Background. Despite the increasing pervasiveness of digital entertainment and serious games in organisational life, there is little evidence for the validity of game-based team training and assessment. Aim. The authors used the game, TEAMUP for a series of team training and assessment sessions, while at the same time researching the internal validity of the game for this purpose. Method. A total of 106 sets of data on games played by teams of professionals (police officers, auditors, consultants, etc.) and undergraduates and postgraduates (in aerospace engineering, entrepreneurship, etc.) were gathered for analysis through pre- and post-game questionnaires focusing on constructs for team quality, such as psychological safety and team cohesiveness. In addition, a large quantity of such data as time to complete task, distance and avoidable mistakes were logged to measure in-game team performance. Correlation and regression analyses were conducted to find relationships between team structure factors, team quality constructs and in-game performance measures. Results. The main finding is that the in-game performance measure ‘avoidable mistakes’ (a proxy for task quality) correlates markedly and pervasively with ‘team cohesiveness’. More important, the findings support the premise that in-game assessment can be internally valid for team research and assessment purposes.