Boundary-Induced Apparent Risk Aversion in Nonergodic Multiplicative Growth

Battle Count: 71

The paper is highly relevant to position sizing and risk management in quantitative trading. It provides a rigorous mechanism showing that the Kelly fraction is suboptimal near ruin/liquidation boundaries, directly applicable to leveraged trading, margin-constrained portfolios, and drawdown management. The boundary-induced compression and local reversal findings inform how traders should adjust exposure based on proximity to stop-loss or margin-call thresholds. However, the model is minimal (binary returns, fixed exposure, single agent) and does not directly address multi-asset portfolios, transaction costs, or market microstructure. The CRRA mapping critique is relevant for calibrating risk models from observed trading behavior.

The core lattice propagation algorithm is straightforward: a recombining binary tree with probability mass transfer to an absorbed state, scaling polynomially in T. The numerical optimization over an exposure grid is simple. However, careful handling of boundary conditions, grid resolution, tie-breaking conventions, and diagnostic refinements (high-rho reversal, horizon sensitivity) adds moderate complexity. No specialized ML frameworks or large-scale computation are needed. The main intellectual complexity lies in the model formulation and interpretation rather than in code implementation.

The paper provides a detailed benchmark parameterization (L=100, S=10, rho=0.1, T=50, p=0.55, a=b=1), grid specifications (601 exposures on [0,0.3], 80 distances on [0.05,3.0]), and exact lattice recursion formulas. A reproducibility package with numerical data and custom code is stated as available upon request to the corresponding author. However, no public repository link is provided, and the code is not openly hosted.