Beyond Forecast Accuracy: Translating Multi-Market Signals into State-Dependent Selling Policies with CVaR Downside Control

This study adopts a two-stage computational workflow. MATLAB is used for data preparation and for constructing market-signal variables from the price series, while GAMS is used to solve the constrained decision-optimization problem that represents real-world selling operations. In the first stage, price data are cleaned, aligned to a common time frequency, and transformed into decision-time observable signals under a strict no–look-ahead condition in MATLAB. The processed dataset is then passed to GAMS to derive disciplined selling rules and/or selling plans that improve risk-adjusted performance subject to operational constraints, including decision timing, cash-flow feasibility, and transaction/marketing rules. Finally, the GAMS outputs are returned to MATLAB for statistical and risk-based evaluation and for producing the study’s tables and figures.