{"paper_id":"2507.10701","title":"Kernel Learning for Mean-Variance Trading Strategies","authors":["Owen Futter","Nicola Muca Cirone","Blanka Horvath"],"paper_date":"2025-07-14","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2507.10701v1","page_url":"https://thequant.space/flowcharts/kernel-learning-for-mean-variance-trading-strategies/","scores":{"empirical_rigor":6,"math_complexity":9.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a kernel-based framework for constructing dynamic, path‑dependent trading strategies under a mean‑variance optimisation criterion.","quote":"In this article, we develop a kernel-based framework for constructing dynamic, path‑dependent trading strategies under a mean‑variance optimisation criterion.","summary":"In this article, we develop a kernel-based framework for constructing dynamic, pathdependent trading strategies under a mean-variance optimisation criterion. Building on the theoretical results of (Muca Cirone and Salvi, 2025), we parameterise trading strategies as functions in a reproducing kernel…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The underlying asset price process X is a non‑negative stochastic process with α‑Holder rough paths.","type":"model","stated":"explicit","quote":"... we assume that the price paths X_{0,t} are directly given as α‑Holder rough paths"},{"assumption":"A feature embedding ψ maps asset trajectories into ℝ^M streams of features.","type":"model","stated":"explicit","quote":"→ Λ_ψ is defined as a feature embedding that maps the input asset trajectories into new streams of features"},{"assumption":"The decay kernel K follows a power‑law form with a given exponent α.","type":"model","stated":"explicit","quote":"we also assume that the decay kernel K is given as a power law which is the same as seen in (4.1) for a given power law decay α"},{"assumption":"A 15‑minute forecast horizon with R^2 of 0.5% is used for predictive signals.","type":"data","stated":"explicit","quote":"... chosen to be small enough to demonstrate performance in low signal to noise ratio circumstances"},{"assumption":"The market impact and transaction‑cost trade‑off is conceptually discussed but not directly incorporated into headline results.","type":"execution","stated":"implied","quote":"This is conceptually similar to the trade‑off introduced when optimising under transaction costs, where smoother trading increases the variance of the position process, but reduces costs"}]},"datasets":{"status":"available","items":[{"name":"Synthetic data","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Through both synthetic and market‑data experiments, we illustrated that kernel trading strategies consistently outperform their Markovian counterparts"},{"name":"Market‑data (Databento)","provider":"Databento","kind":"other","access":"public","period":null,"universe":null,"quote":"We obtained this data from www.databento.com, [Dat25]."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"available","modeled":"unclear","headline_net_of_costs":null,"detail":null,"quote":null,"source":"conflict","rule":"R4b","candidates":{"front_matter":"no","evidence_card":"partial","evidence_card_quoted":true},"conflict_kind":"graded","card_detail":"Transaction costs are discussed conceptually as a trade‑off but are not quantified in the reported performance results.","card_quote":"This is conceptually similar to the trade‑off introduced when optimising under transaction costs, where smoother trading increases the variance of the position process, but reduces costs","meaning":"the platform classifier and the paper-text extraction disagree, so no value is asserted"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"Kernel trading strategies consistently outperform Markovian (Markowitz) strategies in both synthetic and market‑data settings, especially when signals exhibit temporal dependencies.","kind":"performance","quote":"Through both synthetic and market‑data experiments, we illustrated that kernel trading strategies consistently outperform their Markovian counterparts, especially in settings where asset dynamics or predictive signals exhibit temporal structure and non‑Markovianity."},{"finding":"The outperformance holds across varying signal decay speeds and stochastic‑volatility levels, with larger advantages for slower‑decaying, heavy‑tailed signals.","kind":"performance","quote":"The left‑hand panel shows that this outperformance holds regardless of whether the signal decays quickly or slowly, with a marginal advantage when the signal is slower and exhibits greater path‑dependency"},{"finding":"Using the signature kernel within the kernel framework yields comparable performance to the dedicated signature‑based method.","kind":"performance","quote":"conducted a detailed comparison with the signature trading approach of [FHW23], highlighting comparable performance when using the signature kernel."},{"finding":"The robust spectral decomposition reduces sensitivity to hyper‑parameter selection and improves out‑of‑sample stability.","kind":"theoretical","quote":"We also derived a robust spectral decomposition of the optimal solution, which reduces sensitivity to hyperparameter selection and improves out‑of‑sample stability."},{"finding":"Kernel methods remain effective with smaller datasets and scale better with many assets or signals compared to signature methods.","kind":"performance","quote":"Are kernel methods more efficient and robust with smaller datasets? ... Are kernel methods more computationally efficient with a large number of assets/signals?"}],"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":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":1,"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":"2026-11-14T17:32:36Z"},"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/"}