The Effectiveness of a Deep Learning Approach in Enhancing Elementary School Students' Physical Fitness during Physical Education

This classroom action research documented implementation of pedagogical deep learning and within-class changes in physical fitness and participation among 84 fifth-grade students at SDN 2 Pakunden, Blitar City, during 2025/2026. The intervention comprised three iterative cycles with two Physical Education meetings per cycle. Archived records reported a composite fitness score and mastery status at baseline and after each cycle, and observed learning activity during Cycles I-III. Mean fitness scores increased descriptively from 68.12 at baseline to 75.13, 86.16, and 91.45 across the three cycles. The corresponding numbers meeting the school mastery criterion increased from 21/84 (25.00%) to 57/84 (67.86%), 78/84 (92.86%), and 84/84 (100%). Observed activity percentages were 75.08%, 84.67%, and 92.43% in Cycles I-III; no pre-intervention activity measure was available. These temporal changes describe one class cohort and do not establish causal effectiveness because there was no concurrent comparison group, inferential analysis, validated implementation measure, or control for maturation, repeated testing, teacher attention, and cumulative activity. The implementation is best regarded as a locally useful improvement cycle that requires evaluation with standardized measures and a controlled design.