple market regimes.
To understand why, it helps to start with the performance record.
The Performance Record
The S&P 500 Total Return Index from January 2000 to September 2022 tells a familiar story (Figure 1). Two drawdowns of approximately 50% mark the tech bubble collapse of 2002/2003 and the Global Financial Crisis of 2008/2009. A short correction in 2018 coincided with the Trump trade tariffs and rising interest rates. A brief but sharp dislocation arrived in March 2020 with Covid. From January 2022, a building drawdown developed as inflation accelerated, Chinese growth slowed, and geopolitical uncertainty intensified.
Figure 2 adds two further benchmarks: the TTU Trend Following Index, a composite of approximately 55 trend following programs each with a track record exceeding 15 years, and the Vanguard VBIAX ETF, a benchmark for the traditional 60/40 portfolio of US equities and bonds.
Two things stand out immediately. The first is the breakdown of the historically negative correlation between equities and bonds. In the inflationary regime that began in 2022, both asset classes have been falling together. Bonds are no longer providing the protective cushioning that made the 60/40 portfolio an attractive construction for two decades. The second is what trend following has been doing while this has been unfolding: compounding strongly, sitting near high watermarks, and diverging sharply from both benchmarks during the periods of greatest stress.
Figure 3 makes this relationship explicit. In four out of five equity downturns over the 22-year period, the TTU Trend Following Index significantly outperformed. The tech bubble collapse, the GFC, and the inflationary regime of 2022 all produced strong trend following returns while equities deteriorated. The one exception was the short correction of 2018, which reversed quickly before trend following programs had sufficient time to exit prior positions and lock into new emerging trends. Short, sharp corrections are structurally unfavourable to trend following. Protracted regime shifts are where the edge lives.
Figure 4 extends the picture to individual programs. Ten trend following programs are shown alongside Warren Buffett’s Berkshire Hathaway. Most are sitting at or near their high watermarks, having compounded strongly through a period during which value investing and the major US equity indices were struggling.
Trend following does not only perform well during equity crises. Figure 5 shows outperformance in 2011, 2014/2015, and 2020/2021, periods when equity markets were themselves performing strongly. The ability to compound effectively across multiple regime types, not only during stress, is a core contributor to long-run geometric return outcomes. The correlation statistics on the TTU Trend Following Index confirm what the visual evidence suggests: the index is uncorrelated with both the S&P 500 Total Return Index and the 60/40 portfolio, with values approaching zero. This is expected given the extensive diversification that trend following programs deploy across numerous asset classes. Low long-run correlation does not mean the relationship is casual, however. Figures 3 and 5 together show that regime-specific causal relationships clearly exist. The low long-run figure simply reflects the fact that bursts of negative and positive co-movement average out over time.
Why Trend Following Behaves This Way
Two structural features, common to all diversified systematic trend following programs, explain this performance profile: simple price-following rules that cut losses short and let profits run, and extensive diversification across markets, systems, and timeframes.
Simple Rules, Asymmetric Outcomes
The logic is straightforward. Enter small, define the maximum loss with an initial stop, and trail that stop progressively in the direction of the trend as price moves favourably. Trade in both directions. Never use profit targets. The upside is structurally uncapped (Figure 6).
When markets offer protracted trending conditions, trend followers flourish. When they do not, trend followers stagnate or enter drawdown. Because each loss is cut short, it takes a sustained sequence of many small losses to produce a meaningful drawdown. When a favourable trending regime arrives, a small number of large winning trades can recover that ground and drive the equity curve to new highs.
The consequence of this structure, viewed across a large sample of trades, is a positively skewed return distribution: many small losses and small wins, with occasional very large wins that pull the distribution strongly to the right (Figure 7). Approximately 10% of trades account for the bulk of long-run performance. The vast majority of trades are noise. The Outliers are the signal.
This is the essence of Outlier Hunting. The fat-tail events that materialise during periods of market regime shift and uncertainty are the primary source of long-run wealth generation for trend following programs. Everything else is the cost of being in position when they arrive.
Diversification Across Markets, Systems, and Timeframes
Any individual market will produce only two or three genuine Outliers over a twenty-year history. Because Outliers are unpredictable in their timing and occurrence, concentrating in a small number of markets creates unacceptable exposure to long barren periods between events. Wide diversification across markets, systems, and timeframes increases the frequency of Outliers in the portfolio return stream and, critically, distributes them more evenly across the time series.
Figure 8 illustrates this. An ensemble of return streams from an uncorrelated portfolio shows each stream individually volatile and lumpy, with Outlier trades distributed widely across the series.
When consolidated into a portfolio, these distributed streams produce a composite that appears considerably smoother than any individual stream would suggest (Figure 9).
Figure 10 confirms the relationship between diversification and geometric return outcomes. Portfolios drawn from only five markets show wide dispersion in both CAGR and maximum drawdown. As diversification increases to twenty, then forty markets drawn from the same universe, dispersion contracts and the cluster of outcomes shifts toward higher CAGR at comparable drawdown levels.
This is not simply the standard correlation benefit of drawdowns being offset by drawups in an uncorrelated portfolio. A composite of positively skewed return streams, each hunting Outliers across different markets, increases the overall representation of large wins in the portfolio distribution. This raises CAGR through the mechanics of geometric compounding: variance drain is reduced, and the path of the equity curve becomes more favourable for long-run wealth accumulation.
Figure 11 demonstrates the practical application. A diversified suite of trend following systems deployed simultaneously against a single Outlier move in USDJPY, using short, medium, and long-term models, magnifies the impact of that Outlier in the aggregate return distribution.
Why Markets Produce Outliers
Understanding why this works requires understanding what Outliers are and why they exist. The answer lies in the structure of market return distributions.
When the returns of any liquid market are plotted over a long history, the distribution exhibits leptokurtic behaviour: fatter tails and a sharper peak near the mean than the normal distribution predicts (Figure 12). This is not a statistical curiosity. It is a structural feature of how markets function, and it directly determines where trading edges can be found.
A normal distribution implies that all price intervals are independent, with no serial correlation across time. Under those conditions there is no exploitable edge. The leptokurtic distribution tells a different story. It reveals three distinct zones of behaviour: clustering near the mean, and clustering in both tails, far from the mean.
Near the mean, markets exhibit convergent behaviour. Prices oscillate around an equilibrium, serial correlation is negative, and predictability is relatively high. Mean reversion strategies, pattern recognition, high-frequency trading, and value investment all depend on this regime. Central bank intervention, through buying dips and selling rallies, reinforces convergent conditions and can sustain them for extended periods, generating a large population of strategies designed to exploit stability and predictability. In a convergent regime, the multiplicity of competing approaches exerts a negative feedback on volatility, stabilising the market further. Economics here follows the law of diminishing returns: edges are identified and quickly competed away.
In the tails, the character of the market changes entirely. Here, markets exhibit divergent behaviour. Prices extend strongly in one direction, serial correlation turns positive, and return distributions take on pathological forms. Samples drawn from the tail regions follow distributions such as the Cauchy distribution, which have no stable mean and no well-defined standard deviation. This is an information-poor environment in the traditional sense. It is also the environment in which Outlier Hunters operate.
The frequency of these tail events is far higher than the normal distribution would predict. Figure 13 makes this concrete. Over 60 years of S&P 500 daily returns, events of five or more standard deviations occurred 43 times. Under a normal distribution, a single five-sigma event has a probability of approximately 1 in 3,488,555. Markets are not normally distributed at their extremes. They are a complex adaptive system (CAS), and the collective behaviour of their participants, Central Banks, investors, hedgers, traders, speculators, produces non-linear dynamics that the normal distribution cannot capture.
Figure 14 captures the spectrum of market regimes. The relative weighting of convergent and divergent participant behaviours determines where on that spectrum the market sits at any given time. Noisy markets, the dominant state, represent a balance between the two forces. Pure convergence and pure divergence are the extremes.
Trends exist across all regimes, but their character differs. In convergent regimes, trends are segments of broader mean-reverting cycles: directional, but not persistent, resolving back toward equilibrium. In a random price series, trend-like structures can emerge purely by chance with no serial correlation and no persistence whatsoever. Figure 15 illustrates this directly: real EURUSD data and randomly shuffled EURUSD data are visually indistinguishable. The structure of a trend tells you nothing about its persistence without understanding the serial correlation that underlies it.
In divergent regimes, trends display genuine directional persistence. Serially correlated clusters of price data are causally linked and drive momentum. This is where the Outlier Hunter operates.
How Divergence Feeds on Itself
As a market transitions from convergent to divergent, convergent traders find their models stop working. Strategies built for stability and predictability begin generating losses. The negative skew embedded in mean-reverting approaches is fully exposed. Some participants exit quickly. Others average down, adding to losing positions in an attempt to defer recognition of loss, compounding their exposure as the trend extends. Margin calls eventually force liquidation at the worst possible prices.
These behaviours do not dampen the trend. They reinforce it. Each wave of convergent participants capitulating into the divergent move adds order flow in the direction of the trend, feeding positive serial correlation and drawing in further momentum. This is path dependence operating at the portfolio level: prior price moves shape participant behaviour, and that behaviour in turn drives future price moves. It is a non-ergodic process. Transitions are sticky in their early stages and tend to accelerate as collective behaviour shifts.
The result is a mass transfer of wealth, from convergent participants whose models have failed to the small population of divergent participants whose models are designed for exactly this environment. Diversified systematic trend following is one of the very few approaches structured to capture this transfer systematically. Long optionality strategies, the outright purchase of calls and puts, are the other notable category.
The seeds of a divergent regime rarely remain confined to a single market. In a complex adaptive system with nested dependencies, regime shifts propagate across asset classes. A wave of divergence in one market creates pressure in adjacent markets, extending the opportunity set for Outlier Hunters deploying capital broadly. This is why the tech collapse, the GFC, and the inflationary regime of 2022 all produced widespread, multi-market trending conditions rather than isolated single-market moves.
All trends ultimately end. Extreme price moves cannot persist indefinitely. Central Bank intervention and the eventual exhaustion of directional order flow restore convergent conditions. But the tendency of markets to periodically abandon equilibrium and shift toward chaotic, directional behaviour is not an aberration. It is a structural feature of complex adaptive systems. It is why cutting losses short and letting profits run is not just a heuristic. It is the rational response to the environment in which systematic trend followers operate.