Systematic Portfolios: Diversification That Actually Works
Ten strategies deployed in parallel constitute diversification only if they fail to lose capital in the same week.
Macro Context: A Correlation Coefficient Describes a Period, Not a Property
A correlation coefficient calculated over a specific historical period describes how strategies behaved during that period — it does not describe an intrinsic, permanent property of the strategies themselves. Two strategies can show near-zero correlation across years of stable conditions and converge sharply the moment a shared underlying dependency, such as trending volatility or ample liquidity, reverses.
The Structural Challenge: Diversification Hiding a Shared Dependency
A portfolio of systematic strategies can appear diversified on paper, spanning different asset classes and time horizons, while sharing a hidden dependency on the same underlying market condition. When that condition reverses, strategies that appeared uncorrelated in backtesting move together in live trading, precisely when diversification is needed most.
The Methodology: Testing Under Stress, Examining Signal Logic Directly
Genuine independence requires testing strategies against each other under stressed as well as historical market conditions, and examining the underlying signal logic directly, independent of a correlation coefficient calculated over a period that may not include the regime that ultimately breaks the portfolio.
position sizing compounds the exposure when set by conviction alone, disconnected from risk contribution: a strategy sized aggressively because it has performed well recently often proves to be the strategy most correlated with the others in the portfolio, concentrating precisely the risk the allocation was intended to spread.
The Deterministic Outcome
A portfolio evaluated on underlying signal logic and stress-tested correlation, with positions sized by risk contribution rather than conviction, holds its intended diversification under the exact conditions a purely historical correlation analysis would have missed.
Strategic Takeaways
- Test strategy correlation under stressed market conditions, extending beyond the historical backtest period
- Examine the underlying signal logic directly, alongside the calculated correlation coefficient — not as a substitute for it
- Size positions by risk contribution to the portfolio, not by recent performance or conviction alone
Discuss this with our team.
Tell us what you are working through, and we will route you to the right partner.
Backtesting Discipline: The Easiest Number to Fake
The determining question is when the strategy logic was fixed relative to the performance period being shown.
Algorithmic Trading Compliance Amid Shifting Markets
A compliance framework calibrated to the prior market cycle governs a market structure that has already moved on.
Risk Budgeting Across a Multi-Strategy Systematic Portfolio
Equal-capital allocation and equal-risk allocation are structurally distinct exercises, and that distinction is where most portfolios quietly go wrong.