This study set out to understand the association between teaching practices, teacher confidence, competence, self-efficacy, and the resulting student outcomes. Data regarding teaching behaviours were collected via video recording and then evaluated using the MASTER Observation Tool. The information about demographics, self-reported teaching confidence, competence, self-efficacy, and student outcomes was collected using questionnaires. The association between teacher characteristics, and teacher and/or student outcome variables were tested using a one-way analysis of variance (ANOVA). A total of ten primary schools were involved, including 597 children (age range: 10-12 years, grade 4-6) and 16 grade 4-6 PE teachers (with 16 PE classes). Most of the Physical Education (PE) lesson time was spent in training-form activities (60.2% ± 9.1), followed by instructional time (33.1% ± 8.6%), reflection (3.4% ± 2.3%), and warm-up (2.9% ± 2.0 %). It was observed that teaching behaviours and student outcomes were significantly better in urban than rural areas. Smaller class sizes (21-30 children) were found to have more positive feedback than larger ones (41-50 children). PE teachers with more than 10 years of teaching experience reported more teaching competence and self-efficacy than teachers with less than 10 years of experience. PE teachers with class sizes of 21-30 children enjoyed significantly better scores in self-efficacy than classes with 41-50 children. They also scored more highly in confidence and competence than classes with 41-50 and 51-60 pupils. The current study confirmed that teachers dedicated a large proportion of lesson time to PE delivering training-form activities, followed by instructional time. Teaching behaviour and student outcomes were associated with location and class size, but not gender. The study contributes to our understanding of PE instruction in Chinese primary schools and offers preliminary evidence to improve future PE teaching strategies in the country.