{"paper_id":"2509.12456","title":"Reinforcement Learning-Based Market Making as a Stochastic Control on Non-Stationary Limit Order Book Dynamics","authors":["Rafael Zimmer","Oswaldo Luiz do Valle Costa"],"paper_date":"2025-09-15","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2509.12456v1","page_url":"https://thequant.space/flowcharts/reinforcement-learning-based-market-making-as-a-stochastic-c/","scores":{"empirical_rigor":6,"math_complexity":8.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper implements a market‑making agent using Proximal‑Policy Optimization and evaluates it in a non‑stationary limit order book simulator.","quote":"Our contributions include a practical implementation of a market making agent based on the Proximal-Policy Optimization (PPO) algorithm, alongside a comparative evaluation of the agent’s performance under varying market conditions via a simulator-based environment.","summary":"Reinforcement Learning has emerged as a promising framework for developing adaptive and data-driven strategies, enabling market makers to optimize decision-making policies based on interactions with the limit order book environment. This paper explores the integration of a reinforcement learning…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The benchmark closed‑form optimal spread assumes normally distributed non‑mean reverting spreads, constant executed order size and exponentially distributed event time dynamics.","type":"model","stated":"explicit","quote":"The closed-form expression is derived from a simple market model which assumes normally distributed non-mean reverting spreads, constant executed order size and exponentially distributed event time dynamics."},{"assumption":"The training assumes no knowledge of the underlying dynamics of the environment (model‑free learning).","type":"model","stated":"explicit","quote":"since we assume no knowledge of the underlying dynamics of the environment when training the agent, a model-free approach is necessary."}]},"datasets":{"status":"available","items":[{"name":"Limit Order Book simulator with multiple non‑stationary dynamics","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Our main contribution is implementing multiple non-stationary dynamics into a single limit order book simulator, and how using fine‑controlled non‑stationary environments can enhance the resulting agent’s performance under adverse"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"not_extracted","modeled":"unclear","headline_net_of_costs":null,"detail":null,"quote":null,"source":"none","rule":"R0","candidates":{"front_matter":"unclear","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"the classifier was not confident"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"The RL agent achieved a mean financial return of 5.203 × 10⁻⁵ (annualized ≈ +1.31 %), outperforming the benchmark’s 3.037 × 10⁻⁵ (≈ +0.76 %).","kind":"performance","quote":"the reinforcement learning agent exhibited a mean financial return of 5.203 × 10−5 (annualized return of about +1.31%)."},{"finding":"The RL agent’s Sortino ratio was 0.7497, higher than the benchmark agents (0.4271 for Stoikov and –0.0079 for Long‑Only).","kind":"performance","quote":"benchmark agent scores of 0.4271 and −0.0079 for the Stoikov and Long‑Only strategies, respectively."},{"finding":"Processing latency was low: actor network 2.90 ms per episode and critic network 0.21 × 10⁻⁴ s per episode.","kind":"performance","quote":"the actor network taking on average 2.90ms per episode, and the critic network taking 0.21 × 10−4 seconds per episode, well within the acceptable range."}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":25,"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":"2026-11-14T16:38:24Z"},"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/"}