As esports continue to professionalise, understanding the cognitive mechanisms underpinning expert performance becomes increasingly important. However, despite its rapid growth, esports research remains comparatively limited, particularly regarding the cognitive mechanisms underlying expert decision-making. This study explores how a team of semi-professional League of Legends (LoL) players manage the challenges of Team Fights (TFs), moments that heighten the most tactical, cognitive, and emotional demands of a match, and the heuristics they employ during gameplay. To address this need, the study had two main objectives: (1) to identify the heuristics players use when making decisions in TFs; and (2) to examine whether these heuristics are shared among team members, potentially reflecting a Shared Mental Model (SMM) of gameplay within the team. To gather expert insights, we conducted an Applied Cognitive Task Analysis (ACTA) involving five players from the same semiprofessional LoL team, comprising Task Diagram, Knowledge Audit, and a simulation focus group phase, yielding a total of 11 hours of interview data. We applied a thematic analysis to the Task Diagram phase, whereas template analysis was applied to the Knowledge Audit interviews and the final simulation focus group. This analysis revealed a total of 22 heuristics shared by the players, including a potentially novel Inhibitory Heuristic, which aligned with five core cognitive demands experienced during a TF. The findings indicate that these players rely on shared heuristics to guide their decision-making through a SMM during TFs. By developing a taxonomy of heuristics and illustrating their relation to key cognitive demands, this study offers novel insights into the performance psychology of esports. It also provides a foundation for creating training interventions aimed at improving real-time decision making in LoL’s TFs, supporting the alignment of decision-making processes across players and enhancing the management of cognitive load in high-pressure team performance contexts.