The Problem Gambling Severity Index (PGSI) is a widely used nine item scale for measuring the severity of gambling problems in the general population. Of the four gambler types defined by the PGSI, non-problem, low-risk, moderate-risk and problem gamblers, only the latter category underwent any validity testing during the scale's development, despite the fact that over 95% of gamblers fall into one of the remaining three categories. Using Canadian population data on over 25,000 gamblers, we conducted a comprehensive validity and reliability analysis of the four PGSI gambler types. The temporal stability of PGSI subtype over a 14-month interval was modest but adequate (intraclass correlation coefficient = 0.63). There was strong evidence for the validity of the non-problem and problem gambler categories however the low-risk and moderate-risk categories showed poor discriminant validity using the existing scoring rules. The validity of these categories was improved with a simple modification to the scoring system.