AI models act as “black boxes,” with decision processes that aren’t visible. This paper presents an explainable AI (XAI) method for a model predicting basketball shot success or failure. The model, trained on body posture sequences using LSTM, identifies what minimal posture adjustments can turn a failed shot into a success through a new technique called Input Gradient Descent . The system provides players with clear visual and verbal feedback, offering practical, transparent guidance to improve performance. This approach could even be generalized to LLMs.