LambdaRankIC: Directly Optimizing Rank IC for Financial Prediction

Yan LIN · 2026

Read LambdaRankIC: Directly Optimizing Rank IC for Financial Prediction by Yan LIN (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-00501.txt.

Description

Yan Lin, Yihong Su, and Yi Yang train financial prediction models on the very score they will be judged by. That score is Rank IC, the rank correlation between a model's predictions and the asset returns that actually follow. Most models instead train on regression losses or generic ranking goals, which can pull against the target metric.

Their method, LambdaRankIC, optimizes Rank IC directly. Ranking is not differentiable, so they derive closed-form expressions for the gradients created by swapping pairs of assets, letting standard gradient training proceed inside the LambdaRank framework. They implement it as a custom objective in a popular boosting library, and prove it optimizes an upper bound on the target. In simulations it recovered true rankings and beat rivals under noisy, heavy-tailed conditions. On real market data it led on Rank IC, stability, monthly return, and Sharpe ratio.

Catalog record

CREATORYan LIN
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.00501

Browse the full library · About the Tower of Babel