<![CDATA[Purpose of the study: This study investigates the effectiveness of a deep learning–based instructional programme for sprint running among Phase E junior high school students (aged 13–15 years). The primary aim was to determine whether a structured, reflective, technology-enhanced instructional cycle significantly improves students’ sprint technique proficiency, biomechanical knowledge, motor coordination, and sport-related self-efficacy. Materials and methods: A quasi-experimental one-group pre-test/post-test design was employed with 32 Phase E students (17 male, 15 female; mean age = 13.9 ± 0.72 years) selected via purposive sampling at SMP Negeri 35 Medan, Indonesia. An eight-week deep learning programme (16 sessions × 80 min) integrating video analysis, biomechanical instruction, guided peer reflection, and iterative corrective feedback was implemented. Outcome measures included the Sprint Technique Assessment Rubric (STAR; ICC = 0.91), Biomechanical Knowledge Test (BKT; Cronbach’s α = 0.83), and the Physical Education Self-Efficacy Scale (PESES; α = 0.79). Wilcoxon signed-rank tests, paired-samples t-tests, and Cohen’s d effect sizes were applied (α = .05; Bonferroni corrected αadj = .008. Results: All outcome measures improved significantly (p < .001). The overall composite score increased from M = 49.5 (SD = 7.4) to M = 80.5 (SD = 6.8), a gain of 62.6%. Effect sizes ranged from d = 2.56 (stride rhythm stability) to d = 4.61 (biomechanical knowledge), all exceeding Cohen’s large-effect benchmark. Conclusions: The deep learning–based instructional model is an effective, scalable approach for sprint running instruction at the junior high school level, simultaneously enhancing technical proficiency, declarative biomechanical knowledge, motor coordination, and self-efficacy. Physical educators and curriculum developers are encouraged to integrate reflective, video-mediated instructional cycles into school athletics programmes.]]>
