{"paper_id":"2601.05085","title":"Trading Electrons: Predicting DART Spread Spikes in ISO Electricity Markets","authors":["Emma Hubert","Dimitrios Lolas","Ronnie Sircar"],"paper_date":"2026-01-08","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2601.05085v1","page_url":"https://thequant.space/flowcharts/trading-electrons-predicting-dart-spread-spikes-in-iso-elec/","scores":{"empirical_rigor":8,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a multi‑zone DART spike forecasting model combined with a calibrated structural price‑impact model that yields closed‑form optimal virtual trade sizes.","quote":"we extend spike prediction from a single zone to a multi‑zone setting and treat both positive and negative DART spikes within a unified statistical model.","summary":"We study the problem of forecasting and optimally trading day-ahead versus real-time (DART) price spreads in U.S. wholesale electricity markets. Building on the framework of Galarneau-Vincent et al., we extend spike prediction from a single zone to a multi-zone setting and treat both positive and…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Price impact of virtual trades is linear in the aggregate system‑wide quantity and zone‑specific quantity.","type":"model","stated":"explicit","quote":"we approximate the zonal response of prices to aggregate traded quantity St by a piecewise linear function, with energy‑impact coefficients k+E and k‑E, and a separate linear impact from the trade size qt,z in the specific zone with coefficient kz"},{"assumption":"All predictors are constructed using only information available strictly before the day‑ahead bid deadline.","type":"data","stated":"explicit","quote":"As day‑ahead bids must be submitted well in advance of the operating day, all predictors are constructed using only data available strictly before th…"},{"assumption":"Spike events are extremely rare and the feature set is largely linear in its predictive structure.","type":"statistical","stated":"explicit","quote":"the spike events of interest are extremely rare and the feature set is largely linear in its predictive structure; consequently, more flexible models exhibit a tendency to overfit the noise"},{"assumption":"Logistic regression provides stable out‑of‑sample probability forecasts that can be reliably translated into trading signals.","type":"model","stated":"explicit","quote":"Logistic regression, by contrast, delivered the most robust and interpretable out‑of‑sample performance across all three markets, and produced stable probability estimates that translate reliably into trading signals"},{"assumption":"Cross‑zone DART movements are synchronized in ISO‑NE and ERCOT, allowing a single representative zone to capture most variation.","type":"market","stated":"explicit","quote":"ISO‑NE and ERCOT exhibit highly synchronized zonal DART movements, so that a single representative zone captures most of the relevant variation"},{"assumption":"Higher forecasted load increases losses and congestion in expectation, justifying the linear‑quadratic impact model.","type":"model","stated":"explicit","quote":"Both components are consistently positively correlated with forecast load across all seasons and time buckets, supporting the modeling assumption that higher forecasted load increases losses and congestion in expectation"}]},"datasets":{"status":"available","items":[{"name":"NYISO, ISO‑NE, ERCOT market data","provider":"public data portals of the corresponding system operators","kind":"market-prices","access":"public","period":"NYISO: 2015–2019; ISO‑NE and ERCOT: 2018","universe":"hourly panel of day‑ahead and real‑time prices, system and zonal load forecasts, exogenous","quote":"All market data were obtained from the public data portals of the corresponding system operators, namely ERCOT [11], ISO New England [21], and NYISO [29]."},{"name":"GridStatus platform","provider":"GridStatus (https://www.gridstatus.io)","kind":"other","access":"public","period":null,"universe":null,"quote":"Data access and aggregation were facilitated using the GridStatus platform (https://www.gridstatus.io)."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"A linear‑quadratic price‑impact model is calibrated from historical day‑ahead bid stacks and incorporated into P&L calculations.","quote":"we estimate system‑wide energy impact coefficients and zone‑specific congestion sensitivities, yielding a linear–quadratic impact model for virtual load","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"A linear‑quadratic price‑impact model is calibrated from historical day‑ahead bid stacks and incorporated into P&L calculations.","card_quote":"we estimate system‑wide energy impact coefficients and zone‑specific congestion sensitivities, yielding a linear–quadratic impact model for virtual load","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"The impact‑aware strategy yields realized price‑impact slopes between roughly $8 and $41 per MWh per 1,000 MWh, matching calibrated coefficients.","kind":"performance","quote":"The results in Table 17 show that the realized slopes lie between roughly 8 and 41$/MWh per 1000 MWh. These magnitudes are fully consistent with the calibrated Summer–Peak coefficients k+E ∈[34.64, 46.48]/1000MWh"},{"finding":"Logistic regression outperforms more flexible models in out‑of‑sample trading P&L despite similar in‑sample accuracy.","kind":"predictive-accuracy","quote":"While some nonlinear models achieved marginally higher in‑sample classification accuracy, they offered no consistent improvement in out‑of‑sample trading P&L"},{"finding":"NYISO exhibits heterogeneous zone‑level DART distributions, making multi‑zone modeling essential; Long Island is the most profitable zone.","kind":"statistical","quote":"Long Island emerges as the most profitable load pocket, combining frequent extreme spreads with high average load"},{"finding":"Restricting trades to the model‑consistent direction (side‑clipping) improves total P&L and reduces contradictory positions.","kind":"performance","quote":"Figure 13 shows that this clipped strategy performs better on the 2022–2025 test set. The total P&L increases relative to the unconstrained optimizer"},{"finding":"Cross‑zone interactions via the common energy component can make offsetting positions optimal to concentrate risk where marginal price impact is lowest.","kind":"theoretical","quote":"Closed‑form expressions for the optimal zonal quantities show that portfolio‑level decisions are shaped not only by local expected revenues, but also by cross‑zone interactions through the common energy component"}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":36,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:17:09Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}