Learning to Rank Assets for Cross-Sectional Systematic Strategies

Summary

The paper addresses the asset-ranking step used in cross-sectional systematic portfolios. It argues that simple heuristics and rankings formed by sorting predictions from conventional regression or classification models can be suboptimal, drawing an analogy with ranking tasks in information retrieval. Its proposed framework uses learning-to-rank algorithms that learn pairwise and listwise relationships among instruments, with the goal of improving the ordering supplied to portfolio construction.

Cross-sectional momentum serves as the case study. The authors report that modern machine-learning ranking methods improve trading performance substantially, with Sharpe ratios approximately three times those of traditional approaches. The description does not name the algorithms, specify the market universe or evaluation period, or explain transaction costs and validation procedures. The result therefore motivates ranking methods as a strategy-design tool, but the provided summary is insufficient to assess robustness or assume the reported improvement transfers to other markets and portfolios.

Key ideas

• Cross-sectional systematic strategies depend on ranking assets before portfolio construction.
• Sorting regression or classification predictions may be suboptimal for ranking.
• Learning-to-rank methods can model pairwise and listwise relationships among instruments.
• The study demonstrates its approach using cross-sectional momentum.
• It reports approximately threefold higher Sharpe ratios than traditional approaches, without details here on costs or robustness.

Tags

# Building Cross-Sectional Systematic Strategies By Learning to Rank


# Building Cross-Sectional Systematic Strategies By Learning to Rank









The success of a cross-sectional systematic strategy depends critically on accurately ranking assets prior to portfolio construction. Contemporary techniques perform this ranking step either with simple heuristics or by sorting outputs from standard regression or classification models, which have been demonstrated to be sub-optimal for ranking in other domains (e.g. information retrieval). To address this deficiency, we propose a framework to enhance cross-sectional portfolios by incorporating learning-to-rank algorithms, which lead to improvements of ranking accuracy by learning pairwise and listwise structures across instruments. Using cross-sectional momentum as a demonstrative case study, we show that the use of modern machine learning ranking algorithms can substantially improve the trading performance of cross-sectional strategies -- providing approximately threefold boosting of Sharpe Ratios compared to traditional approaches.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.