When trend followers discuss diversification, the conversation typically centres on markets: the breadth of instruments traded, the range of asset classes represented, and the geographic spread of the portfolio. This is the right conversation to have. But it is incomplete. System diversification, the practice of deploying multiple trend-following models rather than a single one, is equally consequential and less frequently examined with the rigour it deserves.
Why a Single System Is a Concentrated Bet
A trend follower who allocates 100% of their capital to a single trend-following system within a market is making a concentrated bet on one specific interpretation of how trends form and develop in that market. The system has a particular lookback period, a particular entry trigger, a particular exit mechanism, and a particular sensitivity to trend duration and pattern. When market conditions align with those characteristics, the system performs. When they do not, it underperforms or produces losses.
The selection of that single system introduces a further problem: selection bias. A system chosen on the basis of its historical backtested performance is, by construction, the system that fitted the historical data most closely. This is the same as saying it is the system most likely to be overoptimised to past conditions and least likely to generalise robustly to future ones. The model that won the backtesting competition is not necessarily the model best suited to the markets as they will behave going forward. It is the model best suited to the markets as they have already behaved.
The Ensemble: Offsetting Correlations and Reducing Selection Bias
Deploying an ensemble of ten distinct trend-following models, each allocated 10% of capital within the same market, produces ten independent return streams. Each stream reflects the unique way its corresponding model interacts with the market’s trending behaviour. The models share the core discipline of trend following, cutting losses short and letting profits run, but differ in their sensitivity to trend duration, their entry and exit criteria, and their responsiveness to different trend patterns.
The aggregate effect of combining these streams is a composite return with inherent correlation offsets. When one model is in drawdown because the market is not producing the trend duration it is calibrated to capture, other models with different characteristics may be performing. The aggregate volatility of the ensemble is lower than the volatility of any individual system within it, not because the individual systems have been made less volatile, but because their correlation structure, driven by their different design characteristics, produces a smoother aggregate outcome.
The ensemble also resolves the selection bias problem structurally. No single model in the ensemble is selected on the basis of its historical superiority. All models share the same foundational edge, and their collective performance is less dependent on any one model’s fit to past conditions. The ensemble is more adaptable to a range of future market environments precisely because it is not optimised to any single past environment.
Direct Control Over Portfolio Correlations
One of the most practically significant properties of system diversification is that it gives the Outlier Hunter direct control over portfolio correlations in a way that market diversification alone does not. Market correlations are largely exogenous: they reflect the structural relationships between asset classes and the behaviour of participants across those markets. A trend follower cannot meaningfully alter the correlation between crude oil and Brent oil by choosing which markets to trade. The correlation is a property of the markets themselves.
System correlations are endogenous. They are a product of the design and configuration of the systems being deployed. By designing systems with different entry criteria, different holding periods, and different sensitivities to trend characteristics, and by staggering their activation so that they operate independently of one another at different times, the portfolio’s correlation structure can be deliberately engineered rather than simply accepted.
The practical consequence is visible in highly correlated markets. When an ensemble of trend-following systems is deployed across two markets with high intrinsic correlation, such as Brent crude and WTI crude, the return streams produced by the ensemble across those markets exhibit materially lower correlation than the markets themselves. The system design has effectively decomposed the market correlation, allowing the portfolio to treat two highly correlated markets as more independent sources of return than they would appear to be on the basis of their price relationship alone. This expands the universe of markets that can be meaningfully added to the portfolio, increasing diversification beyond what market correlations would otherwise permit.
The Staggered Ensemble and Outlier Capture
The staggered activation structure of the ensemble has a specific relationship to Outlier capture that goes beyond correlation management. Each system within the ensemble activates based on its own entry criteria. When a trend begins to develop, some systems will trigger early, others will trigger as the trend extends and confirms, and others will trigger later still as the trend reaches full momentum. The result is that capital is not committed all at once at the onset of a potential trend. It is deployed progressively, as the trend demonstrates material development and strength.
This mirrors the logic of pyramiding: adding to a winning position as it continues to prove itself, leveraging the trend’s developing momentum rather than committing maximum exposure before the trend has established its character. Each system that activates adds to the position at a point where the trend has already provided some confirmation. The aggregate position grows with the trend rather than being fixed at the outset.
The risk management architecture of the ensemble reinforces this. Each system within the ensemble is designed to strictly control adverse risk and protect realized capital. This discipline creates the conditions under which unrealized equity, the profits accrued on open positions as the trend develops, can be deployed more aggressively into the extending move. The realized capital is protected. The unrealized equity is the ammunition for the Outlier capture. The ensemble structure allows both objectives to be pursued simultaneously rather than trading one off against the other.
When market conditions align favourably across an ensemble of well-designed systems, the staggered activation and progressive position building can produce Outlier trades of significant magnitude. The collective strength of multiple systems, each capturing the trend from a slightly different angle and at a slightly different stage of its development, multiplies the performance potential that any single system operating alone would produce.
System Diversification as Structural Architecture
The case for system diversification is not that it improves performance in every period. In strongly trending markets where a single well-chosen system performs exceptionally, the ensemble will produce a lower peak return than that single system at its best. The case for system diversification is structural: it produces more robust geometric return outcomes across the full range of market conditions, including the conditions that any single system is poorly suited to navigate.
A single system is a single bet on a single interpretation of how markets trend. An ensemble of systems is a portfolio of interpretations, each with genuine edge, collectively less vulnerable to the specific market conditions that any one interpretation handles poorly. The ensemble’s reduced volatility, lower selection bias exposure, engineered correlation structure, and staggered Outlier capture architecture make it better suited to the actual distributional properties of financial markets than any single-system approach can be.
For a more detailed technical exploration of system diversification in trend following, including the mechanics of correlation management and ensemble construction, the full analysis is available at Aussie Turtles: https://www.aussieturtles.com/enhancing-trend-following-performance-using-system-diversification/