Electronic gaming machines (EGMs) are widely used and the gambling product most commonly associated with harmful gambling. Understanding factors that increase the risk of problematic EGM play is therefore important. Previous studies into risk factors for EGM gambling have used measures of problem gambling based on an individual's total gambling activity, which therefore do not distinguish harmful gambling specifically associated with EGMs. This study used an EGM-specific measure (PGSI-EGM) to achieve its aim of identifying risk factors specifically associated with problematic EGM play. By removing nuisance effects from other gambling forms that higher-risk gamblers typically engage in, this approach provides a more accurate assessment of the determinants of EGM-related problems. An online survey was completed by 1932 at-least monthly EGM players in Australia. It measured demographics, EGM gambling behaviour, motivations, gambling urges, gambling fallacies, trait self-control, alcohol misuse, and the PGSI-EGM. A penalised regression model identified the most important proximal predictors of higher-risk EGM gambling as: higher gambling urges, higher levels of erroneous cognitions, playing EGMs more frequently, higher session expenditure, longer sessions, usually playing EGMs alone, and playing EGMs in more venues. Lower trait self control was the strongest distal determinant. Higher-risk EGM players tended to be younger, male, more educated, never married, to have higher (although still modest) incomes, and be more likely to have alcohol problems. These findings can inform interventions such as treatment, consumer education and venue interventions.