When a temporary team faces an unfamiliar task environment, it should particularly benefit from including members who have collaborated before. Although several studies have made this prediction, it has not been supported empirically. We reconcile this discrepancy by distinguishing between prior collaborations in tightly versus loosely coupled roles, defined by the degree of interdependence among collaborators during prior work. Both forms of prior collaboration can help teammates communicate effectively, aiding team adaptation in new contexts. However, prior collaboration in tightly coupled roles also fosters shared routines that may create inertia and impede adaptation. As a result, we argue that the value of tightly coupled experience declines in unfamiliar environments, whereas the value of loosely coupled experience increases. We test these ideas using data from esports, where professional players with varying collaborative histories are randomly assigned to temporary teams. Unanticipated changes to the game exogenously alter teams’ familiarity with their task environments. Consistent with our theory, we find that tightly coupled collaborative experience enhances team performance in familiar task environments but degrades it in unfamiliar ones. Loosely coupled experience provides modest benefits in familiar environments but substantially enhances performance in unfamiliar ones. Overall, our findings suggest that when environments change, tightly coupled experience can act as a faulty roadmap—anchoring teams to outdated routines—whereas loosely coupled experience can serve as a compass that promotes coordination and adaptation.