{"paper_id":"2501.06701","title":"Sequential Portfolio Selection under Latent Side Information-Dependence Structure: Optimality and Universal Learning Algorithms","authors":["Duy Khanh Lam"],"paper_date":"2025-01-12","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2501.06701v2","page_url":"https://thequant.space/flowcharts/sequential-portfolio-selection-under-latent-side-information/","scores":{"empirical_rigor":2.5,"math_complexity":8.5,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes learning algorithms that rely only on past asset prices, without side information or knowledge of latent dependence structures, and shows they achieve asymptotic growth comparable to the optimal dynamic log‑optimal strategy.","quote":"Based on our ﬁndings, a class of learning algorithms that relies solely on past assets’ prices, without any prior knowledge of latent dependence structures or side information, is shown to be optimal and comparable to another innovative learning algorithm that is more sophisticated and leverages complete market knowledge.","summary":"This paper investigates the investment problem of constructing an optimal no-short sequential portfolio strategy in a market with a latent dependence structure between asset prices and partly unobservable side information, which is often high-dimensional. The results demonstrate that a dynamic…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The market is stationary, meaning the latent dependence structure is statistically stable over time.","type":"market","stated":"explicit","quote":"Specifically, if the market is stationary, implying that the dependence structure is statistically stable, the growth rate of an optimal dynamic strategy, utilizing the maximum capacity of the entire market information, almost surely decays over time into an equilibrium state, asymptotically converging to the growth rate of a constant strategy.","support":"full"},{"assumption":"The latent dependence structure remains unchanged over time.","type":"model","stated":"explicit","quote":"Generally, if the latent dependence structure remains unchanged over time, a random optimal constant strategy could exist.","support":"full"},{"assumption":"Portfolios are constructed without commission fees for arbitrary fractions.","type":"execution","stated":"explicit","quote":"With the above settings, let the initial capital S0 \u0000b0 \u0001 =: S0 = 1 by convention for any strategy, and assume that the portfolios are constructed without commission fees for arbitrary fractions.","support":"full"},{"assumption":"No‑trash assets condition: the ratio of maximum to minimum asset returns is bounded away from zero.","type":"market","stated":"explicit","quote":"A market is said not to have trash assets if max1≤i≤m Xn,i/min1≤i≤m Xn,i ≥c ≫0 for all n almost surely.","support":"full"},{"assumption":"Results on convergence of the log‑optimal strategy can be proved without the no‑trash assets assumption, using ergodic theorem.","type":"statistical","stated":"explicit","quote":"It is important to highlight that the almost sure convergence of the log-optimal strategy in Theorem 2 can be proven without the assumption of no-trash assets, by simply using Birkhoff’s ergodic theorem, as in the original proof by Algoet and Cover (1988).","support":"full"},{"assumption":"The analysis assumes positive real‑valued asset returns and real‑valued side information defined on a canonical probability space.","type":"model","stated":"explicit","quote":"Assume that the assets’ returns and side information are realized with respect to a pairwise random process \b Xn,Yn ∞ n=1 that is jointly deﬁned on the canonical probability space \u0000Ω,F,P \u0001 , i.e., \u0000Xn,Yn \u0001\u0000ω \u0001 = \u0000Xn \u0000ω \u0001 ,Yn \u0000ω \u0001\u0001 for all n, where ω ∈Ω. Speciﬁcally, the positive real-valued vector Xn := \u0000Xn,1,...,Xn,m \u0001 ∈Rm ++ represents the positive returns of the assets at time n, which is deﬁned as the ratio of","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":"available","modeled":"unclear","headline_net_of_costs":null,"detail":null,"quote":null,"source":"conflict","rule":"R4b","candidates":{"front_matter":"n/a","evidence_card":"no","evidence_card_quoted":true},"conflict_kind":"applicability","card_detail":null,"card_quote":"With the above settings, let the initial capital S0 \u0000b0 \u0001 =: S0 = 1 by convention for any strategy, and assume that the portfolios are constructed without commission fees for arbitrary fractions.","meaning":"the platform classifier and the paper-text extraction disagree, so no value is asserted"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"The best dynamic strategy cannot infinitely often outperform an optimal constant strategy in growth rate or expected growth rate as time goes to infinity.","kind":"theoretical","quote":"We demonstrate that the best dynamic strategy, deﬁned as a sequence of portfolios determined using the complete history of market observations over time, including even the unobservable features, cannot inﬁnitely often outperform an optimal constant strategy, represented as a unique and random portfolio, in terms of both growth rate and expected growth rate as time progresses to inﬁnity.","support":"full"},{"finding":"A random optimal constant strategy exists almost surely and can asymptotically match the growth rate of the log‑optimal dynamic strategy under a stationary market with side information.","kind":"theoretical","quote":"By analyzing the dynamic log-optimal portfolio strategy as the optimal benchmark in a stationary market with side information, we show that a random optimal constant strategy almost surely exists, even when a limiting growth rate for the dynamic strategy does not. However, we show that a random optimal constant strategy can almost surely asymp-totically approach the optimal growth rate of the log-optimal strategy","support":"full"},{"finding":"Under stationarity, the advantage of utilizing the full market information decays over time, leading to equilibrium with the constant strategy’s growth rate.","kind":"theoretical","quote":"It implies that under the stationarity condition, the ad-vantages of utilizing the maximum capacity of market information will almost surely decay over time into an equilibrium state.","support":"full"},{"finding":"An empirical distribution‑based learning algorithm, using only past prices, achieves asymptotic growth comparable to the universal strategy when an optimal constant strategy exists.","kind":"theoretical","quote":"Despite the algorithm’s simplicity, such an empirical strategy, as proposed below, is comparable with the Universal strategy in general markets, as long as there exists an optimal constant strategy that is comparable to the log-optimal one, as demonstrated in Theorem 3. In comparison, Theorem 3 demonstrates the optimality of the empirical distribution-based log-optimal strategy in a more general market, without","support":"full"},{"finding":"The constant strategy does not require side information yet remains asymptotically optimal, contradicting conventional expectations that full information always yields better performance.","kind":"theoretical","quote":"This ﬁnding implies that the constant strategy does not require knowledge of periodic side market information, yet it remains asymptotically optimal. Such a result contradicts conventional perspectives in the ﬁelds of information theory, ﬁnance, and potentially learning theory, which suggest that possessing full knowledge of market information and dependence structures should unequivocally outperform ig-norance.","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":20,"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":"2026-11-14T19:53:06Z"},"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/"}