KellyBoost: Growth-Optimal Portfolio Construction with Gradient-Boosted Trees
2026-08-24 • Machine Learning
Machine Learning
AI summaryⓘ
The authors present KellyBoost, a machine learning model that directly predicts the best way to allocate money across different assets to maximize long-term growth. Unlike typical models that use simplified goals, KellyBoost uses an exact measure called the negative log growth rate to train. The authors also worked out the math needed to efficiently train this model and provided a basic version anyone can use. This approach helps in making portfolio decisions based on features while aiming for optimal growth.
XGBoostsoftmaxportfolio allocationnegative log growth rateKelly criteriongradientHessianmulti-output modelasset returnsmachine learning
Authors
Jiayu Li
Abstract
KellyBoost is a single multi-output XGBoost model whose softmax output is the portfolio: with y the vector of per-asset holding-period returns, the training loss is - log(1 + w y), the negative log growth rate, so the fitted model is the growth-optimal (Kelly) allocation conditioned on the features. The objective is exact rather than a surrogate: we derive the gradient, the analytic diagonal Hessian and the full Hessian in closed form, verify them by finite differences, and ship a dependency-free reference engine.