Tripling the predictive accuracy of a financial AI model can be done by simply changing the target from raw returns to rank-based returns.
April 26, 2026
Original Paper
Getting the Target Right in Return Prediction
SSRN · 6615698
The Takeaway
The biggest bottleneck in stock market AI is not the complexity of the neural network or the amount of data used. It is the overlooked choice of how the target variable is defined during training. Switching to ranks instead of raw numbers doubles the returns for an investment portfolio. This simple shift helps the model ignore market-wide noise and focus on the relative strength of different stocks. It proves that basic data engineering is often more powerful than the latest model architectures.
From the abstract
We show that a largely overlooked design choice-how returns are defined as prediction targets-drives machine learning performance in stock returns. Using a large international panel of equities from 35 markets over 1994 to 2024, we compare transformations applied to stock characteristics and target returns. Transforming the target from raw to standardized or rank-based returns nearly triples predictive accuracy and doubles portfolio returns. Feature transformations play a secondary role. Rank-ba