{"paper_id":"2406.08041","title":"HARd to Beat: The Overlooked Impact of Rolling Windows in the Era of Machine Learning","authors":["Francesco Audrino","Jonathan Chassot"],"paper_date":"2024-06-12","paper_type":"dataset-benchmark","paper_url":"https://arxiv.org/pdf/2406.08041v1","page_url":"https://thequant.space/flowcharts/hard-to-beat-the-overlooked-impact-of-rolling-windows-in-th/","scores":{"empirical_rigor":8.5,"math_complexity":4,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"The paper proposes that a carefully fitted HAR model provides superior forecasting accuracy compared to machine learning methods.","quote":"Our analysis focuses on the role of fitting schemes, particularly the training window and re-estimation frequency, in determining the HAR model's performance","summary":"We investigate the predictive abilities of the heterogeneous autoregressive (HAR) model compared to machine learning (ML) techniques across an unprecedented dataset of 1,455 stocks. Our analysis focuses on the role of fitting schemes, particularly the training window and re-estimation frequency, in…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"HAR models are estimated using a rolling window with regular re‑estimation.","type":"model","stated":"explicit","quote":"HAR models are typically estimated in a rolling window fashion, where the model is re-estimated at regular intervals using a fixed (or increasing) number of past observations."},{"assumption":"Transaction costs are proxied by the rolling median bid‑ask spread over the last nine months.","type":"execution","stated":"explicit","quote":"In this setting, we follow Bollerslev et al. (2018) and use the rolling median of the bid-ask spread over the last nine months as a proxy for transaction costs."},{"assumption":"Realized utility is interpreted as the percentage of wealth an investor would pay for access to the model’s forecasts.","type":"other","stated":"explicit","quote":"the values may be interpreted as the percentage of wealth that an investor would be willing to pay in order to access a portfolio that leverages the model’s forecasting capabilities compared to a portfolio that uniquely allocates capital in the risk‑free asset."},{"assumption":"The analysis uses only realized volatility and VIX as predictors for all models.","type":"data","stated":"explicit","quote":"HAR consistently outperforms its ML counterparts when both rely solely on realized volatility and VIX as predictors."}]},"datasets":{"status":"available","items":[{"name":"CRSP U.S. Stock Database","provider":"Center for Research in Security Prices","kind":"market-prices","access":"public","period":"January 2015 to October 2023","universe":"1,445 U.S. stocks consistently present over nine years","quote":"We query high-frequency data for a total of 6,061 stocks. Among these, 1,445 stocks remain consistently present throughout the nine years.","known_id":"crsp"},{"name":"NYSE Trade and Quote (TAQ) database","provider":"New York Stock Exchange","kind":"order-book","access":"commercial","period":"January 1, 2015 to October 31, 2023","universe":"high‑frequency intraday trade data for the 1,445 selected stocks","quote":"We source high-frequency intraday trade data for all selected stocks from the TAQ database from January 1, 2015, to October 31, 2023.","known_id":"taq"}],"known":[{"id":"taq","name":"NYSE TAQ","provider":"NYSE","kind":"market-prices","access":"commercial","quote":"We source high-frequency intraday trade data for all selected stocks from the TAQ database from January 1, 2015, to October 31, 2023.","mentions":2,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"rolling median of the bid‑ask spread over the last nine months used as proxy","quote":"In this setting, we follow Bollerslev et al. (2018) and use the rolling median of the bid-ask spread over the last nine months as a proxy for transaction costs.","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"rolling median of the bid‑ask spread over the last nine months used as proxy","card_quote":"In this setting, we follow Bollerslev et al. (2018) and use the rolling median of the bid-ask spread over the last nine months as a proxy for transaction costs.","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"HAR consistently outperforms ML models in predictive accuracy across QLIKE, MSE, and realized utility metrics.","kind":"predictive-accuracy","quote":"HAR consistently outperforms its ML counterparts when both rely solely on realized volatility and VIX as predictors."},{"finding":"The HAR model estimated via WLS yields the highest realized utility, even when transaction costs are included, regardless of VIX inclusion.","kind":"performance","quote":"the HAR model estimated via WLS markedly outperforms the other models, regardless of whether VIX is included."},{"finding":"Larger training windows generally lead to lower RMSE for the HAR model.","kind":"statistical","quote":"larger training windows generally lead to lower prediction errors."},{"finding":"ML models underperform relative to HAR, with the gap more pronounced for DJIA and Nasdaq‑100 subsets.","kind":"negative","quote":"the underperformance of ML models is even more pronounced in these subsets."}],"reported_performance":null,"out_of_sample":{"method":"walk-forward","headline_out_of_sample":null}},"independent_evidence":{"status":"available","supports":[{"paper":{"paper_id":"2408.01499","title":"NeuralFactors: A Novel Factor Learning Approach to Generative Modeling of Equities","paper_date":"2024-08-02","authors":["Achintya Gopal"],"empirical_rigor":6,"page_url":"https://thequant.space/flowcharts/neuralfactors-a-novel-factor-learning-approach-to-generativ/"},"question":"Does a larger training window length improve model performance (lower RMSE)?","this_paper_finding":"Larger training windows generally lead to lower RMSE for the HAR model.","other_paper_finding":"Increasing training data length improves performance, contradicting the belief that non‑stationarity requires shorter windows.","other_paper_quote":"the performance improved monotonically with more training data. In other words, the belief that non-stationarity requires us to use fewer years of training data appears to be incorrect and that more data consistently improves performance.","note":"Both findings report that increasing the training window length leads to better performance—A observes lower RMSE with larger windows, and B finds longer data improves results, so B supports A.","confidence":0.798},{"paper":{"paper_id":"2502.15813","title":"Stock Price Prediction Using a Hybrid LSTM-GNN Model: Integrating Time-Series and Graph-Based Analysis","paper_date":"2025-02-19","authors":["Meet Satishbhai Sonani","Atta Badii","Armin Moin"],"empirical_rigor":7,"page_url":"https://thequant.space/flowcharts/stock-price-prediction-using-a-hybrid-lstm-gnn-model-integr/"},"question":"Does increasing the training window size improve forecasting accuracy and adaptability?","this_paper_finding":"Larger training windows generally lead to lower RMSE for the HAR model.","other_paper_finding":"Expanding window training improved adaptability, mitigated concept drift, and enhanced generalisation.","other_paper_quote":"The expanding window training approach offers several key advantages. Firstly, it enhances adaptability","note":"Both findings report that larger/expanding windows lower error (RMSE) and enhance generalisation, showing consistent benefits across HAR and LSTM‑GNN approaches.","confidence":0.478}],"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":1,"extracted_at":"2026-11-14T18:07:51Z"},"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/"}