{"paper_id":"2510.22206","title":"Right Place, Right Time: Market Simulation-based RL for Execution Optimisation","authors":["Ollie Olby","Andreea Bacalum","Rory Baggott","Namid Stillman"],"paper_date":"2025-10-25","paper_type":"software-tool","paper_url":"https://arxiv.org/pdf/2510.22206v1","page_url":"https://thequant.space/flowcharts/right-place-right-time-market-simulation-based-rl-for-exec/","scores":{"empirical_rigor":7,"math_complexity":5.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a reinforcement learning framework for discovering optimal execution strategies, evaluated within a reactive agent‑based market simulator.","quote":"In this work, we present a reinforcement learning (RL) framework for discovering optimal execution strategies, evaluated within a reactive agent-based market simulator.","summary":"Execution algorithms are vital to modern trading, they enable market participants to execute large orders while minimising market impact and transaction costs. As these algorithms grow more sophisticated, optimising them becomes increasingly challenging. In this work, we present a reinforcement…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Price dynamics follow an arithmetic Brownian motion, implying normally distributed trading costs with known mean and variance.","type":"market","stated":"explicit","quote":"They assumed that price dynamics followed an arithmetic Brownian motion. This means trading costs are normally distributed with a known mean and variance."},{"assumption":"Slippage is the sum of market risk and market impact.","type":"execution","stated":"explicit","quote":"Slippage, 𝜁, can be defined as the sum of market risk, 𝜁𝑀𝑅, and market impact, 𝜁𝑀𝐼, 𝜁= 𝜁𝑀𝑅+ 𝜁𝑀𝐼"},{"assumption":"Order arrivals for both limit and market orders follow a Poisson process.","type":"model","stated":"explicit","quote":"The rate of order submission (both market and limit) can be estimated by assuming a Poisson model."},{"assumption":"The agent‑based market simulator can generate realistic reactive order flow that decomposes market impact and risk from slippage.","type":"other","stated":"implied","quote":"ABMs provide the functionality to model the diverse strategies, reactions, and mechanisms of various market participants, generating synthetic data that reflects plausible market conditions."},{"assumption":"The RL policy network receives a constant dummy state because no contextual information is provided.","type":"model","stated":"explicit","quote":"where 𝑥 is the input state vector (which we set as a constant dummy state to account for the lack of context in the model)"},{"assumption":"EU AI Act of 2024 requires clear explanations for high‑risk AI systems used in trading decisions.","type":"other","stated":"explicit","quote":"the EU AI Act of 2024 dictates that, if using a high-risk AI system (such as is typically used in trading decisions), any person subject to the output of the AI has the right to clear and meaningful explanations"}]},"datasets":{"status":"available","items":[{"name":"Simudyne Pulse intraday market simulator","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Figure 1: Overview of the proposed framework using the Simudyne Pulse intraday market simulator."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"unclear"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"Slippage is modeled as the sum of market impact and market risk; the RL loss is computed from market slippage, impact and risk measured in the simulator.","quote":"The loss is determined by executing the order schedule through our market simulator and computing the market slippage, impact and risk.","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Slippage is modeled as the sum of market impact and market risk; the RL loss is computed from market slippage, impact and risk measured in the simulator.","card_quote":"The loss is determined by executing the order schedule through our market simulator and computing the market slippage, impact and risk.","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"execution-cost","findings":[{"finding":"RL‑derived strategies consistently outperform TWAP and VWAP baselines in slippage and lie close to the Almgren‑Chriss efficient frontier.","kind":"performance","quote":"Our RL agent, trained in a realistic agent-based simulation, produced execution strategies that outperform traditional baselines such as TWAP and VWAP in terms of slippage and proximity to the efficient frontier."},{"finding":"The bi‑modal unbounded distribution trained under the slippage objective yields the most efficient execution strategy by optimizing impact and risk separately.","kind":"performance","quote":"The bi-modal unbounded distribution trained under the slippage objective produced the most efficient execution strategy by using its two modes to optimise impact and risk separately."},{"finding":"When optimizing directly for market impact, the RL agent reduces impact values compared to baselines but incurs higher market risk.","kind":"performance","quote":"... the RL agent consistently reduced impact values compared to both baselines and the slippage‑optimised strategies. However, this came at the cost of increased market risk."},{"finding":"Under the slippage loss function, RL strategies achieve similar market impact to baselines but improve overall transaction costs by reducing risk.","kind":"performance","quote":"strategies... we see improvement for all distributions over the baselines and a considerable improvement for the unbounded distribution."},{"finding":"The uni‑modal distribution trained with the slippage function can achieve optimal VaR at a 95% confidence level.","kind":"performance","quote":"The uni-modal distribution trained with the slippage function produces an execution strategy which can produce optimal VaR at a 95% confidence level."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":27,"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":"2026-11-14T16:38:40Z"},"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/"}