{"paper_id":"2407.01566","title":"A Parametric Contextual Online Learning Theory of Brokerage","authors":["François Bachoc","Tommaso Cesari","Roberto Colomboni"],"paper_date":"2024-05-22","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2407.01566v2","page_url":"https://thequant.space/flowcharts/a-parametric-contextual-online-learning-theory-of-brokerage/","scores":{"empirical_rigor":2,"math_complexity":8.5,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes a contextual online learning model for brokerage between traders and introduces algorithms (Algorithm 1 and Algorithm 2) that achieve optimal regret bounds in both 2‑bit feedback and full‑feedback settings.","quote":"We propose a contextual version of the online brokerage problem, that is of significant practical interest given that the broker often has access to meaningful information about the asset being traded and the 1In the US alone, the value of assets traded in OTC markets exceeded a remarkable 50 trillion USD in 2020, surpassing centralized markets by more than 20 trillion USD [64]. 3 Learning in Contextual Brokerage In","summary":"We study the role of contextual information in the online learning problem of brokerage between traders. In this sequential problem, at each time step, two traders arrive with secret valuations about an asset they wish to trade. The learner (a broker) suggests a trading (or brokerage) price based…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"There exists an unknown linear relation between the hidden market value mt and the observed context ct, i.e., mt = c_t^⊤ φ for some unknown φ.","type":"model","stated":"explicit","quote":"The first assumption (Theorem 1.1) states that an unknown linear relation exists between the unknown market value mt and the corresponding context ct the broker observes before proposing a trading price. There exists ϕ ∈[0,1]d, unknown to the broker, such that, for each t ∈N, it holds that mt = c⊺ t ϕ.","support":"full"},{"assumption":"Traders’ valuations are zero‑mean perturbations of the market value: V_t = m_t + ξ_t and W_t = m_t + ζ_t with E[ξ_t]=E[ζ_t]=0.","type":"statistical","stated":"explicit","quote":"The second assumption accounts for variability due to personal preferences or individual needs of the traders by modeling traders’ valuations as zero-mean perturbations of the market values. There exists an independent sequence of random variables ξ1,ζ1,ξ2,ζ2,... such that, for each t ∈N, it holds that E[ξt] = 0 = E[ζt] and Vt = mt + ξt and Wt = mt + ζt.3 Contexts.","support":"full"}]},"datasets":{"status":"none_found","items":[],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"none"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"Algorithm 1 attains a regret of order Ō(√{L d T}) in the 2‑bit feedback setting, matching a matching lower bound Ω(√{L d T}).","kind":"theoretical","quote":"3 Learning in Contextual Brokerage In this section, we introduce an algorithm (Algorithm 1) for the contextual brokerage problem for which we prove regret guarantees of order̃ O( √ LdT). We conclude this section by stating a matching (up to logarithmic terms) worst-case Ω( √ LdT) regret lower bound for any algorithm, proving the optimality of Algorithm 1.","support":"full"},{"finding":"Algorithm 2, using full feedback, achieves a logarithmic worst‑case regret of O(L d log T), with a matching lower bound Ω(L d log T).","kind":"theoretical","quote":"If the learner runs Algorithm 2 and the traders’ valuations admit a density bounded by L ≥1, then, for any time horizon T ∈N, it holds that RT ≤1 + 4LdlnT. We conclude this section by stating a matching worst-case Ω(LdlnT) regret lower bound for any algorithm in the full-feedback case, proving the optimality of Algorithm 2.","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":5,"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":2,"extracted_at":"2026-11-14T19:57:00Z"},"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/"}