Handcrafted Portfolio Construction with Hierarchical Grouping and Risk Targets
Summary
The document outlines a discretionary, spreadsheet-friendly approach to constructing long-only portfolios or allocating among trading strategies. It emphasizes choosing inputs that are easier to estimate and interpret, especially Sharpe ratios and volatility, while recognizing that expected returns, volatilities, and correlations are uncertain. It compares equal weighting with inverse-volatility weighting and describes scaling a portfolio to a risk target, using leverage when available or adding cash when the target is below the portfolio’s natural risk.
A central method is hierarchical grouping: allocate risk or capital across broad groups, then divide each group among its subgroups and assets. Examples show how nested allocations translate into final weights, and the document suggests organizing assets by categories such as regions, sectors, bond types, commodities, currencies, or trading rules. It also raises the need to account for diversification, correlations, constraints, and Sharpe ratios. The outline promises later backtesting, but this excerpt gives no empirical performance evidence; its long-only scope makes it unsuitable for long-short portfolios, and group choices remain consequential.
Key ideas
- The method is intended for long-only assets or trading strategies and is designed to be implemented in a spreadsheet.
- It favors intuitive portfolio inputs such as Sharpe ratios and volatility while acknowledging that all estimates are uncertain.
- Inverse-volatility weighting converts risk allocations into capital weights, with cash or leverage used to reach feasible risk targets.
- Hierarchical grouping allocates weight at multiple levels, making the resulting asset weights depend on the chosen structure.
- Diversification, correlations, constraints, and expected Sharpe ratios may require additional adjustments.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.