Data and code for: One market-volume cap or one per battery? A formal decomposition of multi-asset battery back-tests in Great Britain

This record contains the data, code, retained analytical outputs and validation materials supporting the article “One market-volume cap or one per battery? A formal decomposition of multi-asset battery backtests in Great Britain” by Ethan B. Li and Chris J. Regan. The study examines an upstream modelling choice in multi-asset battery backtests: whether a common market-volume allowance is assigned separately to each asset, to intermediate economic-unit groupings, or once to the portfolio. It formalises cap assignment as a partition of physical assets, establishes the fixed-schedule ordering under partition refinement, evaluates all five partitions of a synthetic three-battery Great Britain portfolio, and separates three effects: partition accounting, dispatch reoptimisation and changes in the total participation budget. The empirical analysis covers 1 January 2023 to 22 December 2025, comprising 1,081 days and 51,888 half-hour periods. The deposit includes: 702 archived Elexon Market Index Data, Physical Notification and B1610 response files with request metadata; processed analytical inputs, including the 2022 Market Index warm-up data and a processed BOALF descriptive extension; Python code for the reference analysis, five-partition extension, robustness checks, source reconstruction, validation and figure generation; period-level and daily schedules, parameter surfaces, cap-preserving counterfactuals and reported result tables; a result registry linking each article and supplementary item to its source files and validation command; and a scoped evidence map of 15 adjacent storage-market formulations. The supplied scripts reproduce the retained results and check the 2023–2025 MID, PN and B1610 analytical inputs against the archived public responses. Market Index Volume is used as a matched-volume proxy, not participant-specific executable liquidity. The batteries form a synthetic test portfolio, and the reported conditional scores are not realised revenues. See LICENSES.txt for licensing details.