Soccer teams frequently adapt their playing style based on the game context. For example, once a team takes the lead it may sit a little deeper and take fewer offensive actions. This paper focuses on automatically inferring how a team adjusts its in-possession playing style to the game context. Our approach hinges on analyzing passing dynamics as a proxy for a team’s in-possession playing style and involves two components. First, we enhance the SoccerMap architecture by incorporating game state variables and team identities, resulting in a more context-aware model that can predict passing tendencies for individual passes. Second, we introduce an encoder-decoder architecture to isolate the impact of game state on passing decisions, providing insights into how teams adapt their strategies in various contexts. By studying passing patterns and decisions made by teams in different game contexts, we aim to provide clubs with concrete insights that can inform tactical adjustments and optimize gameplay in various scenarios.