This paper researches strategic decision-making in simulated curling using advanced computational methods. A simulated solution could revolutionize tactical discussions in curling and enhance the strategy of professional curlers. By focusing on position analysis and the integration of supervised learning algorithms, the study aims to enhance the accuracy and effectiveness of strategic calls in curling simulations. The research reveals that supervised neural networks and graph neural networks outperform heuristic approaches and exhibit high accuracy in complex game scenarios. However, the study also shows that current methods and datasets seem to be insufficient within the realm of machine learning and pattern recognition to enable an artificial system to consistently outperform professional curlers (particularly in complex mid-end positions). The findings also indicate that a hybrid approach, combining methods from statistical and structural pattern recognition, might better address the challenges of the continuous space of curling situations.