The Finance Gazette

Nov 2 – Nov 8, 2026 · Start with what you can read, then let one unfamiliar paper widen the week.

PanelGP: A Probabilistic Factor Model for Asset-Return Panels

PanelGP predicts stock returns by letting company characteristics and broader economic conditions affect them in complex, non-straightforward ways. It offers investors flexible statistical tools as markets become highly dynamic and complex, the authors suggest.

PanelGP: A Probabilistic Factor Model for Asset-Return Panels · Andrea Ruglioni, Damir Filipovic · arXiv:2611.04040

Neural Score-Driven Volatility and Risk Forecasting

A small AI learns the actual patterns of market surprises to forecast how much markets will fluctuate, rather than assuming a standard distribution. It proved exceptionally good at predicting market risk during the turbulent COVID-19 period, beating traditional methods.

Neural Score-Driven Volatility and Risk Forecasting · Qianli Zhao, Chao Wang, Minh-Ngoc Tran et al. · arXiv:2611.05778

Neural Realized Measures: Forecast-Supervised Intraday Aggregation for Volatility Forecasting

A method calculates daily volatility from high-frequency price changes by learning which intraday periods best predict future volatility, instead of a fixed rule. It significantly improves forecast accuracy across models and assets, especially volatile cryptocurrencies, and shows which market moments are informative.

Neural Realized Measures: Forecast-Supervised Intraday Aggregation for Volatility Forecasting · Brian Sing Fan Chan, Jennifer So Kuen Chan · arXiv:2611.07287

Penalize the Common Trade: A Collision-Aware Layer for Learned Portfolios

A look at how different AI investment models, even if diverse, can converge on similar trades, creating market "crowding" and potential instability. A simple post-model step to reduce exposure to these common market movements, which in simulations significantly improves portfolio risk metrics like drawdown and volatility.

Penalize the Common Trade: A Collision-Aware Layer for Learned Portfolios · Eric Benhamou, Jean-Jacques Ohana, Beatrice Guez et al. · arXiv:2611.03453

Aggregate Calibration Can Mask Stress-State Undercoverage: Bounded-Memory Forecast Intervals under Recurrent Financial Stress

Prediction ranges set from recent data can look accurate but miss extreme events when market stress returns after a long gap. It warns that short-term memory in models may create overconfidence during recurring periods of economic uncertainty.

Aggregate Calibration Can Mask Stress-State Undercoverage: Bounded-Memory Forecast Intervals under Recurrent Financial Stress · Ruihang Han, Li-Hsiang Lin · arXiv:2611.04223

All Gazettes · Finance & Markets Research Club