J.P. Morgan Macrosynergy Quantamental System (JPMaQS): Portfolio optimization with macro factors and neural networks

Research evidence excerpt

J.P. Morgan Macrosynergy Quantamental System (JPMaQS): Portfolio optimization with macro factors and neural networks

Products & Services J P M O R G A NJPMaQS

27 July 2026

This article shows a practical method for optimizing equity portfolios with point-in-time macroeconomic information and

sequential statistical learning. The learning process relies on neural networks, as they learn portfolio weights directly from a full

historical panel of macroeconomic divergence factors and return data. They do not require stock-by-stock theoretical priors for

model construction. The signals of various candidate networks are weighted during the statistical learning process based on their

historical out-of-sample performance.

In this post, the network-based learning method is applied to portfolio allocation across U.S. equity sectors. Even in its most

restrictive version, the allocation generates notable economic value-added without any hindsight. When given full freedom to

adjust sector weights and the option to also invest in a safe-haven Treasury bond, macro factor-based learning doubles long-term

wealth generation. The method also produced profitable signals for a long-short equity sector strategy.

Full post and python code here