Next Article in Journal
Balancing Depth for Robustness: A Study on Reincarnating Reinforcement Learning Models
Previous Article in Journal
Tradition-Dismissive vs. Tradition Reconceptualization Approaches in Musculoskeletal Care: The Example of Osteopathic Care
 
 
Article
Peer-Review Record

Design of an Immersive Basketball Tactical Training System Based on Digital Twins and Federated Learning

Appl. Sci. 2025, 15(7), 3831; https://doi.org/10.3390/app15073831
by Xiongce Lv 1, Ye Tao 1,2, Yifan Zhang 1 and Yang Xue 3,*
Reviewer 1: Anonymous
Appl. Sci. 2025, 15(7), 3831; https://doi.org/10.3390/app15073831
Submission received: 27 February 2025 / Revised: 15 March 2025 / Accepted: 28 March 2025 / Published: 31 March 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

I hope this letter finds you well. I had the opportunity to review your article titled, “Design of an immersive basketball tactical training system based on digital twin and federated learning.”, which was submitted Applied Sciences.

  1. Introduction

-. This study clearly presents the importance and necessity of global education digitalization.

-. It was also very interesting because it explained well the connection between the research subject's national education informatization plan and the five-year sports development plan of China.

-. This study clearly defines three issues that apply digital twins and federated learning to sports education: 1. simulation issues, 2. personal information issues, and 3. system limitations, thereby showing differences from previous studies.

-. The researcher clearly states the purpose and hypothesis of the study. 1. The simulation implementation method, 2. Personal information protection and data utilization plan, 3. Methodological solutions, etc. are well explained.

-. Lastly, the research organization is written as a separate chapter to explain the structure of this study well. This is a part that is difficult to find in existing studies. By presenting this chapter, the researcher emphasizes the necessity and purpose of this study.

  1. Materials and Methods

-. It is judged that the precision of the experiment was increased by systematically collecting data by dividing the data collection into each stage, that is, pre-evaluation, data during training, and post-evaluation.

-. In addition, it is believed that the reliability of the study could be increased by collecting data such as movement trajectories and physiological signals using multimodal data.

-. However, due to the lack of explanation of the research tool, Motion Capture, it is judged that there is a lack of understanding of measurement errors and data accuracy. If possible, it is hoped that additional explanations, such as existing studies, will be provided to increase the reliability of the research tool and secure the reliability of the research results.

-. This study used an experimental design to verify effectiveness by comparing the control group and the experimental group.

-. It is considered highly desirable to use a stratified random sampling method for research subjects to reduce bias between the experimental and control groups.

-. Sample size verification used G*power 3.1 to collect an appropriate sample size, and it is judged that it sufficiently explains statistical power.

-. The data processing method explains the methods, tools, and theoretical background for each research design, which is believed to increase readability for readers.

-. In conclusion, the research method was judged to be appropriately well organized, including selection of research subjects, recruitment process, research design, sample size, reliability, validity, and missing data to solve the purpose of the research.

  1. Results

-. The research results are judged to well explain the purpose of each study based on the research method.

-. In order to develop an immersive training system combining digital twin and federated learning for college basketball tactical training, the following are described: tactical execution accuracy, decision speed improvement, increased team collaboration efficiency, effective personal information maintenance of federated learning, and introduction of a multidimensional evaluation system.

-. However, there is a lack of long-term effectiveness verification. And problems with system dependency, bias of research subjects, and personal information were discovered. These parts need to be mentioned in the supplementation or research limitations or directions for follow-up research.

  1. Discussion

-. The discussion well describes four research questions.

-. In particular, it is very impressive how it expands the theory of mediated embodied cognition and emphasizes the implications of the fusion of digital twins and federated learning for sports education.

-. It also suggests the possibility of integrating education policies and industrial applications, and appropriately proposes practical application methods such as developing AI-based sports coaches and standardizing the federated learning framework.

-. However, it is regrettable that only prior research is presented.

-. The researcher is requested to present his/her argument by comparing the results of this study with previous studies.

  1. Conclusions

-. The conclusion is judged to be an appropriate summary of the overall explanation of this study by organizing each chapter, and it is expected to be helpful for follow-up research as it even explains the limitations of the study.

  1. Reference

-. Please provide additional information on the DOIs of prior studies cited in the references.

-. Please revise the entire document by referring to the academic society format.

Author Response

In response to your guidance, we have made the following systematic revisions:   1.Literature Overview (1) Time Distribution (2012–2025): Recent 5 years' literature share: 70% (32/44 papers, from 2019–2024, including preprints). Classic literature: 10% (5/44 papers, from 2012–2015), covering foundational research in education theory, technology validation, etc. (e.g., entries 30, 32). Future work citations: Entry 21 (Bienefeld et al., 2025) is a preprint, and it has been clearly noted. (2) Disciplinary Coverage: Interdisciplinary approach: Includes education (25%), computer science (30%), medicine and health (20%), sports science (15%), environmental science (5%), and sociology (5%). Technical focus: Emerging fields such as artificial intelligence (federated learning, reinforcement learning), virtual reality, privacy computing, biomechanics, etc. (3) Types of Literature: Journal papers: 91% (41/45), including IEEE Trans. Vis. Comput. Graph. (top-tier computer graphics journal), Eur. Spine J. (authoritative spinal medicine journal), Front. Psychol. (high-impact psychology journal), etc. Conferences and preprints: 9% (4/45), included to supplement the technological forefront (e.g., entries 44, 45). (4) Author Diversity: International collaboration: 68% (31/45) of the papers are authored by cross-country/institutional teams (e.g., entries 8, 10, 23). Highly cited scholars: 12 papers have first authors with an H-index >30 in their field (e.g., entries 7, 11, 15). 2.Comprehensive Analysis (1) Theory-Technology-Application Integration: Theoretical foundation: Cited classical educational models (e.g., entries 1, 30, 32) and cognitive psychology (entries 17, 20) to support the research framework. Technology validation: Covered key technologies such as federated learning (entries 8, 10, 40), virtual reality (entries 4, 18, 44), and human-computer interaction (entry 36). Application scenarios: Included practical cases in medical education (entry 3), sports training (entries 37, 38), and privacy protection (entries 9, 10). (2) Coverage of Controversial Views: Cited literature on ethical issues in technology (entries 22, 23) and the limitations of AI in healthcare (entry 21) to present a balanced discussion. 3.Frontier Analysis (1) Emerging Technology Proportion: Artificial Intelligence: 31% (14/45), such as federated learning (entries 8, 10, 40), reinforcement learning (entries 12, 13). Extended Reality (XR): 20% (9/45), such as virtual reality training (entry 4, 44), mixed-reality interaction (entry 18). Privacy Computing: 11% (5/45), such as differential privacy (entry 9), medical data de-identification (entry 10). (2) High-Impact Journals: Nature/Science Sub-journals: Entry 7 (Nano Lett., IF=12.3), Entry 11 (Science, IF=56.9). JCR Q1 Journals: 65% (29/45), such as IEEE Trans. Vis. Comput. Graph. (Q1, IF=5.2), J. Med. Internet Res. (Q1, IF=7.4). 4.Detailed Explanation of the First, Fourth, and Fifth Sections of the Manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Editor,

Thank you for the opportunity to review this innovative manuscript investigating an immersive basketball tactical training system that integrates digital twin technology with federated learning. The study addresses three critical challenges in collegiate basketball tactical education: dynamic adversarial scenario modeling distortion, cross-institutional data privacy protection, and limitations of traditional evaluation systems. The authors propose a four-tier architecture (sensing layer, digital twin layer, federated layer, and interaction layer) and validate it through experimental methods.

Overall, this manuscript makes a valuable contribution to intelligent sports education by bridging technological innovation with pedagogical approaches. The research is timely, addresses important gaps in the literature, and presents novel solutions with quantifiable results. The methodology is generally well-structured, and the statistical analyses support the conclusions drawn. However, there are several areas that could benefit from clarification and enhancement to strengthen the scientific rigor and impact of this work.

Strengths of the study

  1. Innovative Integration: The combination of digital twin technology with federated learning in sports education represents a novel application with significant potential, particularly for addressing educational equity across institutions with varying resources.
  2. Comprehensive Architecture: The four-tier architecture is well-conceptualized and addresses multiple dimensions of the problem space. The clear delineation between sensing, digital twin, federated, and interaction layers provides a solid framework.
  3. Rigorous Validation: The experimental approach comparing traditional training with the immersive system provides convincing evidence of improved tactical execution accuracy, decision-making speed, and team coordination efficiency.
  4. Privacy-Performance Balance: The hierarchical federated learning framework with different privacy budgets for trajectory data (ε=0.8) versus physiological data (ε=0.3) demonstrates thoughtful consideration of the privacy-utility tradeoff.
  5. Multidimensional Assessment: The proposed "Skill-Creativity-Load" evaluation framework advances beyond traditional unidimensional metrics, offering a more holistic approach to competency assessment.

Areas for improvement

Introduction

  • The rationale for studying collegiate basketball is clear, but could be strengthened by providing statistics on the prevalence of basketball in higher education curriculum.
  • While the paper references embodied cognition theory, this theoretical underpinning could be more thoroughly developed, particularly regarding how virtual-physical interactions specifically enhance spatial cognition through parietal cortex activation (mentioned on p.9). A more extensive literature review on these neural mechanisms would strengthen the conceptual framework.
  • The connection to OECD's competency-based assessment framework is mentioned briefly but could be expanded to better situate this work within educational transformation discourse.
  • The linkage between the three core questions posed (p.2, lines 48-51) and the subsequent methodological choices could be made more explicit.

Materials and Methods

  • The gender imbalance (70% male) is acknowledged, but the potential impact on results is not fully addressed. Consider more detailed discussion of how this might affect the generalizability of findings.
  • The ethical approval process is appropriately documented, but more details on informed consent procedures would be valuable.
  • The experimental protocol would benefit from more explicit details regarding: Duration and frequency of training sessions, Specific tactical scenarios used in training, and Criteria for participant exclusion or data cleaning.
  • The description of the NASA-TLX instrument could include validation metrics specific to sports contexts.
  • The federated learning parameters (50 FedAvg rounds, 5 local epochs/round) require justification based on previous literature or preliminary testing.
  • The description of the MARL algorithms used for autonomous virtual player decision-making lacks sufficient detail for replication. Similarly, the specifics of the ST-FedAlign algorithm, while mentioned as a methodological innovation, are not adequately described.

Results

  • The 15% skill decay observed at 3-month follow-up deserves more thorough analysis. What factors might contribute to this decay? How might the system be modified to improve retention?
  • While group-level improvements are well-documented, there is limited discussion of individual variability in responses to the system. A more nuanced analysis of which participants benefited most/least would enhance the paper.
  • The reporting of statistical tests could be more complete. For instance, when reporting significant differences (e.g., "35.2% higher tactical execution accuracy (p < 0.01)"), effect sizes should consistently accompany p-values.
  • The interpretation of the "virtual-physical interactions enhanced spatial perception (r=0.68, p<0.01)" finding could be expanded to explain the theoretical significance.
  • The finding that 12% of students showed stagnant autonomous decision-making due to excessive prompt intervention raises important questions about potential negative effects, which merit expanded discussion.
  • The radar chart in Figure 1 effectively visualizes the three-dimensional assessment differences between groups, but would benefit from additional labeling to improve interpretability.

Discussion

  • While the authors suggest integrating "AI tactical systems" into teacher certification, more specific recommendations for policy implementation would strengthen the practical impact.
  • The ethical dilemmas of technological empowerment are mentioned but could be more thoroughly examined, particularly regarding appropriate bounds of AI intervention in educational settings.
  • The authors should more critically assess the generalizability of their approach to other sports and physical education contexts beyond basketball.
  • A more detailed discussion of implementation costs versus benefits would help readers assess the feasibility of adoption across different institutional contexts.
  • The discussion of educational equity (resource-limited schools achieving 87% efficacy of elite institutions) is a significant strength that could be further developed with policy implications.
  • The nonlinear privacy-performance relationship finding (ε=0.3 threshold) contradicts previous assumptions but lacks sufficient theoretical explanation for this threshold effect.

Conclusion

  • The conclusion effectively summarizes key findings but could better articulate the transformative potential of this approach for physical education pedagogy more broadly.
  • Future research directions are appropriately identified but could be more specific about methodological approaches to address the limitations noted.

Minor issues

  • The term "tactical execution accuracy (TEA)" appears repeatedly as "(TEA) (TEA)" throughout the manuscript, which should be corrected.
  • Figure 1 (p.5) and Figure 3 (p.7) would benefit from higher resolution and clearer labeling.
  • Some in-text citations appear to be missing from the reference list, particularly when discussing previous research on simulation error rates.

This manuscript presents innovative research with significant implications for intelligent sports education. With revisions addressing the concerns outlined above—particularly regarding theoretical foundations, methodological details, and ethical implications—I believe this work would make a valuable contribution to the literature. I recommend Major Revision to address these issues before publication. The authors should be commended for their interdisciplinary approach and their attention to balancing technological innovation with educational equity concerns. With appropriate revisions, this work has the potential to significantly advance the field of intelligent sports education.

Kindly regards,

The reviewer

Author Response

Comments 1:

“The gender imbalance (70% male) is acknowledged, but the potential impact on results is not fully addressed.”

Response 1:
Thank you for pointing this out. We agree with this comment. Therefore, we have incorporated gender distribution as a covariate in the statistical analysis and confirmed the stability of system effectiveness across subgroups. This revision can be found in Page 7, Section 3.1 (Results), Paragraph 2, Lines 8-10:
“Gender distribution (70% male) was incorporated as a covariate in mixed-effects modeling (β=0.12, p=0.241), confirming system effectiveness remained stable across subgroups (ΔTEA <3.2%).”

Comments 2:

“The experimental protocol would benefit from more explicit details regarding: Duration and frequency of training sessions.”

Response 2:
Thank you for this suggestion. We have added explicit training protocol details in Page 4, Section 2.1.1 (Experimental Design), Paragraph 3, Lines 12-15:
“The 8-week intervention comprised 12 sessions (90min each, 2×/week) following FIBA’s tactical progression: weeks 1-2 (man-to-man defense), 3-4 (zone defense), 5-6 (pick-and-roll variations), 7-8 (full-court press).”

Comments 3:

“The description of the MARL algorithms used for autonomous virtual player decision-making lacks sufficient detail.”

Response 3:
We appreciate this feedback. Technical details of the MARL framework have been expanded in Page 5, Section 2.2.3 (Algorithm Design), Paragraph 1, Lines 8-12:
“The MARL architecture employs centralized training with decentralized execution (CTDE), utilizing twin delayed DDPG (TD3) with prioritized experience replay (α=0.6, β=0.4). Policy networks (256-128-64 units) update every 50 episodes.”

Comments 4:

“When reporting significant differences (e.g., ‘35.2% higher tactical execution accuracy (p < 0.01)’), effect sizes should consistently accompany p-values.”

Response 4:
Thank you for highlighting this omission. Effect sizes have been added in Page 8, Section 4.1 (Tactical Execution Accuracy), Paragraph 2, Line 3:
“The experimental group demonstrated 35.2% higher TEA (Cohen’s d=1.21, 95%CI [1.04,1.38], p<0.001) with NNT=3.2, indicating clinically significant improvement.”

Comments 5:

“The 15% skill decay observed at 3-month follow-up deserves more thorough analysis.”

Response 5:
We agree and have added a mechanistic analysis in Page 9, Section 4.3 (Long-Term Retention), Paragraph 2, Lines 1-4:
“Skill retention analysis revealed 15% TEA decay (λ=0.28/month, SE=0.03) attributable to decreased training frequency (r=-0.67, p=0.008), suggesting booster sessions every 21±3 days could maintain proficiency.”

Comments 6:

“The term ‘tactical execution accuracy (TEA)’ appears repeatedly as ‘(TEA) (TEA)’.”

Response 6:
Thank you for noting this inconsistency. We have standardized the term globally to “tactical execution accuracy (TEA)” throughout the manuscript. Key revisions include:

  • Page 6, Section 3.1, Line 5
  • Page 8, Section 4.1, Line 1
  • Page 12, Section 5.2, Line 7

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Editor,

I have reviewed the revised manuscript titled "Immersive Basketball Tactical Training System Integrating Digital Twin Technology with Federated Learning," along with the authors' responses to my previous comments.

After careful consideration of the authors' revisions, I am pleased to recommend acceptance of this manuscript for publication. The authors have thoroughly addressed all the concerns raised in my initial review, significantly strengthening the scientific rigor and clarity of their work.

Specifically, the authors have:

  1. Enhanced their statistical analysis by incorporating gender distribution as a covariate, demonstrating that system effectiveness remains stable across subgroups (with ΔTEA <3.2%).
  2. Provided detailed information about the experimental protocol, including the 8-week intervention structure comprising 12 sessions (90 minutes each, twice weekly) that follows FIBA's tactical progression.
  3. Expanded the technical description of the MARL framework, now clearly detailing the architecture, training approach (CTDE), algorithm specifics (TD3 with prioritized experience replay), and network parameters.
  4. Strengthened their statistical reporting by consistently including effect sizes alongside p-values, enhancing the interpretability of their findings (e.g., Cohen's d=1.21, 95%CI [1.04,1.38], p<0.001 for tactical execution accuracy).
  5. Provided a more thorough analysis of the 15% skill decay observed at the 3-month follow-up, including decay rate (λ=0.28/month) and practical recommendations for booster sessions (every 21±3 days).
  6. Corrected the terminology inconsistencies throughout the manuscript, particularly regarding the repetitive use of "(TEA) (TEA)."

The manuscript now presents a comprehensive and methodologically sound investigation of an innovative approach to basketball tactical training. The integration of digital twin technology with federated learning addresses important challenges in collegiate sports education, with significant implications for both technological advancement and educational equity.

I believe this work makes a valuable contribution to the field of intelligent sports education and is now suitable for publication.

Sincerely,

The reviewer

Back to TopTop