Solving Kelly Ruin via Wasserstein DRO and Convex Optimization (github.com)

A new framework for managing risk in predictive modeling, developed by Quant Research Labs, aims to tackle epistemic uncertainty through robust optimization techniques. It utilizes Wasserstein Distributionally Robust Optimization (DRO) to create an ambiguity set around predicted probabilities, transforming the traditional Kelly Criterion approach. Instead of focusing on a fixed distribution, this innovative method addresses potential worst-case scenarios, maximizing log-growth while incorporating Wasserstein-2 constraints to ensure stability in uncertain environments. The framework is modularized into seven Python components, including a Stochastic Engine leveraging Hierarchical Bayesian Inference and an Optimization Engine using cvxpy and the Splitting Conic Solver (SCS). It employs advanced techniques such as isotonic regression for calibration and a walk-forward validation process to mitigate biases. With a complete suite provided in the Commercial Asset Kit, this resource not only enhances predictive accuracy but also safeguards against "optimal ruin," making it a significant tool for the AI/ML community focused on financial applications and risk management. However, users are cautioned that past performance does not guarantee future results, as the tool is provided "as is."