Deep momentum networks with market trend dynamics

Jaemin Song; Jaegi Jeon

doi:10.1371/journal.pone.0331391

Peer Review History

Original SubmissionMarch 18, 2025
13 Jun 2025 Decision Letter - Tomo Popovic, Editor PONE-D-25-14477Deep Momentum Networks with Market Trend DynamicsPLOS ONE Dear Dr. jeon, Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Specifically, Criterion 3 (Technical quality and detailed description): The manuscript presents a technically compelling approach, but the absence of publicly available code, lack of detail on final hyperparameter settings, and limited transparency in the feature engineering process hinder reproducibility and independent validation of the results. Criterion 4 (Conclusions supported by data): While the conclusions are broadly consistent with the reported findings, they are presented without statistical testing or confidence measures and rely on results from a single asset class, limiting the generalizability of the claims. Criterion 7 (Data availability): The dataset used is not openly accessible, and the reliance on a third-party data provider with restricted access directly conflicts with PLOS ONE’s open data policy, preventing full compliance with community standards for data transparency. Also, please include discussion on the limitations of the study and provide the next steps. Please submit your revised manuscript by Jul 28 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. Please include the following items when submitting your revised manuscript: A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter. If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols. We look forward to receiving your revised manuscript. Kind regards, Tomo Popovic, Ph.D. Academic Editor PLOS ONE Journal requirements: When submitting your revision, we need you to address these additional requirements. 1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf 2. Please note that PLOS ONE has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse. 3. Thank you for stating the following financial disclosure: “This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2023-00242528 and RS-2024-00355646).” Please state what role the funders took in the study. If the funders had no role, please state: "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript." If this statement is not correct you must amend it as needed. Please include this amended Role of Funder statement in your cover letter; we will change the online submission form on your behalf. 4. Thank you for stating the following in the Acknowledgments Section of your manuscript: “This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2023-00242528 and RS-2024-00355646).” We note that you have provided funding information that is currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form. Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows: “This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2023-00242528 and RS-2024-00355646).” Please include your amended statements within your cover letter; we will change the online submission form on your behalf. [Note: HTML markup is below. Please do not edit.] Reviewers' comments: Reviewer's Responses to Questions Comments to the Author 1. Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. Reviewer #1: Partly Reviewer #2: Yes ******** 2. Has the statistical analysis been performed appropriately and rigorously? Reviewer #1: I Don't Know Reviewer #2: Yes ****** 3. Have the authors made all data underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: No Reviewer #2: Yes ****** 4. Is the manuscript presented in an intelligible fashion and written in standard English? PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here. Reviewer #1: Yes Reviewer #2: Yes ****** 5. Review Comments to the Author Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters) Reviewer #1: The paper extends the Deep-Momentum Network by injecting nine probabilistic market-state signals— “Market-Trend Dynamic Points” (MTDPs)—and letting an LSTM learn how those states alter momentum exposure. Tests on 99 continuous futures from 1995-2021 show consistent, if moderate, gains in Sharpe, Sortino, and Calmar ratios over the vanilla DMN, indicating that a finer regime lens can sharpen trend-following models. Several methodological details remain opaque. Hyper-parameter tuning is described as a grid search, yet the specific ranges, random seeds, early stopping rules, and final choices are not reported, so reproducibility and computational cost are hard to judge. The walk-forward protocol in Scenario 3 needs a stricter chronology: wording suggests cross-validation folds may straddle the 2019-2021 boundary, letting pandemic-era information leak into model selection. A purely rolling-origin evaluation would eliminate that risk. Performance differences would carry more weight with statistical context. Weekly returns are autocorrelated, yet no confidence intervals or bootstrap p-values accompany the Sharpe and drawdown numbers. A stationary bootstrap would clarify whether the 0.05–0.10 uptick in Sharpe is likely to persist. Sensitivity to friction also deserves deeper treatment: Sharpe dips below zero once round-trip costs reach 5 bps for several window settings, whereas real futures trading often exceeds that level. Extending the cost sweep to 2, 5, 10, and 25 bps and reporting turnover by asset class would show where the strategy remains viable. The nine MTDP probabilities provide a natural hook for attribution, but only qualitative remarks appear. Gradient-based or SHAP analysis highlighting which states drive position shifts, plus an ablation that feeds the LSTM only bull- and bear-state probabilities, would demonstrate the concrete value of the intermediate transitions. Minor presentation issues persist. Repeated phrases (“…as well as a stronger as well as a stronger momentum effect”), overlapping figure labels, and undefined terms such as “ordinary point (OP)” distract from the core contribution; trimming these and moving dense asset-level plots to an appendix would improve readability. Reviewer #2: The manuscript presents an innovative and well-executed study on enhancing time-series momentum (TSMOM) strategies using deep momentum networks (DMNs) enriched with market trend dynamic points (MTDPs). The integration of XGBoost-based feature engineering and LSTM networks is methodologically sound and offers valuable insights for both academics and practitioners in algorithmic trading. 1. Clarification on Model Selection Criteria: The choice of five fast momentum look-back windows is reasonable, but further clarification on how these specific intervals were determined would enhance the methodological transparency. 2. Discussion of Model Robustness: The results suggest no universally optimal window size. A brief discussion on potential hybrid or adaptive models that dynamically adjust window lengths could be added in the conclusion or future work section. 3, Figures & Legends: Some figures (e.g., Figures 6 and 7) could benefit from more detailed captions to guide readers unfamiliar with the model architecture. 4. Typos and Formatting: A few minor typographical issues (e.g., “Correaction” instead of “Correction” in Table 2) should be corrected. ****** 6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: No Reviewer #2: No ******** [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step. https://doi.org/10.1371/journal.pone.0331391.r001
Revision 1
28 Jul 2025 Author Response Responses to Reviewer 1 We thank Reviewer 1 for their constructive feedback, which has helped us improve both the clarity and the scope of the manuscript. We address each comment in turn. Comment 1. Hyper-parameter tuning is described as a grid search, yet the specific ranges, random seeds, early stopping rules, and final choices are not reported, so reproducibility and computational cost are hard to judge. Response. We have clarified in Section 3.2 (lines 174–179) that early stopping was not employed; model selection relied solely on validation log-loss. Furthermore, the asset-wise optimal hyperparameters obtained from the grid search are now provided in the Supporting Information (S1 Table). The hyperparameter selections for the DMN model, which vary across each fold of the rolling-window backtest, have been omitted for brevity due to their prohibitive volume. These additions supply the previously missing details for the XGBoost MTDP classifier. Comment 2. The walk-forward protocol in Scenario 3 needs a stricter chronology: wording suggests cross-validation folds may straddle the 2019-2021 boundary, letting pandemic-era information leak into model selection. A purely rolling-origin evaluation would eliminate that risk. Response. To eliminate any ambiguity regarding the chronology of Scenario 3, we have substantially revised “Section 4.3 Backtesting Setups and Section 4.4 Training Details”. Key changes include: 1. The three evaluation schemes are now referred to as backtesting setups, clarifying that they differ in data-splitting rules. 2. For Setups 1 and 3, we explicitly describe a rolling-origin expanding-window framework: a fixed 5-year test window, an in-sample window that grows annually, and a strictly chronological 90:10 train/validation split (lines 273–277 & 284–299). 3. Setup 2 is now presented as a 2020 hold-out COVID-19 stress test without rolling expansion (lines 278–283). 4. Figure 5 on page 12 has been redrawn to clearly mark fixed 5-year test windows and now includes a concise summary in the caption. 5. Section 4.4 also clarifies that hyperparameter tuning (100-trial random search) is performed within each iteration, using only the current train/validation slice; the test set remains untouched throughout (Table 4 caption revised). These changes ensure that validation folds do not cross the 2019–2021 boundary, and that Setups 1 and 3 adhere strictly to a rolling-origin evaluation. Comment 3. Performance differences would carry more weight with statistical context. Weekly returns are autocor- related, yet no confidence intervals or bootstrap p-values accompany the Sharpe and drawdown numbers. A stationary bootstrap would clarify whether the 0.05-0.10 uptick in Sharpe is likely to persist. Response. To address the concern regarding statistical inference, we have incorporated a full stationary-bootstrap analysis and integrated its results into the manuscript. Key additions include: 1. Section 5.2 “Performance Evaluation” (lines 351–356) briefly introduces the one-sided stationary bootstrap analysis [1], using B = 50,000 resamples and block probability p ≈ 0.50 (mean block length ≈ 2 weeks) for each asset, selected via the plug-in rule of [2]. We confirm that weekly returns exhibit weak autocorrelation, justifying the short block length. 2. Table 5 has been retitled and now reports one-sided p-values. Superscripts * and ** denote significance at the 5% and 1% levels, respectively. 3. Section 5.2 “Performance Evaluation” has been revised to reflect the statistical significance of Sharpe and related metrics across models and backtesting setups. These revisions provide the requested statistical context: the observed 0.05–0.10 Sharpe-ratio gains in Setups 1 and 3 are statistically significant (p < 0.05), while differences in Setup 2 are not. Comment 4. Sensitivity to friction also deserves deeper treatment: Sharpe dips below zero once round-trip costs reach 5 bps for several window settings, whereas real futures trading often exceeds that level. Extending the cost sweep to 2, 5, 10, and 25 bps and reporting turnover by asset class would show where the strategy remains viable. Response. We have expanded the manuscript to provide a deeper treatment of transaction-cost sensitivity: 1. “Sharpe Ratios and Transaction Costs in Setup 3” part in Section 5.2 (lines 400–435) now evaluates round-trip costs of 2, 5, 10, and 25 bps and reports the corresponding cost-adjusted Sharpe ratios and annual turnover by asset class. 2. Figure 7 visualizes the erosion of Sharpe ratios across the four cost levels, and Table 6 reports the corresponding cost-adjusted Sharpe ratios together with annual turnover for each asset class. These additions identify the cost thresholds at which the proposed strategies remain profitable and pinpoint the asset classes with the greatest exposure to slippage. Comment 5. The nine MTDP probabilities provide a natural hook for attribution, but only qualitative remarks ap- pear. Gradient-based or SHAP analysis highlighting which states drive position shifts, plus an ablation that feeds the LSTM only bull- and bear-state probabilities, would demonstrate the concrete value of the intermediate transitions. Response. A new Section 5.3 “Feature Attribution and Ablation Study on MTDP Scores” now provides the requested quantitative evidence. 1. Sign-preserving, L1-normalised Integrated Gradients (Figure 8; full plots in S2 Appendix) show that flows between Correction and Bull (BuToCo, CoToBu) are the most influential channels in the long-horizon setups, directly linking specific MTDP components to position-sizing decisions. 2. Bull–Bear ablation (Table 7) collapsing the nine states to three (Bull, Bear, OP) keeps headline returns unchanged but lowers Sharpe (1.589 → 1.555), Sortino (2.479 → 2.425), and Calmar (1.583 → 1.487) ratios, confirming that intermediate transitions contain irreplaceable information. S2 Appendix documents the IG calculation procedure and presents feature-importance plots for all MTDP models. The combined gradient analysis and ablation demonstrate that the full nine-component MTDP score is essential for robust, risk-efficient trading. Comment 6. Minor presentation issues persist. Repeated phrases (“. . . as well as a stronger as well as a stronger momentum effect”), overlapping figure labels, and undefined terms such as “ordinary point (OP)” dis- tract from the core contribution; trimming these and moving dense asset-level plots to an appendix would improve readability. Response. Minor issues fixed (duplicates removed, figure labels tidied, OP defined at line 28). To improve readability, detailed asset-level plots and diagnostics are now in the Supporting Information. References [1] Politis DN, Romano JP. The stationary bootstrap. Journal of the American Statistical association. 1994;89(428):1303–1313. [2] Politis DN, White H. Automatic block-length selection for the dependent bootstrap. Econometric reviews. 2004;23(1):53–70. Responses to Reviewer 2 We thank Reviewer 2 for their careful reading and valuable comments regarding the motivation, theo- retical formulation, and contribution of our work. Comment 1. Clarification on Model Selection Criteria: The choice of five fast momentum look-back windows is reasonable, but further clarification on how these specific intervals were determined would enhance the methodological transparency. Response. We selected the five fast windows (2, 4, 8, 16, and 20 weeks) to follow an approximately exponential grid, balancing coverage of various time horizons while keeping the model parsimonious. As now explained in Section 2.3 (lines 99–107), this approach extends prior work to a weekly frequency. We also clarify that intermediate windows (e.g., 6 and 12 weeks) were excluded for tractability, as preliminary tests showed they did not materially improve risk-adjusted outcomes while significantly increasing the hyper-parameter search space. Comment 2. Discussion of Model Robustness: The results suggest no universally optimal window size. A brief discussion on potential hybrid or adaptive models that dynamically adjust window lengths could be added in the conclusion or future work section. Response. We agree and have expanded the conclusion to outline a regime-aware hybrid design. Specifically, Section 6 (lines 494–501) now proposes a two-stage architecture that (i) employs a Hidden Semi- Markov Model to infer latent trend regimes and (ii) routes inputs through a Multi-Task Mixture- of-Experts network whose expert weights are modulated by the inferred regime probabilities. This adaptive approach enables smooth re-weighting among 2-, 4-, 8-, 16-, and 20-week experts and will be the focus of our follow-up study. Comment 3. Figures & Legends: Some figures (e.g., Figures 6 and 7) could benefit from more detailed captions to guide readers unfamiliar with the model architecture. Response. To improve clarity and manuscript flow, we have thoroughly revised our figures and moved the asset- specific examples to the Supporting Information (S1 Appendix). – The figures referenced by the reviewer (formerly Figures 6 and 7) are now located in S1 Appendix as Figures A1–A5. – All captions have been rewritten to be self-contained. For example, Figure A1’s title now specifies the assets and timeframe ("DMN-based trading strategies for the S&P 500 and Euro-Bobl futures (2016-–2020)"), while the captions for Figures A2–A5 clarify the exact asset and backtesting setup used in each plot. This approach ensures each visual is self-explanatory, directly addressing the reviewer’s concern. Comment 4. Typos and Formatting: A few minor typographical issues (e.g., “Correaction” instead of “Correction” in Table 2) should be corrected. Response. We performed a manuscript-wide proofreading pass and fixed all detected issues. Attachments Attachment Submitted filename: Responses to Reviewer 2.pdf https://doi.org/10.1371/journal.pone.0331391.r002
15 Aug 2025 Decision Letter - Tomo Popovic, Editor Deep Momentum Networks with Market Trend Dynamics PONE-D-25-14477R1 Dear Dr. jeon, We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements. Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication. An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support. If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. Kind regards, Tomo Popovic, Ph.D. Academic Editor PLOS ONE Additional Editor Comments (optional): Reviewers' comments: Reviewer's Responses to Questions Comments to the Author 1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation. Reviewer #1: All comments have been addressed Reviewer #2: All comments have been addressed ******** 2. Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. Reviewer #1: Yes Reviewer #2: (No Response) ****** 3. Has the statistical analysis been performed appropriately and rigorously? Reviewer #1: N/A Reviewer #2: Yes ****** 4. Have the authors made all data underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: Yes Reviewer #2: Yes ****** 5. Is the manuscript presented in an intelligible fashion and written in standard English? PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here. Reviewer #1: Yes Reviewer #2: Yes ****** 6. Review Comments to the Author Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters) Reviewer #1: All comments have been adressed and all the previous ,,problems" and shortcomings are resolved. The paper is technicaly very sound. Reviewer #2: (No Response) ****** 7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: No Reviewer #2: No ******** https://doi.org/10.1371/journal.pone.0331391.r003
Formally Accepted
Acceptance Letter - Tomo Popovic, Editor PONE-D-25-14477R1 PLOS ONE Dear Dr. jeon, I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now being handed over to our production team. At this stage, our production department will prepare your paper for publication. This includes ensuring the following: * All references, tables, and figures are properly cited * All relevant supporting information is included in the manuscript submission, * There are no issues that prevent the paper from being properly typeset You will receive further instructions from the production team, including instructions on how to review your proof when it is ready. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few days to review your paper and let you know the next and final steps. Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. You will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing. If we can help with anything else, please email us at customercare@plos.org. Thank you for submitting your work to PLOS ONE and supporting open access. Kind regards, PLOS ONE Editorial Office Staff on behalf of Prof. Tomo Popovic Academic Editor PLOS ONE https://doi.org/10.1371/journal.pone.0331391.r004

Open letter on the publication of peer review reports

PLOS recognizes the benefits of transparency in the peer review process. Therefore, we enable the publication of all of the content of peer review and author responses alongside final, published articles. Reviewers remain anonymous, unless they choose to reveal their names.

We encourage other journals to join us in this initiative. We hope that our action inspires the community, including researchers, research funders, and research institutions, to recognize the benefits of published peer review reports for all parts of the research system.

Learn more at ASAPbio .