<p>Problem sports betting is an increasing public health concern, especially among young adults who are developmentally vulnerable to gambling-related harm. Although past studies have identified correlates of sports betting harm, most have not accounted for the potential influence of wide ranges of other variables in multivariable models. This study applied machine learning to identify the most influential predictors of problem sports betting in a sample of 221 young adults (ages 18–29; 77% male) from 35 U.S. states who reported recent engagement in sports betting. Two machine learning algorithms, regularized regression and random forest, were used to predict scores on the Problem Gambling Severity Index–Sports Betting (PGSI-SB) from a broad set of 80 candidate predictors. Nested cross-validation was used to tune model parameters and obtain unbiased out-of-sample performance estimates in this modest sample. The regularized regression model reduced out-of-sample prediction error by 39% relative to a naïve baseline model, whereas the random forest reduced error by 34%. Across both models, sports betting motives, social functioning, amount wagered, and perceived harm from sports betting emerged as the most influential predictors. Given that both models left considerable prediction error unexplained, it is likely that important determinants of problem sports betting were not captured in the available feature set. Findings demonstrate the utility of machine learning in identifying key risk factors across multiple domains and highlight the importance of focusing on sports betting motives and social functioning in prevention efforts. This study contributes to a growing evidence base by identifying factors that are uniquely predictive of problem sports betting among young adults.</p>