This work presents a novel application of the Stochastic Dual Dynamic Problem (SDDP) to large-scale asset allocation. We construct a model that delivers allocation policies based on how the portfolio performs with respect to user-defined (synthetic) indexes, and implement it in a SDDP open-source package. Based on US economic cycles and ETF data, we generate Markovian regime-dependent returns to solve an instance of multiple assets and 28 time periods. Results show our solution outperforms its benchmark, in both profitability and tracking error.

中文翻译：

---

需要实现投资目标？使用 SDDP 方法跟踪合成索引

这项工作提出了随机对偶动态问题 (SDDP) 在大规模资产配置中的新应用。我们构建了一个模型，该模型根据投资组合在用户定义（合成）指数方面的表现来提供分配策略，并在 SDDP 开源包中实现它。基于美国经济周期和 ETF 数据，我们生成依赖于马尔可夫机制的回报来解决多个资产和 28 个时间段的实例。结果表明，我们的解决方案在盈利能力和跟踪误差方面均优于其基准。