Little is known about how gamblers estimate probabilities from multiple information sources. This paper reports on a preregistered study that administered an incentivized Bayesian choice task to n=465 participants (self-reported gamblers and non-gamblers). Our data failed to support our main hypotheses that experienced online gamblers would be more accurate Bayesian decision-makers compared to non-gamblers, that gamblers experienced in games of skill (e.g., poker) would be more accurate than gamblers experienced only in non-skill games (e.g., slots), or that accuracy would differ in females compared to males. Pairwise comparisons between these types of participants also failed to show any difference in decision weights placed on the two information sources. Exploratory analysis, however, revealed interesting effects related to self-reported gambling frequency. Specifically, more frequent online gamblers had lower Bayesian accuracy than infrequent gamblers. Also, those scoring higher in a cognitive reflection task were more Bayesian in weighting information sources when making belief assessments. While we report no main effect of sex on Bayesian accuracy, exploratory analysis found that the decline in accuracy linked to self-reported gambling frequency was stronger for females. Decision modeling found a decreased weight placed on new evidence (over base rate odds) in those who showed decreased accuracy, which suggests a proper incorporation of new information into one's probability assessments is important for more accurate assessment of probabilities in uncertain environments. Our results link frequency of gambling to worse performance in the critical probability assessment skills that should benefit gambling success (i.e., in skill-based games).