Objective This study examined the longitudinal effects of Home-visit sports service on loneliness and social isolation in children and adolescents, focusing on social connectedness as a mediator, and assessed its predictive value for individual-level psychosocial outcomes.Methods A three-wave longitudinal design was employed, surveying 627 adolescents from five Chinese provinces at one-month intervals. The Co-linearity of Latency and Path Moderation (CLPM) method validated temporal causality among variables, while machine learning algorithms (Ridge regression and XGBoost) assessed predictive gains at the individual level.Results Cross-lagged analysis revealed that prior Home-visit sports service participation significantly and positively predicted subsequent social connectedness. Specifically, the path coefficient from Phase 1 to Phase 2 was β = 0.18 (p < 0.001); The predictive effect of Phase 2 on Phase 3 was also significant, with coefficients β = 0.06/0.07 (p < 0.05). Social connectedness exhibited stable longitudinal mediating effects in the relationships between Home-visit sports service and loneliness/social isolation. Machine learning analyses confirmed that incorporating Phase 2 data into Phase 1-only models improved predictive accuracy for T3 outcomes, with ΔR² = 0.171 for loneliness and ΔR² = 0.168 for social isolation. Stratified analyses showed that adolescents with higher baseline risk derived the greatest predictive benefit from sustained Home-visit sports service participation.Conclusion Home-visit sports service effectively reduces loneliness and social isolation by strengthening social connectedness. Sustained participation provides psychological and social resources, particularly for high-risk youth, supporting precision interventions. Machine learning predictions further demonstrate the practical utility of continuous engagement, highlighting the value of Home-visit sports service in promoting adolescent psychosocial well-being and reducing mental health risks.