The paper introduces the "experience-weighted attraction" (EWA) model, a general framework that combines elements of both belief-based models and choice reinforcement models. EWA allows strategies to be reinforced based on both actual payoffs and hypothetical payoffs for unchosen strategies, with a parameter δ controlling the relative weight of these two types of reinforcement. The model includes two discount rates, φ and ρ, which control the decay of past attractions and the normalization of attractions, respectively. The EWA model is estimated using three sets of experimental data, and the parameters are calibrated to predict holdout samples. The results show that EWA generally outperforms both belief and reinforcement models, though belief models perform better in some constant-sum games. The EWA model is flexible, allowing for a wide range of initial attractions, growth rates, and discount rates, making it a powerful tool for understanding human behavior in games.The paper introduces the "experience-weighted attraction" (EWA) model, a general framework that combines elements of both belief-based models and choice reinforcement models. EWA allows strategies to be reinforced based on both actual payoffs and hypothetical payoffs for unchosen strategies, with a parameter δ controlling the relative weight of these two types of reinforcement. The model includes two discount rates, φ and ρ, which control the decay of past attractions and the normalization of attractions, respectively. The EWA model is estimated using three sets of experimental data, and the parameters are calibrated to predict holdout samples. The results show that EWA generally outperforms both belief and reinforcement models, though belief models perform better in some constant-sum games. The EWA model is flexible, allowing for a wide range of initial attractions, growth rates, and discount rates, making it a powerful tool for understanding human behavior in games.