Unlocking Hidden Patterns
How Daily Returns Predict Future Stock Performance
Nusret Cakici, Christian Fieberg, Gabor Neszveda, Robert Bianchi, and Adam Zaremba, authors of the January 2026 study “A Unified Framework for Anomalies based on Daily Returns,” challenged how we think about short-term return patterns in stock markets. Their research reveals that the wealth of information contained in daily stock returns has been hiding in plain sight—and when properly extracted, it generates remarkable predictive power for future performance.
What the Researchers Examined
The academic literature is filled with anomalies that attempt to predict stock returns using recent daily price movements. Some focus on when returns occurred (like short-term reversal strategies), while others emphasize how extreme they were (like the MAX effect, which looks at maximum daily returns). But here’s the puzzle: all these strategies draw from the same raw material—the sequence of daily returns over the past month—yet each isolates just one specific aspect.
The authors asked a more fundamental question: What if we let the data tell us how to weight and combine information from recent daily returns, rather than imposing arbitrary functional forms?
Using nearly a century of U.S. stock data (1937-2024), they employed machine learning techniques (elastic net regression) to systematically extract two core dimensions from the past month’s daily returns:
Chronological information: The time-ordered sequence of daily returns, capturing when returns occurred within the month.
Rank information: The magnitude-ordered returns, capturing how extreme each daily outcome was relative to others in that month.
From these components, they constructed the Daily Return Information (DRI) signal and its corresponding factor portfolio, DRIF (Daily Return Information Factor).
Key Findings
You can read the rest of my Alpha Architect article here.