Affective forecasting refers to the process of predicting emotional reactions to future events. It plays an important role in decision making, but is also prone to errors, such as the ‘impact bias’: a tendency to overestimate the intensity of future reactions. The impact bias has been considered evolutionarily adaptive, as it performs a protective function in motivating people to avoid risky behaviour. Affective forecasting may be qualitatively different in a risk-taking population such as problem gamblers (PGs). In particular, PGs may fail to show the impact bias. This study was the first to examine affective forecasting in PGs. PGs (N = 25) and controls (N = 29) were asked to predict how they would feel after completing a guessing task. As hypothesized, controls exaggerated how bad they would feel after losing at the task, whereas PGs accurately predicted their reactions. Encouraging PGs to focus on anticipated emotions may be a novel target for treatment interventions.