Research / Theme
Portfolio Construction as a System
Portfolio construction is often treated as an optimization problem — weights, correlations, expected returns. In practice, it is a systems problem shaped by assumptions, constraints, time horizons, and governance.
Decisions made at the portfolio level interact across asset classes, managers, liquidity, and risk tolerance. A change in one component frequently alters the behavior of the whole system in ways that are not immediately visible. Correlations shift under stress. Factor exposures cluster invisibly. Liquidity assumptions prove fragile precisely when they need to hold.
Many portfolios fail not because individual allocations are incorrect, but because interactions between components are poorly understood. The problem is not the inputs — it is the architecture.
This research theme examines portfolio construction as an integrated discipline rather than a sequence of independent decisions. It emphasizes structural resilience, second-order effects, and the alignment of portfolio design with long-term objectives and fiduciary constraints.
MARY's Portfolio Construction agent uses Black-Litterman optimization with J.P. Morgan capital market assumptions as the quantitative foundation — then applies Bayesian blending to incorporate regime tilts, quality signals, and momentum overlays. The tactical overlay weights regime 40%, valuation 30%, momentum 30%. Stock-level quality scores from the Fundamental agent feed directly into security selection. Every portfolio is stress-tested across macro scenarios before a recommendation is made. Construction is treated as a system, not a spreadsheet.