All Days Are Not Created Equal: Understanding Momentum by Learning to Weight Past Returns
A new study introduces the Characteristic-Managed Momentum (CMM) strategy, which uses machine learning to weight past stock returns based on their importance. CMM outperforms traditional momentum investing, especially during market crises and recent years. The research finds that only a few days, like those with earnings announcements or large market moves, predict future returns. CMM’s success is linked to investor underreaction to news, not risk, though its complexity may limit practical use.
What it examines
This study introduces a new momentum investing strategy that uses machine learning to flexibly weight past stock returns, instead of treating all days equally. The goal is to improve momentum profits and understand why momentum works, by analyzing which past returns are most informative for predicting future stock performance.
What it concludes
The new strategy outperforms traditional momentum, especially by focusing on key days like earnings announcements and large price moves. It works well even after costs and in global markets. This approach can help investors build better trading strategies and offers insights for future research on market behavior and asset pricing.
Evidence objects
A new Characteristic-Managed Momentum (CMM) strategy uses machine learning to flexibly weight past stock returns, dramatically outperforming traditional momentum investing, especially during market crises and recent decades of poor momentum performance.
key_findings bullet 1 · key_findings · validation V0
Surprisingly, the study finds that only a handful of dayssuch as earnings announcements, market-wide jumps, or unusually large returnshold most predictive power for future stock performance, challenging conventional wisdom about momentum signals.
key_findings bullet 2 · key_findings · validation V0
CMMs success is mainly driven by investors underreaction to important news, not risk factors. Despite relying on complex neural networks, the approach remains highly profitable after transaction costs, though transparency may limit practical adoption.
key_findings bullet 3 · key_findings · validation V0
This paper introduces a novel machine learning-based 'characteristic-managed momentum' (CMM) strategy, which flexibly weights past returns to outperform traditional momentum. Its originality lies in data-driven, conditional weighting, revealing that only select daysoften with large returns or newsmatter most. Robust empirical results offer compelling insights for both academic research and trading applications.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- … We do so by setting up a simple machine learning framework that extracts these weights from the formation period returns themselves and firm characteristics that serve as …
Source row: 159 · abstract type: snippet