Decision-Induced Ranking Explains Prediction Inflation and Excessive Turnover in SPO-Based Portfolio Optimization

TL;DR

This paper analyzes how decision-focused learning in portfolio optimization can lead to inflated returns and high turnover, proposing stabilization methods to improve practical implementation.

Contribution

It provides a KKT-based interpretation of SPO-based DFL and evaluates practical mechanisms to reduce prediction inflation and excessive turnover.

Findings

Clipping, rescaling, and partial adjustment help stabilize portfolios.

Realistic constraints improve the implementability of SPO strategies.

Empirical evidence shows reduction in prediction inflation and turnover.

Abstract

Decision-focused learning (DFL) is attractive for portfolio optimization because it trains predictors according to downstream decision quality rather than prediction accuracy alone. However, SPO(Smart, Predict then Optimize surrogate)-based DFL may produce inflated return signals and unstable portfolio reallocations. This study provides a KKT-based interpretation showing that portfolio decisions can be viewed as ranking over risk- and transaction-cost-adjusted marginal scores. Empirically, we examine prediction inflation and excessive turnover in SPO-trained portfolios, and evaluate clipping, min-max rescaling, and partial portfolio adjustment as practical stabilization mechanisms. The results suggest that realistic output constraints and portfolio-level turnover control improve the implementability of SPO-based portfolio strategies.