{"paper_id":"2511.10478","title":"Noise-proofing Universal Portfolio Shrinkage","authors":["Paul Ruelloux","Christian Bongiorno","Damien Challet"],"paper_date":"2025-11-13","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2511.10478v1","page_url":"https://thequant.space/flowcharts/noise-proofing-universal-portfolio-shrinkage/","scores":{"empirical_rigor":6,"math_complexity":8,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes two robustness methods for UPSA—time‑averaging of optimal ridge penalty weights and pre‑filtering covariance matrices with Average Oracle eigenvalues—and combines them into a new estimator AvgUPSA‑AO.","quote":"We propose two methods to make UPSA more robust and improve its efficiency: time-averaging of the optimal penalty weights and using the Average Oracle correlation eigenvalues to make covariance matrices less noisy and more robust to covariate shift.","summary":"We enhance the Universal Portfolio Shrinkage Approximator (UPSA) of Kelly et al. (2023) by making it more robust with respect to estimation noise and covariate shift. UPSA optimizes the realized Sharpe ratio using a relatively small calibration window, leveraging ridge penalties and…","strategy":"AvgUPSA-AO"},"assumptions":{"status":"available","items":[{"assumption":"The empirical analysis uses monthly returns of 153 characteristic‑managed US equity portfolios from 1970 to 2024.","type":"data","stated":"implied","quote":"We use the same dataset as Kelly et al. (2023): monthly returns of 153 characteristic-managed portfolios of US equities between 1970 and 2024 from Jensen et al. (2023)."},{"assumption":"Portfolios are calibrated on an in‑sample window of 120 months (10 years).","type":"execution","stated":"implied","quote":"The in-sample calibration window is set to T = 120 months."},{"assumption":"A ridge‑penalty grid of 20 logarithmically spaced points in the range [10⁻⁸, 10⁻¹] is used for all estimators.","type":"model","stated":"implied","quote":"fix the ridge penalty grid to 20 logarithmically spaced points in [10⁻⁸"},{"assumption":"Out‑of‑sample Sharpe ratio is computed on a 6‑month holding period after each monthly calibration.","type":"statistical","stated":"explicit","quote":"We define the OOS Sharpe ratio of a portfolio π, held for 6 months, with realized returns and risk: SR = πᵀ μ̂_OOS / √(πᵀ"},{"assumption":"Portfolios are rebalanced monthly and returns are scaled to target an annualized volatility of 10 %.","type":"execution","stated":"implied","quote":"Portfolios are rebalanced every month, and their return is measured as the realized return over the following month. The resulting cumulative log‑returns are rescaled using a centered 6‑month rolling standard deviation, targeting an annualized volatility of 10%."}]},"datasets":{"status":"available","items":[{"name":"Monthly returns of 153 characteristic‑managed US equity portfolios","provider":"Jensen et al. (2023) / Kelly et al. (2023)","kind":"market-prices","access":"public","period":"1970‑2024","universe":"153 factor portfolios","quote":"We use the same dataset as Kelly et al. (2023): monthly returns of 153 characteristic-managed portfolios of US equities between 1970 and 2024 from Jensen et al. (2023)."}],"known":[],"sample":{"start":"1985","end":"2024","market":"US equity characteristic-managed portfolios","asset_class":"equities","frequency":"monthly"},"data_resolution":"weekly-monthly"},"transaction_costs":{"status":"available","modeled":"no","headline_net_of_costs":false,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"no","evidence_card":"no","evidence_card_quoted":false},"meaning":"a trading or execution result is reported without costs"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"Combined estimators (UPSA‑AO and AvgUPSA‑AO) consistently outperform UPSA across calibration window sizes, as shown by higher mean annualized Sharpe ratios.","kind":"performance","quote":"We can observe a consistent over‑performance of combined estimators UPSA‑AO and AvgUPSA‑AO over UPSA."}],"reported_performance":{"strategy":"AvgUPSA-AO","sharpe":2.493,"sharpe_benchmark":null,"annual_return_pct":null,"total_return_pct":null,"max_drawdown_pct":2.73},"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":8,"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":"2026-11-14T17:21:59Z"},"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/"}