{"paper_id":"2506.22611","title":"Deep Hedging to Manage Tail Risk","authors":["Yuming Ma"],"paper_date":"2025-06-27","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2506.22611v1","page_url":"https://thequant.space/flowcharts/deep-hedging-to-manage-tail-risk/","scores":{"empirical_rigor":7,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes using deep neural networks to parameterize convex‑risk minimization (CVaR/ES) for portfolio tail‑risk hedging.","quote":"we innovatively employ deep neural networks to parameterize convex-risk minimization (CVaR/ES) for the portfolio tail-risk hedging problem.","summary":"Extending Buehler et al.’s 2019 Deep Hedging paradigm, we innovatively employ deep neural networks to parameterize convex-risk minimization (CVaR/ES) for the portfolio tail-risk hedging problem. Through comprehensive numerical experiments on crisis-era bootstrap market simulators – customizable…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The primal portfolio contains m component assets and the hedging portfolio contains d component assets.","type":"model","stated":"explicit","quote":"Here, we assume that the primal portfolio contains m component assets and the hedging portfolio contains d component assets."},{"assumption":"Trading constraints such as long‑only or order‑size limits are omitted for simplicity.","type":"execution","stated":"explicit","quote":"trading constrains, e.g., long-only, rebalancing order size limitation and etc., for simplicity"},{"assumption":"The hedging portfolio's asset price is identical to the primary portfolio's asset price in the corollary analysis.","type":"model","stated":"explicit","quote":"Let the hedging portfolio's asset price be identical to the primary portfolio's asset price"},{"assumption":"Transaction costs can be customized and are incorporated via an explicit cost function C(·).","type":"market","stated":"explicit","quote":"To introduce explict trading costs for customizability, e.g., trasaction fees, short selling interests, margin interests, dealers' compensation, management fees and stamp duties etc."}]},"datasets":{"status":"available","items":[{"name":"Training data set","provider":null,"kind":"synthetic","access":"unknown","period":"2004/01/01 to 2005/12/31","universe":null,"quote":"Training data set is from 2004/01/01 to 2005/12/31."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"Explicit trading costs are accounted in cash account via a cost function C(·).","quote":"explicit trading costs are accounted in cash account, then we additionally put the explicit trading costs function C(·) into consideration","source":"evidence_card","rule":"R2","candidates":{"front_matter":"unclear","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Explicit trading costs are accounted in cash account via a cost function C(·).","card_quote":"explicit trading costs are accounted in cash account, then we additionally put the explicit trading costs function C(·) into consideration","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"unclear","findings":[{"finding":"The end‑to‑end framework achieves a significant one‑day 99% CVaR reduction.","kind":"performance","quote":"our end-to-end framework not only achieves significant one-day 99%CVaR reduction but also yields practical insights into friction-aware strategy adaptation"},{"finding":"Numerical experiments show the approach reduces the tail‑risk measure 1‑Day 99CVaR efficiently.","kind":"performance","quote":"We concluded from numerical experiments that our approach reduces tail risk measure, 1-Day 99CVaR, efficiently."},{"finding":"Performance of the MLP optimizer improves with more hidden layers on the tail‑risk hedging problem.","kind":"performance","quote":"Generally, performance of MLP optimizer increases on tail risk hedging problem with more hidden layers."},{"finding":"MLP optimizer trained on more volatile market data yields stronger profit generation while maintaining tail‑risk hedging performance.","kind":"performance","quote":"MLP optimizer learning from a relatively more volatile market empirical data leads to a stronger profit generative performance while remaining tail risk hedging performance without deterioration."}],"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":"2026-11-14T17:34:40Z"},"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/"}