Hierarchical Risk Parity for Robust Multi-Asset Portfolio Allocation
Summary
The document explains Hierarchical Risk Parity (HRP), a portfolio allocation method that uses asset similarity and risk rather than directly optimizing expected returns. It outlines the process: estimate returns, covariance and correlation; cluster assets by correlation distance; reorder them according to the resulting tree; then recursively divide capital between clusters in proportion to their risk. This lets related instruments share a risk budget without inverting the covariance matrix.
The article describes an MQL5 implementation and says each pipeline stage is checked against an independent Python reference. It also presents a rebalancing Expert Advisor that aligns prices across instruments and sizes positions by account currency. Its comparison with mean-variance optimization argues that HRP can produce stable, long-only, fully invested allocations when correlated inputs make optimizer weights unstable. HRP does not forecast returns or guarantee lower variance, and the article’s examples and implementation do not establish that it will outperform in live trading.
Key ideas
- HRP groups instruments by correlation distance before allocating capital.
- Quasi-diagonalization orders similar instruments next to one another in the cluster tree.
- Recursive bisection gives a larger share to the lower-risk side of each split.
- Avoiding covariance matrix inversion reduces sensitivity to noisy correlation estimates.
- HRP allocates risk but does not predict returns or ensure lower variance than Markowitz optimization.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.