Multi-System Portfolios and Correlation

One system is one point of failure. A portfolio of uncorrelated systems is a business.

Why trade multiple systems

A single system has one edge, one regime, one set of weaknesses. When its edge decays, you stop earning. A portfolio of systems with low correlation smooths returns and reduces drawdown — when one system is in drawdown, others can carry the account.

Diversification across systems is the trading equivalent of not putting all eggs in one basket. The math is identical to portfolio theory in finance.

The correlation principle

Two systems with correlation ρ:

• ρ = 1.0: identical — no diversification benefit
• ρ = 0: uncorrelated — variance drops ~30% with two equal systems
• ρ = -1.0: opposite — variance can drop dramatically (rare and difficult to engineer)

The lower the correlation between systems, the smoother the combined equity curve.

Sources of diversification

Dimension	Example
Strategy type	Trend-following + mean reversion
Timeframe	H1 swing + D1 position
Asset class	Forex + futures + equities
Direction bias	Long-bias + short-bias systems
Regime bias	Volatility-expansion + low-vol range
Geography	US equities + Asian FX
Setup type	Breakout + reversal

A portfolio combining trend-following on commodities + mean reversion on forex + breakout on equity indices tends to have low correlation across its components.

Measuring correlation between systems

Compute the monthly or weekly return series for each system. Calculate pairwise correlation:

ρ(A, B) = Cov(RA, RB) / (σA × σB)

Where RA, RB are the return series.

Correlation	Action
0.8–1.0	Don't combine — duplicates risk
0.5–0.8	Combine only if setups are very different
0.2–0.5	Good diversification
-0.2 to 0.2	Excellent diversification
< -0.2	Excellent — possibly engineered

Build a correlation matrix for all candidate systems before allocating.

Capital allocation strategies

Equal weighting

Allocate the same risk budget to each system. Simple, robust, ignores performance differences.

Volatility weighting

Allocate inversely to each system's volatility. A system with half the volatility gets twice the allocation. Equalizes risk contribution.

Sharpe weighting

Weight by Sharpe ratio — better risk-adjusted systems get more capital.

Risk parity

Each system contributes equal risk to the portfolio (variance contribution, not capital). Sophisticated; better for larger portfolios.

For beginners, equal weighting is sufficient. Add complexity only when the portfolio grows beyond 4–5 systems.

The benefit in numbers

Consider two systems:

• A: 15% annual return, 12% MDD
• B: 15% annual return, 12% MDD
• Correlation between A and B: 0.3

Combined (50/50):

• Annual return: 15% (unchanged)
• Expected MDD: ~8% (lower than either alone)

Return unchanged, drawdown reduced by a third. This is the free lunch of diversification.

Common pitfalls

1. Hidden correlation

Two systems look different on the surface but trade the same edge. A "moving-average crossover" and a "Donchian breakout" are highly correlated despite different mechanics — both are trend-following.

Fix: compute actual return correlation, don't assume.

2. Correlation increases in crisis

Systems that look uncorrelated in normal markets often correlate in crashes ("all correlations go to 1"). A portfolio that survived the backtest may collapse in a 2020-style event.

Fix: stress-test with rolling correlation during crisis periods. Don't allocate based on full-sample correlation alone.

3. Over-diversification

Adding systems beyond 5–7 produces diminishing returns and complexity. Each additional system requires monitoring, journaling, and capital.

Fix: stop adding systems once marginal diversification benefit drops below 10% drawdown reduction.

4. Correlated instruments across systems

A trend system on gold and a breakout system on gold miners are highly correlated — both depend on the same underlying price.

Fix: diversify across asset classes, not just strategies.

5. Changing correlations

Correlations drift over time. A system that was uncorrelated last year may be correlated today.

Fix: recompute correlation quarterly. Re-balance if drift is significant.

Practical portfolio construction

1. Backtest 4–6 candidate systems independently
2. Compute pairwise correlation of monthly returns
3. Eliminate candidates with >0.7 correlation to an existing system
4. Equal-weight the survivors initially
5. Run a portfolio backtest — does combined MDD improve?
6. Run Monte Carlo on the combined trade list
7. Allocate to live trading with reduced size, validate forward

Monitoring the portfolio

Monthly:

• Compute each system's contribution to return
• Compute each system's contribution to drawdown
• Check correlation drift vs. baseline

Quarterly:

• Re-weight if risk contribution has shifted significantly
• Add or retire systems based on failure criteria

The psychological benefit

Multiple systems reduce the emotional swings of any single system's drawdown. When your trend system is down 15%, your mean-reversion system being up 8% keeps you rational. This is why even discretionary traders benefit from running a few systematic strategies alongside their discretionary trades.

---

Next: turn your rules into automated code — the technical path from manual to algorithmic trading.