{"paper_id":"2503.00603","title":"Understanding the Commodity Futures Term Structure Through Signatures","authors":["Hari P. Krishnan","Stephan Sturm"],"paper_date":"2025-03-01","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2503.00603v1","page_url":"https://thequant.space/flowcharts/understanding-the-commodity-futures-term-structure-through-s/","scores":{"empirical_rigor":3,"math_complexity":8,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes a regular perturbation expansion of the signature of commodity futures term structures, linking signature variability to model parameters of the Gibson–Schwartz model, particularly the volatility of the convenience yield.","quote":"The present article aims to address this issue directly, by introducing and developing the concept of signature perturbations. In particular, we construct a regular perturbation of the signature of the term structure of log prices for various commodities, in terms of the convenience yield.","summary":"Signature methods have been widely and effectively used as a tool for feature extraction in statistical learning methods, notably in mathematical finance. They lack, however, interpretability: in the general case, it is unclear why signatures actually work. The present article aims to address this…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Constant risk‑free rate and annualized cost of storage are assumed for pricing futures contracts.","type":"model","stated":"explicit","quote":"Assuming a constant risk-free rate and annualized cost of storage, we can then price a futures contract of any given maturity.","support":"full"},{"assumption":"Futures contracts are continuously rolled, approximating actual market practice.","type":"market","stated":"explicit","quote":"Several futures contracts with diﬀerent maturities have been written on the underlying asset X. To simplify the model, we assume that the futures are continuously rolled. This approximates actual market practice.","support":"full"},{"assumption":"Interest rate, storage costs and convenience yield can be time‑dependent and stochastic in the general model.","type":"model","stated":"explicit","quote":"In the general case, r, s and C can all be time-dependent and stochastic.","support":"full"},{"assumption":"A probability measure Q exists under which the discounted asset process is a local martingale (no arbitrage).","type":"model","stated":"explicit","quote":"To exclude arbitrage in the sense of no free lunch with vanishing risk (see [DS06]), we require the existence of a probability measure Q under which the discounted asset process ( ˜Xte−rt)t≥0 is a local G-martingale.","support":"full"},{"assumption":"The filtration is the augmented natural filtration of the two Brownian motions driving spot and convenience yield.","type":"model","stated":"explicit","quote":"We assume that the ﬁltration G is the (augmented) natural ﬁltration of the Brownian motion (W 1, W 2).","support":"full"},{"assumption":"A small parameter δ>0 scales the speed of the convenience‑yield process, enabling a regular perturbation around a constant convenience yield.","type":"model","stated":"explicit","quote":"For this we introduce the parameter δ > 0 that can, thanks to Brownian scaling, be considered as a (slow) speed of the convenience yield modeled, dXδ t = \u0000r −σ2 2 −Cδ t \u0001 dt + σ dW 1 t , Xδ 0 = x, dCδ t = δκ \u0000θ −Cδ t \u0001 dt + √ δγ dW 2 t , Cδ 0 = c. In the case of small δ ≪1 we can think about the convenience yield moving at a glacial pace relative to the spot price dynamics.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Figure 1 data","provider":"Allasso SA","kind":"market-prices","access":"commercial","period":null,"universe":null,"quote":"Data Availability Statement The data for Figure 1 were retrieved from Allasso SA, https://www.allasso.ch/.","support":"full"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"unclear"},"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 volatility of the convenience yield is identified as the major discriminant driving signature‑based classification performance.","kind":"theoretical","quote":"Our perturbation expansion and rigorous convergence estimates help explain the success of signature-based classification of commodities markets according to their term structure, with the volatility of the convenience yield as the major discriminant. Overall, the strongest eﬀect of the model parameters (besides the initial convenience yield) has by γ, the volatility of the convenience yield.","support":"full"},{"finding":"Signature variability can be linked to a measure of uncertainty in term‑structure dynamics, given fixed variability in the shape of the term structure.","kind":"theoretical","quote":"Given a ﬁxed amount of variability in the shape of the term structure, we want to estimate how much the signature will change. Assuming that our estimates are sharp enough, we can then relate signature variability to a measure of uncertainty in term structure dynamics.","support":"full"},{"finding":"Third‑order terms (order √δ³) isolate the mean‑reversion speed κ, while storage costs do not appear in the signature expansion.","kind":"theoretical","quote":"A possibility to separate their eﬀect comes only with the third order term (i.e., the order √ δ 3) where κ appears isolated in contributing terms. None of the terms depends on storage costs (as they were assumed constant).","support":"full"},{"finding":"A regular perturbation of the Gibson–Schwartz model yields a convergent expansion of the futures‑term‑structure signature in a weighted signature norm.","kind":"theoretical","quote":"In particular, we construct a regular perturbation of the signature of the term structure of log prices for various commodities, in terms of the convenience yield. We introduce a convenient notion of convergence of signature by deﬁning the weighted signature norm on the ambient space of sequences of random variables and derive sharp constants for the approximation of iterated integrals and signatures.","support":"full"},{"finding":"Empirically, the signature‑based method outperforms alternative approaches in distinguishing storable from non‑storable commodities, driven by convenience‑yield volatility.","kind":"performance","quote":"Thus if we employ a classiﬁcation algorithm on the feature set given by tthe (truncated) signature vecor, it is most likely to pick up on diﬀerences in the volatility of the convenience yield; this is what drives the superior performance of the signature-based method in discerning storable from non-storable commodities..","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":15,"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":"2026-11-14T19:52:03Z"},"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/"}