Domain-specific contexts promote model-based decision making for basketball players

Abstract Decision-making is a critical component of basketball performance. Within the framework of Reinforcement Learning (RL), understanding the environment and task structure can facilitate model-based (MB) behavior. This study tested whether domain-specific contexts promote MB decision strategies by using the two-stage task in a 2 (Group: basketball players vs. novices) × 2 (Condition: abstract symbols vs. basketball tactical diagrams) mixed design. Thirty-five national basketball players and twenty-seven novices completed both conditions. A hierarchical one-trial-back logistic regression estimated intercept, reward, transition, and reward × transition coefficients; two-factor repeated-measures ANOVAs assessed coefficients, reaction times, and subjective evaluations. Compared to novices, basketball players exhibited a more MB decision-making strategy across all condition, accompanied by higher subjective understanding. Novices exhibited higher reward coefficients indicated a more model-free (MF) tendency. For basketball players, MB strategies were associated with longer RTs, yet longer RTs in novices did not necessarily indicate MB strategies use. These findings show that expertise facilitates the integration of outcome feedback with the task’s structure, thereby leading to MB decision-making. Incorporating sport-specific displays into training may enhance decision-making performance without relying on explicit structural instruction.