Imagine a busy restaurant kitchen where the last dish ordered each night determines the price every diner pays. For decades in Britain's electricity grid, that last dish was always gas — a combined-cycle gas turbine (CCGT) sitting at the expensive margin of the dispatch stack. This paper proves, with granular unit-level data, that batteries have muscled into that kitchen and are now cooking the marginal dish more than a third of the time on the bid side. That is not a forecast. It is a reconstruction of 50,684 settlement periods. The committed claim: batteries in the GB Balancing Mechanism rose from 0.8% to 36.2% of bid-active marginal price-setting and from 3.2% to 26.9% on the offer side between 2023 and 2025, displacing CCGTs and pumped storage respectively. Capacity-normalised marginal capture increased on both sides, and a unit-quarter fixed-effects regression finds that each additional 100 MW of BM-active battery capacity was associated with a 0.55 and 0.48 percentage-point increase in quarterly bid- and offer-side marginal share. By 2025, battery-set prices were approximately £9–10/MWh more favourable than volume-matched non-battery alternatives drawn from the same ladder. The architecture here is not a model or simulation — it is an empirical reconstruction. Dalton and O'Sullivan rebuild the price-ranked eligible stack for every half-hour settlement period, attribute marginality to individual Balancing Mechanism Units (BMUs), and separate long-system (bid-active) from short-system (offer-active) conditions while retaining co-marginal ties. The method leans on Elexon's public BM data, which provides unit-level bid/offer acceptances, and uses panel regressions with unit-quarter fixed effects to control for fleet heterogeneity. This is applied statistics on administrative data, not a novel algorithm — and that is its strength. The integrity regime is solid for an empirical energy economics paper. The data is public (Elexon), the sample is the universe of settlement periods (not a subsample), and the methodology — reconstructing the marginal stack and running panel regressions — is transparent and reproducible. The £9–10/MWh favourability figure is calculated against volume-matched non-battery alternatives from the same ladder, a thoughtful counterfactual. The key caveat the authors flag honestly: batteries are under-represented by 12.9 percentage points in the highest-priced 5% of short-system periods. They don't oversell. Individual batteries behave as price-takers; the fleet effect is endogenous to routine balancing but not to stress events. Where does this sit on the ladder? The novelty is methodological specificity, not conceptual surprise. Everyone in GB energy policy suspected batteries were becoming marginal price-setters. What this paper delivers is the first unit-level, exhaustive reconstruction proving it and quantifying the rate. Prior work on marginal price-setting in GB either used aggregate fuel-type proxies or shorter sample windows. The 0.8%→36.2% trajectory across two years, with named displaced incumbents (CCGTs on bids, pumped storage on offers), is the kind of number policymakers and market designers can actually use. The milestone question is about the 12.9 percentage-point under-representation gap in high-price periods. Batteries set routine prices but duck when the system is truly stressed — exactly the moments that matter most for security of supply and consumer cost. Closing that gap requires either longer-duration storage, market redesign, or both. The next concrete number to watch: battery marginal share in the top-5% priced periods. If it climbs from its current deficit toward parity with the fleet's overall share, the decarbonisation of price formation extends from routine to crisis conditions. The obvious experiment not run: decomposing the £9–10/MWh favourability by battery duration (1-hour vs 2-hour vs 4-hour), location (constraint zones), and time of day. The authors have the unit-level data to do this. The honest read is (c) — they are saving it for the next paper. This paper establishes the headline result; the granular decomposition is the natural sequel for Applied Energy or a follow-up.
