Hierarchical Risk Parity (HRP)
Hierarchical Risk Parity is a portfolio construction method that groups assets by similarity using machine learning, then allocates risk across the resulting hierarchy without inverting an unstable covariance matrix.
Hierarchical Risk Parity, introduced by Marcos Lopez de Prado, is an allocation technique that builds diversified portfolios in three steps. First, it uses hierarchical clustering on the correlation matrix to group assets that behave similarly into a tree structure. Second, it reorders the covariance matrix so related assets sit next to each other (quasi-diagonalization). Third, it walks down the tree and recursively splits capital between clusters in inverse proportion to their risk. Crucially, it never inverts the covariance matrix, which is the step that makes traditional optimization so fragile.
HRP matters as a direct response to the well-known instability of classical mean-variance optimization. Markowitz-style optimizers require inverting a covariance matrix, which is notoriously sensitive to small estimation errors and tends to produce concentrated, extreme allocations that perform poorly out of sample. By relying on the hierarchy of relationships instead of matrix inversion, HRP produces allocations that are more stable, more diversified, and more robust when the input estimates are imperfect, which they always are in practice.
HRP sits at the intersection of machine learning and portfolio theory, applying clustering algorithms to a problem long dominated by quadratic optimization. It does not guarantee the theoretically optimal portfolio under perfect inputs, but it typically delivers better realized risk-adjusted performance precisely because real-world inputs are noisy. It has become a respected tool in the quantitative allocation toolkit, especially for portfolios with many correlated assets.
Hedgewing.ai includes HRP among its institutional risk analytics, alongside Sharpe and Sortino ratios, Value at Risk, and Fama-French factor analysis. It is a natural fit for a machine-learning-first platform: just as the four-model deep-learning ensemble emphasizes robustness and out-of-sample validation through nightly walk-forward backtesting, HRP brings that same robustness-over-fragility philosophy to how risk is allocated across a portfolio, favoring stable, diversified weights over the brittle precision of classical optimizers.
Related terms
Diversification · Value at Risk (VaR) · Sharpe Ratio · Correlation
Back to the hedgewing.ai glossary · See AI stock forecasts