Group affective tone is an emergent state that can be shared by group members during interdependent tasks. Groups can experience positive group affective tone (PGAT), a shared feeling of, e.g., excitement, enthusiasm, or activation, as well as negative group affective tone (NGAT), a shared feeling of, e.g., distress, anxiety, and hostility. So far, previous cross-sectional research suggests that PGAT and NGAT are related to team performance outcomes. However, little is known about how the dynamic and fluctuating group affective states are related to team performance over an extended period of time. Therefore, the current study investigated the relation between PGAT, NGAT, and performance over the course of 34 software engineering projects. We hypothesized that PGAT is positively related to team performance, whereas NGAT is negatively related to team performance. Based on the punctuated equilibrium model and the feeling-as-information theory, we expected that these associations become stronger in the second half of the project. Using week-level design with 165 participants in 34 software engineering teams, we repeatedly assessed PGAT, NGAT, and team performance over 14 weeks. Data were analyzed using multilevel structural equation modeling. As expected, PGAT was positively related to team performance, whereas NGAT was negatively related to team performance -between teams over the course of the projects as well as within teams over time. More importantly, the weekly relationships were stronger in the second half of the project. Our study indicates that weekly variations in group affective tone are more relevant after projects reach a temporal midpoint. We discuss theoretical and practical implications for project teams.