Derived Feature Dataset for "Forecast-to-Trade under Information Constraints"

This record contains the exact derived feature-matrix snapshot used for the empirical analyses reported in “Forecast-to-Trade under Information Constraints: Frozen Out-of-Sample Evidence from the Germany–Luxembourg Day-Ahead Electricity Market.” The archived dataset, delu_features.parquet, contains 66,455 observations and 31 variables, covering the research sample through 31 July 2026. It includes electricity prices, system-fundamental forecasts, calendar variables, lagged price features, rolling price statistics, and derived renewable and residual-load variables. The dataset was constructed from public ENTSO-E Transparency Platform data using the feature-construction pipeline archived with the associated software release: https://doi.org/10.5281/zenodo.22878011 This Parquet file is preserved as the exact research snapshot used for the reported analysis. A newly retrieved ENTSO-E dataset may not be byte-for-byte identical because source-platform historical records can subsequently be revised. The manuscript distinguishes a Full information set from a Tier-1 robustness information set because the exact historical publication vintages of some renewable forecast variables at the simulated 11:45 D-1 decision cutoff could not be reconstructed. File integrity is documented in SHA256SUMS.txt. The SHA-256 checksum of delu_features.parquet is: C9B64768F98857E92EAEB72D0BD3D107770541FD563CE0D14AEFC6D4A46795AD Underlying market data originate from the ENTSO-E Transparency Platform. Reuse of the underlying source data remains subject to the applicable ENTSO-E terms and licensing conditions.