Topological Risk Parity

Battle Count: 73

Highly relevant to quantitative trading. The paper directly addresses portfolio construction for long/short, market-neutral strategies, which are core to quantitative trading desks. It provides a practical alternative to HRP that preserves signal direction, supports L/S portfolios natively, and allows explicit control over sector/factor exposures. The framework is designed for equity stat-arb and L/S trend strategies where enforcing neutrality at market, sector, or factor level is critical. The method avoids covariance inversion and is computationally simple, making it suitable for live trading workflows.

Moderate complexity. The core pipeline is straightforward: compute correlation matrix, derive Mantegna distance, construct MST, choose root, compute topological factors via tree traversal, multiply signals by factors, and L1 normalize. The Semi-Supervised TRP variant adds complexity with dummy root insertion, augmented graph construction, and DFS spanning tree extraction. Post-processing (clipping, thresholding, sector neutralization) adds further steps. No covariance inversion or recursive cluster variance estimation is needed, making it simpler than HRP. Main implementation challenges include handling MST ties, ensuring deterministic DFS ordering, and managing edge cases (no sector ETFs present, inactive assets).

The paper provides a complete mathematical description of the framework with exact formulas, propositions, and proofs. However, no code repository, empirical backtesting results, or specific dataset references are provided in the extract. The algorithmic steps are clearly defined (Steps 1-6 for Variant I, Steps 1-2 for Variant II), making theoretical reproduction feasible, but practical implementation details (e.g., specific MST library, DFS ordering) are not fully specified.