{"paper_id":"2508.01880","title":"Time-Varying Factor-Augmented Models for Volatility Forecasting","authors":["Duo Zhang","Jiayu Li","Junyi Mo","Elynn Chen"],"paper_date":"2025-08-03","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2508.01880v3","page_url":"https://thequant.space/flowcharts/time-varying-factor-augmented-models-for-volatility-forecast/","scores":{"empirical_rigor":8,"math_complexity":7,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a novel, model-agnostic Factor-Augmented Volatility Forecast framework.","quote":"we propose a novel, model-agnostic Factor-Augmented Volatility Forecast framework.","summary":"Accurate volatility forecasts are vital in modern finance for risk management, portfolio allocation, and strategic decision-making. However, existing methods face key limitations. Fully multivariate models, while comprehensive, are computationally infeasible for realistic portfolios. Factor models…","strategy":"Factor-Augmented Volatility-Scaled Pairs Trading"},"assumptions":{"status":"available","items":[{"assumption":"Daily position sizes are scaled to achieve a 25% annualized volatility target.","type":"execution","stated":"explicit","quote":"Our strategy employs a volatility-targeting approach, where daily position sizes are scaled to achieve a 25% annualized volatility target."},{"assumption":"A transaction cost of 5 basis points per leg is applied in the backtest.","type":"execution","stated":"explicit","quote":"The backtest is initialized with an equity of $50,000 and incorporates a 5 basis point transaction cost per leg."},{"assumption":"Trading signals are generated from a rolling z‑score of the spread with thresholds ±1.5.","type":"execution","stated":"explicit","quote":"Trading signals are generated from a rolling z-score of the spread, where 𝑍𝑡=𝑆𝑡−𝜇𝑆𝑡 / 𝜎𝑆𝑡 . We initiate a long position when 𝑍𝑡< -1.5 and a short position when 𝑍𝑡> +1.5, closing the position when the z-score reverts to zero."},{"assumption":"Statistical models use a 70‑day rolling window; LSTM uses a 30‑day window.","type":"model","stated":"explicit","quote":"The rolling window length (𝑊) is set to 70 days for the statistical models to balance responsiveness to regime changes with statistical stability, while a shorter window of 30 days is used for the LSTM to align with its input structure."}]},"datasets":{"status":"available","items":[{"name":"Equity high‑frequency dataset","provider":"Wharton Research Data Service","kind":"market-prices","access":"commercial","period":"January 2018 to December 2020","universe":"MSFT, AMD, INTC, ORCL, CSCO (S&P 500 Information Technology)","quote":"The equity dataset focuses on five prominent firms within the S&P 500 Information Technology Sector, selected for their large market capitalization and historical sectoral comovements: Microsoft (MSFT), Advanced Micro Devices (AMD), Intel (INTC), Oracle (ORCL), and Cisco (CSCO). For these firms, we obtained high-freque"},{"name":"Cryptocurrency high‑frequency dataset","provider":"Kaiko","kind":"crypto-onchain","access":"commercial","period":"January 2018 to April 2021","universe":"BTC, ETH, XRP, ADA, LTC","quote":"The cryptocurrency dataset, sourced from Kaiko, covers January 2018 to April 2021. This dataset comprises detailed, high-frequency trading data for five leading cryptocurrencies: Bitcoin (BTC), Ethereum (ETH), Ripple (XRP), Cardano (ADA), and Litecoin (LTC).","known_id":"kaiko"}],"known":[{"id":"kaiko","name":"Kaiko","provider":"Kaiko","kind":"market-prices","access":"commercial","quote":"The cryptocurrency dataset, sourced from Kaiko, covers Jan-uary 2018 to April 2021.","mentions":2,"used_in":1}],"sample":{"start":null,"end":null,"market":"cryptocurrencies","asset_class":"crypto","frequency":"daily"},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":true,"detail":"5 basis point transaction cost per leg","quote":"The backtest is initialized with an equity of $50,000 and incorporates a 5 basis point transaction cost per leg.","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"5 basis point transaction cost per leg","card_quote":"The backtest is initialized with an equity of $50,000 and incorporates a 5 basis point transaction cost per leg.","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"Factor augmentation reduces forecast errors by up to 12.8% for equities and 22.8% for cryptocurrencies, lowering MSE and QLIKE.","kind":"predictive-accuracy","quote":"eases by up to 12.8% for equities and 22.8% for cryptocurrencies, with corresponding reductions in Mean Squared Error (MSE) and Quasi-likelihood loss (QLIKE)."},{"finding":"AR model’s annualized return improves by 23.6% after factor augmentation (5.35% → 6.61%) and Sharpe ratio rises from 1.437 to 1.443.","kind":"performance","quote":"... specification boosts its annualized return by a remarkable 23.6%."},{"finding":"Factor‑augmented LSTM flips a -5.5% annualized loss into a +7.3% gain and improves Sharpe from -0.467 to 0.787 in a challenging market period.","kind":"performance","quote":"In a stunning reversal, our factor-augmented LSTM transformed this failing strategy into a resounding success, flipping the annualized loss into a +7.3% gain. This phenomenal turnaround propelled its Sharpe ratio from a negative -0.467 to a positive 0.787."}],"reported_performance":{"strategy":"Factor-Augmented Volatility-Scaled Pairs Trading","sharpe":1.485,"sharpe_benchmark":null,"annual_return_pct":7.3,"total_return_pct":null,"max_drawdown_pct":null},"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":30,"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":"2026-11-14T17:31:08Z"},"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/"}