Past research on gambling behavior has focused largely on cognitive and affective explanations, but has not explored the influence of past gambling behavior or habit. In this study, it was hypothesized that gambling behavior followed predictable patterns based on past behavior. We found that these patterns could be described by the negative binomial distribution (NBD) and NBD–Dirichlet models at game and category levels, respectively. An empirical study using the 1999 Australia Productivity Commission survey dataset provided some support for the usefulness of these models in describing gambling consumption. The paper explores the implications for identifying problem gambling and influencing gamblers’ behavior.