Multi-strategy Portfolio

Introduction

A multi-strategy portfolio combines more than one trading strategy and manages them as one capital system. The central idea is simple: the portfolio should be evaluated not only by how good each strategy is individually, but also by how the strategies interact.

Two strategies can both have positive expectancy and still create a poor combined portfolio if they lose at the same time. Conversely, strategies with different return behavior can make the overall equity path more diversified.

The professional objective is not to collect as many strategies as possible. It is to combine sufficiently independent sources of edge, allocate capital deliberately, and understand the portfolio-level consequences of correlation, drawdown, sizing and regime dependence.

1. Why Combine Multiple Strategies?

A single strategy may be exposed to a specific market behavior.

For example:• momentum may depend on persistent directional movement;• mean reversion may depend on temporary dislocations;• breakout systems may benefit from volatility expansion.

If their return patterns differ, combining them can reduce dependence on one source of edge.

But diversification is not guaranteed simply because the strategy names are different.

2. Strategy Diversity Is Not Strategy Count

Ten strategies can behave like one strategy if they all respond to the same underlying market driver.

Ask:• Do they trade the same instruments?• Do they use the same timeframe?• Do they depend on the same market direction?• Do they react to the same volatility regime?• Do they generate similar R-multiple sequences?

The number of strategies is a poor proxy for diversification.

3. Think in Strategy Families

Group strategies by their underlying behavior.

Possible families:• trend following;• momentum;• mean reversion;• breakout;• event-driven;• volatility-based;• market-neutral approaches.

Two strategies inside the same family may still be useful together, but their common exposure should be understood before capital is allocated.

4. The Common Unit: R-Multiple

R-multiples provide a practical language for comparing strategy returns.

If Strategy A produces +2R and Strategy B produces −1R, the combined result can be expressed in risk units after accounting for their allocated risk.

This makes it easier to compare strategies with different instruments, position sizes and nominal capital requirements.

5. Strategy Correlation

Measure the relationship between strategy returns, preferably using comparable units such as daily returns or R-multiples.

Positive correlation means strategies tend to move in the same direction.Negative correlation means they tend to move in opposite directions.Near-zero correlation means little linear relationship in the measured sample.

Correlation is not permanent and does not guarantee diversification.

6. Correlation Can Change in Stress

Strategies that appear diversified in normal conditions can become highly synchronized during market stress.

A portfolio review should therefore examine:• normal-period correlation;• rolling correlation;• stress-period correlation;• simultaneous loss clusters.

The question is not only “How correlated are these strategies on average?” but also “What happens when the portfolio is under pressure?”

7. Portfolio Expectancy

If strategies are allocated with different capital or risk weights, their individual expectancy does not translate directly into portfolio expectancy without considering those weights.

Suppose:Strategy A: +0.25R expectancy.Strategy B: +0.15R expectancy.

If A receives 80% of the portfolio risk and B 20%, the contribution of each strategy must be evaluated relative to its allocation.

Portfolio-level analysis therefore starts with the allocation rule, not just the strategy statistics.

8. Equal Capital vs. Equal Risk

Allocating the same dollar amount to each strategy does not create equal risk.

A volatile strategy can contribute much more to portfolio variability than a stable one.

Possible allocation frameworks include:• equal capital;• equal risk;• volatility-adjusted;• risk-budgeted;• strategy-edge-weighted approaches.

The choice should be defined by the portfolio objective and tested rather than assumed to be universally optimal.

9. Capital Allocation and Risk Budget

A portfolio needs an explicit risk budget.

For example, the researcher may define:• maximum risk per strategy;• maximum combined risk;• maximum risk for a strategy family;• maximum correlated exposure;• portfolio drawdown threshold.

This turns capital allocation into a controlled system rather than a collection of independent trades.

10. Concentration Hidden Inside Diversification

A portfolio can contain many strategies and still be concentrated.

Example:Five different momentum systems all trade technology stocks during bullish conditions.

The strategies are technically separate, but their exposure to the same market driver can create synchronized losses.

Look beyond strategy labels and examine instruments, sectors, direction, volatility and event exposure.

11. Drawdown at Portfolio Level

Individual strategy drawdowns do not simply add together in every portfolio.

If losses are weakly related, the combined path may be smoother.If losses are synchronized, portfolio drawdown can become much larger.

Measure:• individual maximum drawdown;• portfolio maximum drawdown;• drawdown duration;• simultaneous drawdowns;• recovery behavior.

The portfolio equity curve is its own object of analysis.

12. Risk of Ruin and Multiple Strategies

Adding strategies does not automatically reduce Risk of Ruin.

If the strategies are strongly correlated, combined losses can still be severe. If they are genuinely differentiated, the distribution of portfolio outcomes may improve.

Ruin analysis should therefore use the portfolio's actual risk structure, not simply the number of strategies.

13. Position Sizing Across Strategies

Position sizing must be coordinated.

If three strategies each risk 2% per trade, a simultaneous signal can create 6% gross risk before correlation and common exposure are considered.

Possible controls:• strategy-level risk caps;• portfolio-level risk caps;• correlated-cluster limits;• total open-risk limits.

The exact limits are design choices, but the principle is universal: individual sizing decisions interact.

14. Strategy Capacity and Trade Overlap

A multi-strategy portfolio may produce overlapping trades.

This matters because:• capital may be committed simultaneously;• liquidity requirements increase;• correlated positions can accumulate;• execution costs can rise.

Research should measure not only how often strategies trade, but how often their positions overlap and how much capital is required at those moments.

15. Regime Dependence at Portfolio Level

A portfolio of regime-dependent strategies can unintentionally become a regime bet.

Example:Strategy A performs in high volatility.Strategy B also performs in high volatility.Strategy C has little exposure to low volatility.

The portfolio may look diversified by strategy count while remaining highly dependent on one market environment.

Portfolio research should therefore map strategy performance against common regimes.

16. Multi-Strategy Optimization

Optimizing the allocation between strategies can be useful, but it introduces another research layer.

Testing thousands of allocation combinations can overfit the historical strategy-return matrix.

A disciplined approach:• define a small set of allocation rules;• use development data;• measure portfolio distribution and drawdown;• examine allocation stability;• freeze the allocation method;• validate out of sample.

The allocation itself is a model and can be overfit.

17. Case Study — Four Strategies, One Risk

A trader combines four strategies:

A: trend following.B: momentum.C: breakout.D: short-term reversal.

At first glance the portfolio looks diversified.

Further analysis shows A, B and C all perform poorly during sideways markets and hold similar long exposures. Their losses cluster.

D behaves differently and offsets some of the drawdown.

The portfolio is therefore not four independent edges. It is closer to a directional/trend cluster plus one diversifying strategy.

18. Case Study — Different Edges, Better Interaction

Another portfolio contains three strategies with similar individual expectancy:

A: +0.20R.B: +0.18R.C: +0.17R.

Their R-multiple correlations are low over the relevant sample, and their largest drawdowns occur at different times.

The combined portfolio shows lower drawdown than a concentrated allocation to any single strategy, although the result is not guaranteed to persist.

The important observation is not simply that there are three strategies. It is that their return paths interact differently.

19. Professional Multi-Strategy Workflow

A practical workflow:

1. Define the role of each strategy.2. Verify that each has independent research evidence.3. Standardize performance in R or another comparable unit.4. Measure strategy correlation.5. Examine stress and rolling correlation.6. Map common instrument and regime exposure.7. Define the portfolio risk budget.8. Test allocation rules.9. Analyze combined expectancy and distribution.10. Analyze portfolio drawdown.11. Test trade overlap and capacity.12. Freeze the allocation method.13. Validate the portfolio out of sample.14. Monitor whether the diversification relationship persists.

20. Common Mistakes

• Assuming different strategy names mean diversification.• Allocating equal dollars without considering risk.• Ignoring strategy correlation.• Looking only at average correlation.• Ignoring stress correlation.• Combining strategies before validating each individually.• Optimizing allocation on the same data repeatedly.• Ignoring simultaneous open risk.• Ignoring common instruments and sectors.• Assuming diversification permanently reduces drawdown.• Treating the number of strategies as a risk measure.

21. Expert Exercise — Build a Strategy Interaction Matrix

Select three to five strategies.

Create a matrix containing:• expectancy;• average R;• maximum drawdown;• trade count;• R-multiple correlation;• major market regimes;• common instruments;• common directional exposure.

Then define a hypothetical allocation and calculate the combined risk contribution.

Do not ask first which strategy is “best.” Ask:“Which combination produces a portfolio behavior consistent with my risk budget and research objectives?”

22. Knowledge Check

1. Why is strategy count not the same as diversification?2. What makes strategy correlation different from instrument correlation?3. Why can average correlation hide portfolio risk?4. Why is R useful for comparing strategies?5. Why is equal capital not necessarily equal risk?6. What is a portfolio risk budget?7. How can several strategies create hidden concentration?8. Why does trade overlap matter?9. How can regime dependence create a portfolio-level bet?10. Why can allocation optimization itself become overfit?

Key idea: a multi-strategy portfolio is not a pile of independent systems. It is one integrated risk and capital system whose behavior emerges from the interaction of its strategies.

Copying, reproducing, adapting, distributing, or publicly displaying any part or all of the Academy’s textual content without the author’s prior written consent is prohibited.