{"paper_id":"2405.01598","title":"Predictive Decision Synthesis for Portfolios: Betting on Better Models","authors":["Emily Tallman","Mike West"],"paper_date":"2024-04-30","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2405.01598v1","page_url":"https://thequant.space/flowcharts/predictive-decision-synthesis-for-portfolios-betting-on-bet/","scores":{"empirical_rigor":4,"math_complexity":8,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes Bayesian Predictive Decision Synthesis (BPDS), a fully Bayesian framework that combines model forecasts and decision outcomes for portfolio analysis.","quote":"Bayesian Predictive Decision Synthesis (BPDS), introduced in Tallman and West (2023), closes these gaps. BPDS is a fully Bayesian, theoretically founded framework to evaluate, compare, and combine sets of candidate models based on both their anticipated and historical decision outcomes as well as pure predictive perfor","summary":"We discuss and develop Bayesian dynamic modelling and predictive decision synthesis for portfolio analysis. The context involves model uncertainty with a set of candidate models for financial time series with main foci in sequential learning, forecasting, and recursive decisions for portfolio…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Returns are approximated as normally distributed for score uncertainty analysis.","type":"statistical","stated":"explicit","quote":"With daily returns on FX and stock indices, f and q1/2 are typically of the order of 0.01 or less; this is relevant in exploring score uncertainty via the variance matrix Vp = Vp[s(y, x)]. Were p(r) to be normal, then it is easy to show that Vp(−e2/2) = (f 2 + q/2)Vp(e)."},{"assumption":"Initial model probabilities are derived from traditional Bayesian model averaging (BMA).","type":"model","stated":"explicit","quote":"Initial model probabilities πtj = Pr(Mj|Dt−1), such as from traditional Bayesian model averaging (BMA) based on prior data and information, and/or from historical performance in goal-focused forecasting..."},{"assumption":"The tilting vector is constrained so that the risk‑reward ratio dt = τ1/τ2 is less than the target return r*.","type":"execution","stated":"explicit","quote":"We additionally restrict the tilting vector such that dt < r∗ t so that the portfolio target can be at most doubled; this is in order to obviate large values earlier seen in Figure 3, and is enforced in the computations to evaluate τ t at each time point."},{"assumption":"The data are split into training, model‑selection, and hold‑out periods.","type":"data","stated":"explicit","quote":"The dataset is divided into 3 time periods. The period to the beginning of 2015 is used to fit a large set of models, the next 4 years (2015-2018 inclusive) is used to select a subset of “better-performing models”, and the final 3 years (2019-2021 inclusive) defines a hold-out data test period."},{"assumption":"Portfolio decisions are based on one‑step ahead forecasts unless otherwise specified.","type":"execution","stated":"explicit","quote":"BMA focuses wholly on 1-step ahead predictive accuracy, so the initial probabilities used here in BPDS generally favour models based on predictive performance at this short-term horizon."}]},"datasets":{"status":"available","items":[{"name":"Daily FX returns for 9 currencies","provider":null,"kind":"market-prices","access":"public","period":"January 2001 – December 2021","universe":"9 major foreign exchange currencies","quote":"The data set includes daily returns for q = 9 currencies beginning in January 2001 and ending in December 2021."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"no","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"no","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"a trading or execution result is reported without costs"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"BPDS achieves superior cumulative returns and Sharpe ratios compared to BMA and improved BMA across all settings of the score‑improvement parameter.","kind":"performance","quote":"BPDS is able to achieve superior cumulative returns and Sharpe ratios compared to BMA and “improved BMA”, across all settings of mt."},{"finding":"Five‑day horizon portfolios show weaker improvements than one‑day portfolios, likely due to using initial BMA‑based weights that favor short‑term predictive accuracy.","kind":"performance","quote":"The 5-day portfolios improve over BMA comparisons but generally see relatively lack‑luster results compared to the 1-day portfolios. This is mainly due to the use of initial BMA-based weights."},{"finding":"BPDS results are close to those of a “BMA 2” benchmark and do not consistently outperform it in the first set of comparisons.","kind":"performance","quote":"BPDS outcomes are generally close to those of BMA 2, but do not empirically improve upon the latter in this first set of comparisons."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":24,"citations":{"cited_by_in_corpus":1,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":1},"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":"2026-11-14T18:10:15Z"},"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/"}