Objective: Although coach–athlete relationships (CAR) have been widely studied, less is known about how perceived coach behavior (PCB) operates through group-based identity processes and how individual emotion regulation and developmental factors shape this mechanism. Drawing on Social Identity Theory and Emotion Regulation Theory, this study examined a dual-moderated mediation model in which team identification (TI) mediates the association between PCB and CAR, with emotional intelligence (EI) and age defining key boundary conditions. Methods: A cross-sectional questionnaire survey was conducted among 581 athletes from provincial-level competitive sports teams in China. Results: Positive coaching behaviors strengthened CAR indirectly by enhancing athletes’ team identification, whereas autocratic behavior undermined this relationship by weakening TI. Moreover, EI moderated the TI–CAR pathway, such that TI was a stronger predictor of CAR among athletes with lower EI. This moderating effect was further conditioned by age, remaining significant among younger athletes but diminishing and becoming non-significant among older athletes. Conclusion: These findings advance research on coach–athlete relationships by integrating group-based identity processes with individual emotion regulation and developmental perspectives. From a practical perspective, the results suggest that fostering team identity may be particularly effective for younger or lower-EI athletes, whereas more individualized leadership and communication strategies may be more appropriate for older or emotionally intelligent athletes.

During interpersonal interaction, understanding others’ actions requires the brain to continuously integrate information as behavior unfolds. Yet an action may deviate from its expected course and subsequently be corrected, raising the question of whether neural representations reflect only its eventual outcome or also retain information about how it developed. Here, we recorded electroencephalography from 20 expert table tennis players and 21 novices observing dyadic actions involving uninterrupted success, successfully corrected deviations, or goal-disrupting deviations. These conditions allowed us to distinguish information about an action’s preceding course from its eventual outcome. We combined time-resolved neural decoding, temporal generalization, and representational similarity analysis to characterize the differentiation and organization of unfolding actions. Neural activity differentiated the three action courses, with motor expertise selectively modulating their neural separability and temporal generalization. Critically, in experts, neural representational geometry corresponded to the shared preceding deviation of corrected and goal-disrupting actions despite their different outcomes. This correspondence remained significant after controlling for competing theoretical models and low-level visual and motion similarity. These findings suggest that neural representations of others’ actions incorporate information about their preceding course beyond their immediate observable state, highlighting how the brain preserves action history during unfolding dyadic actions.