{"paper_id":"2511.07834","title":"Levy-stable scaling of risk and performance functionals","authors":["Dmitrii Vlasiuk"],"paper_date":"2025-11-11","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2511.07834v1","page_url":"https://thequant.space/flowcharts/levy-stable-scaling-of-risk-and-performance-functionals/","scores":{"empirical_rigor":5.5,"math_complexity":9,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a finite‑horizon model in which liquid‑asset returns follow a Lévy‑stable scaling on a data‑driven window [τUV, τIR] and provides closed‑form, horizon‑correct formulas for VaR, ES, Sharpe, Information ratios, Kelly fractions and drawdown thresholds.","quote":"We develop a finite-horizon model in which liquid-asset returns exhibit Levy-stable scaling on a data-driven window [tau_UV, tau_IR] and aggregate into a finite-variance regime outside.","summary":"We develop a finite-horizon model in which liquid-asset returns exhibit Levy-stable scaling on a data-driven window [“tau_UV, tau_IR”] and aggregate into a finite-variance regime outside. The window and the tail index alpha are identified from the log-log slope of the central body and a two-segment…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"On the window [τUV, τIR] returns admit a location‑scale representation with a standardized α‑stable driver: Rτ = µτ + σ τ^{1/α} Z.","type":"model","stated":"explicit","quote":"There exist 0 < τUV < τIR < ∞, parameters α ∈(1, 2), σ > 0, µ ∈R, and a standardized α-stable random variable Z with skewness β such that, for all τ ∈[τUV, τIR], Rτ d= µτ + σ τ 1/αZ.","support":"full"},{"assumption":"The driver Z has tail index α∈(1,2), skewness β∈[−1,1], zero mean, a continuous density at the origin and finite moments of order p<α.","type":"statistical","stated":"explicit","quote":"The driver has index α ∈(1, 2), skewness parameter β ∈[−1, 1], and characteristic function φZ(u) = exp n −|u|α\u0010 1 −i β tanπα 2 sign u \u0011o (α̸ = 1), with the Zolotarev modification when α = 1, as shown by Nolan (2013) and Rachev and Mittnik (2000). Assume EZ = 0, that fZ exists and is continuous at the origin, and that E|Z|p < ∞holds for all p < α.","support":"full"},{"assumption":"The return process is strictly stationary and α‑mixing with summable mixing coefficients, ensuring uniform laws of large numbers for the statistics used.","type":"statistical","stated":"explicit","quote":"The return process is strictly stationary and α-mixing with coefficients α(h) →0 and P∞ h=1 α(h)δ/(2+δ) < ∞for some δ > 0, which yields uniform laws of large numbers for the frequency and scale statistics used below. Assumption 2.1.","support":"full"},{"assumption":"For τ>τIR the dispersion aggregates to a finite‑variance √τ regime.","type":"model","stated":"explicit","quote":"For τ > τIR the dispersion obeys a finite-variance law Sτ = C √τ \u00001 + o(1) \u0001 (τ →∞).","support":"full"},{"assumption":"No inference is drawn for horizons τ<τUV.","type":"model","stated":"explicit","quote":"No inference is drawn for τ < τUV.","support":"full"},{"assumption":"An anchor horizon τ0 is used to match the Lévy‑stable scale to a Gaussian surrogate for bias correction.","type":"model","stated":"explicit","quote":"Matching at τ0 gives σ τ 1/α 0 mZ(q) = σG √τ0mN(q) =: Ξ0(q) and therefore ESτ(q) −ESG τ (q) = Ξ0(q) \"\u0012 τ τ0 \u00131/α − \u0012 τ τ0 \u00131/2# .","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":"The Lévy propagation yields flat exception rates for VaR and ES across horizons on the Lévy window.","kind":"statistical","quote":"Empirically, the Lévy propagation delivers flat exception rates for VaR and ES across horizons on the window, horizon-invariant p-Sharpe and p-Information ratios, Kelly fractions that decay with τ as τ 1−2/α, and drawdown thresholds whose realized breach frequencies match their design levels.","support":"full"},{"finding":"Horizon‑invariant p‑Sharpe and p‑Information ratios are obtained, while Gaussian √τ scaling distorts these metrics for α<2.","kind":"statistical","quote":"On the Lévy window, SRα is horizon-invariant and reorders strategies primarily by tail thickness; Gaussian annualization with √τ spuriously depresses Sharpe as τ grows when α < 2, a distortion removed by the τ 1/α scale. Empirically, the Lévy propagation delivers flat exception rates for VaR and ES across horizons on the window, horizon-invariant p-Sharpe and p-Information ratios, Kelly fractions that decay with τ","support":"full"},{"finding":"Kelly fractions decay with horizon as τ^{1−2/α} under the Lévy‑stable model.","kind":"theoretical","quote":"Empirically, the Lévy propagation delivers flat exception rates for VaR and ES across horizons on the window, horizon-invariant p-Sharpe and p-Information ratios, Kelly fractions that decay with τ as τ 1−2/α, and drawdown thresholds whose realized breach frequencies match their design levels.","support":"full"},{"finding":"Drawdown thresholds derived from the Lévy model have realized breach frequencies that match their design levels across horizons.","kind":"statistical","quote":"Empirically, the Lévy propagation delivers flat exception rates for VaR and ES across horizons on the window, horizon-invariant p-Sharpe and p-Information ratios, Kelly fractions that decay with τ as τ 1−2/α, and drawdown thresholds whose realized breach frequencies match their design levels.","support":"full"},{"finding":"All derived formulas differ from Gaussian propagation by an explicit exponent‑gap term (τ/τ0)^{1/α} − (τ/τ0)^{1/2}.","kind":"theoretical","quote":"Closed-form, horizon-correct formu-las were derived for VaR, ES, p-Sharpe, p-Information, Kelly under a Value-at-Risk constraint, and drawdown; in each case the Gaussian propagation differs by an explicit exponent-gap term (τ/τ0)1/α −(τ/τ0)1/2.","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":19,"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":"2026-11-14T19:46:16Z"},"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/"}