Traditional portfolio optimization typically relies on a single fitted return distribution, thus ignores non-normality and uncertainty across distributional states. Mean–variance–ambiguity (MVA) models, motivated by the smooth-ambiguity preferences of Klibanoff, Marinacci, and Mukerji (2005) and the KMM-to-MVA Arrow–Pratt reduction of Maccheroni, Marinacci, and Ruffino (2013), extend conventional risk penalties to account for distributional uncertainty, but they missed how state-conditional return distributions affect portfolio optimization stability. We develop Deep Mean–Variance–Ambiguity (Deep MVA) to resolve this gap by learning state-conditional return distributions and embedding the resulting conditional information loss into the ambiguity penalty. The resulting local certainty equivalent combines reduced-law mean–variance risk with a quadratic penalty for second-order disagreement. Conditional-mixture state resolution yields a Deep ambiguity covariance equal to the Traditional ambiguity covariance plus a positive-semidefinite state-feature increment. This increment is the conditional L2 loss from averaging state-specific distributions before optimization. On compact convex feasible sets, we show that the state-resolving program has stronger ambiguity-relevant curvature, lower mean-input sensitivity, and a unique global solution under restricted positive curvature or a common ridge; its projected iterations converge linearly. An LSTM-MDN learns state-conditional distribution densities from 60-minute bars, and Probability and VoV scores enter weekly MVA, with GARCH_Resid as a comparator. Tests on the SSE 50, CSI 300, and CSI 500 universes from 2017–2025 find statistically significant Sharpe out-of-sample improvements over Traditional MVA and MVO across all three universes and ambiguity specifications.

We propose a deep reinforcement learning method to improve pairs trading by identifying nonlinear relationships in stock news. Using the CSI 300 index constituents from 2015 to 2022, we integrated cointegration and news co-occurrence analysis in asset pairing and used a threshold-based approach in trading design. Results showed our NEWS-CO-DRL method, fusing deep learning and news co-occurrence, outperformed in return generation and risk control, indicating its potential for the Chinese A-share market.

This paper investigates market effects associated with China Security Index 300 (CSI 300) reconstitutions with sample period from April 2005 to Feb 2008. Several findings are listed as followings: Firstly, cumulative abnormal returns for added stocks increase slightly after announcements, while the returns for removed stocks decrease significantly though reverse immediately after index reconstitutions. Considering the whole event period, prices for deletions do not fall dramatically; it’s consistent with asymmetric change of investors’ awareness proposed by H Chen et al (2004). Secondly, both the results of cumulative abnormal returns and volume ratios do not provide evidence to support price pressure hypothesis or index membership hypothesis. We attribute those results to few funds tracking stock indices exactly with the same components and weights as which in the underlying indices in emerging markets, i.e., enhanced index funds are more familiar. Thirdly, the percentage of the additions’ (or deletions’) shares held by funds is not affected obviously by CSI 300 reconstitutions. Finally, we examine index change effects due to IPO that frequently occur in emerging markets, and find that additions witness a full reversal after the first trading day.