Promotion of Sports for All (SFA) can contribute to improving health and physical fitness status, and ultimately national competitiveness. The main purpose of this study is to explore predictors for health and physical fitness status by SFA participation or non-participation via machine learning. In particular, all possible variables and respondents of the 2021 National Survey on Sports for All were analyzed, using penalized regression. After data cleaning and missing data imputation, we analyzed 292 variables of 6,535 participants and 212 variables of 2,465 non-participants, respectively. As a result of group Mnet, 14 and 11 groups of variables were selected as key predictors of the participants and non-participants, respectively. Commonly selected predictors included household income, practices for health and physical fitness, variables on sports club, and prerequisites for promoting sports activites. Gender, education level, the degree of safety rule compliance, reasons for not using physical fitness management service, and accompanying participants in sports activities were predictors for SFA participants. Predictors for non-participants included awareness of neighborhood sports facilities and presence of sports clubs willing to join in the future. We recommend that SFA policies should target prospective SFA participants for more effective implementation of SFA policies. In-depth research is warranted on sports instructors’ quality improvement and sports clubs’ roles and usages in SFA.