{"paper_id":"2412.10947","title":"Auto-Regressive Control of Execution Costs","authors":["Simeon Kolev"],"paper_date":"2024-12-14","paper_type":"methodological","paper_url":"https://arxiv.org/pdf/2412.10947v1","page_url":"https://thequant.space/flowcharts/auto-regressive-control-of-execution-costs/","scores":{"empirical_rigor":7,"math_complexity":6,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes an Auto‑Regressive approach that iteratively derives OLS estimates of market parameters period‑to‑period, enabling uninformed investors to adjust their execution strategies over time.","quote":"Therefore, this study aims to bridge this gap by proposing an approach that iteratively derives OLS estimates of the market parameters from period to period. This methodology enables uninformed investors to engage in the market dynamically, adjusting their strategies over time based on evolving estimates, thus offering a practical solution for navigating the complexities of execution cost optimization without","summary":"Bertsimas and Lo’s seminal work established a foundational framework for addressing the implementation shortfall dilemma faced by large institutional investors. Their models emphasized the critical role of accurate knowledge of market microstructure and price/information dynamics in optimizing…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Investors do not have perfect initial knowledge of market dynamics (θ, γ, ρ) and volatility.","type":"market","stated":"explicit","quote":"However, this paper recognizes that perfect initial knowledge may not be a realistic assumption for new investors entering the market. Without knowing the true market dynamics ( 𝜃, 𝛾, 𝜌) and volatility (𝜎𝜀 2, 𝜎η 2) conditions, the optimal strategy is inefficient, and thus crucial to develop Auto-Regressive estimates of the model parameters (𝜃𝑡̂, 𝛾𝑡̂, 𝜌𝑡̂, 𝜎η𝑡 2 ,̂ 𝜎ε𝑡 2 ̂).","support":"full"},{"assumption":"UNC has access to some serially correlated market information but cannot observe the true correlation ρ.","type":"data","stated":"explicit","quote":"15 The Auto-Regressive Model Now suppose that UNC has access to some serially correlated market information, however, they are not able to estimate future (𝑋𝑡) as they do not know how correlated (𝜌) the data really is.","support":"full"},{"assumption":"UNC does not know the impact of market information (γ) nor the price impact of its own orders (θ).","type":"model","stated":"explicit","quote":"Additionally, they do not know market information impacts stock price (𝛾) nor do they know the price impact of their own market orders (𝜃).","support":"full"},{"assumption":"The price‑impact of market orders and information changes are linear and permanent (additive permanent price impact).","type":"model","stated":"explicit","quote":"The Auto-Regressive model, more technically known as Additive Permanent Price Impact with Auto-Regressive Estimate of Information, is such that the price-impact of market orders and changes in market information are linear and permanent.","support":"full"},{"assumption":"Parameter estimates (θ̂_t, γ̂_t, ρ̂_t, σ̂_ε_t², σ̂_η_t²) converge over time to their true values.","type":"statistical","stated":"explicit","quote":"The optimal strategy would be to derive unbiased estimators of the market parameters, such that the guesses are dynamic (𝜃𝑡̂, 𝛾𝑡̂, 𝜌𝑡̂), and convergent, lim 𝑇→∞(𝜃𝑡̂, 𝛾𝑡̂, 𝜌𝑡̂) = (𝜃, 𝛾, 𝜌).","support":"full"},{"assumption":"Trading is conducted over a fixed finite horizon with a fixed block of shares to be executed.","type":"execution","stated":"explicit","quote":"Specifically, given a fixed block of shares (𝑆𝑡) to be executed within a fixed finite number of periods (𝑇).","support":"full"}]},"datasets":{"status":"none_found","items":[],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"yes","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"unclear","findings":[{"finding":"Across 100 simulated sequences, the Naive, Informed, and Auto‑Regressive strategies outperform the naive expected best‑execution cost by approximately $0.0065, $0.0441, and -$0.0358 per share each period, respectively.","kind":"performance","quote":"Specifically, the Naive, Informed, and Auto-Regressive strategies outperform the Naive expected best-execution cost by around $0.0065, $0.0441, and -$0.0358 a share at each period, respectively.","support":"partial"},{"finding":"The simulated costs of the Naive, Informed, and Auto‑Regressive strategies show periodic variances of about $0.1096, $0.7550, and $0.9301 per share, indicating substantial risk‑adjusted differences.","kind":"performance","quote":"24 Additionally, the Naive, Informed, and Auto-Regressive strategies simulated costs show a periodic variance of around $0.1096, $0.7550, and $0.9301 a share, which is significant in comparison to the average improvement, implying that the optimal strategy, though better with reliable market information, may produce very favorable/unfavorable results and should be considered when comparing these models by their risk","support":"full"},{"finding":"The variance of the average improvement across simulations increases exponentially as the information relevance factor (γ) rises or as the information correlation (ρ) rises while information variance (σ_η²) falls.","kind":"theoretical","quote":"25 Conclusion From my experimentation with the models, the variance of the average improvement across 100 simulations seems to increase exponentially as either the market information relevance factor (𝛾) increases, or as the information correlation coefficient (𝜌) increases and the information variance (𝜎η 2) decreases.","support":"full"},{"finding":"When γ → 0, ρ → 0, or σ_η² → ∞, the optimal strategy collapses to the Naive approach regardless of predictability of market information.","kind":"theoretical","quote":"Logically, these parameter adjustments make sense as (𝛾) dictates the maximum expected improvement that the Informed strategy can produce in comparison to the Naive approach. As 𝛾 →0 the best strategy becomes Naive, regardless of if serial market information is 100% predictable.","support":"full"},{"finding":"The Auto‑Regressive approach, while slightly suboptimal, matches the Naive strategy until parameter estimates converge; with high ρ̂ or low σ̂_η² it emulates the Informed strategy.","kind":"performance","quote":"The Auto-Regressive approach, though showing slightly suboptimal performance, is limited to the same assumptions and restrictions that allow the Informed strategy to dominate. However, the Auto-Regressive approach shows autonomy when prioritizing other price impact parameters (𝜃𝑡̂, 𝛾𝑡̂, 𝜎ε𝑡 2 ̂) in the initial periods, as until (𝜌𝑡̂, 𝜎η𝑡 2 ̂) converge, the Auto-Regressive best execution will emulate the Naive","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":24,"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":"2026-11-14T19:53:33Z"},"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/"}