Sports performance analysis has become increasingly crucial for coaches and analysts aiming to enhance player development and optimize game strategies. However, acquiring extensive, real-world player performance data is often limited by privacy concerns and accessibility. This study addresses the problem by developing a simulation-based system that generates a synthetic dataset for sports performance analytics. The dataset includes diverse features such as player demographics, game metrics, physical and psychological attributes, and environmental factors, creating a comprehensive view of player performance metrics. The method involves generating synthetic data that mimics realistic distributions for each feature. This dataset was used to train machine learning models with two objectives: (1) a classification model to predict if a player is “In-Form” based on recent performance metrics, and (2) a regression model to estimate a player’s Game Impact Score. Feature engineering, categorical encoding, and scaling were applied to optimize the dataset, and model tuning was performed using Gradient Boosting and RandomizedSearchCV. Evaluation metrics such as accuracy, F1-score, ROC-AUC, and mean squared error (MSE) were employed to assess model performance. Results showed that key features like Passing Accuracy, Weight, and Match Result significantly influenced the classification task, with a weighted average F1-score of 0.56 and an AUC of 0.58. For the regression task, the Game Impact Score prediction achieved an MSE of 228.4, though the R-squared score indicated that further feature interactions or advanced models may be necessary for precise predictions. The implications of this study highlight the potential of synthetic data to address data scarcity in sports analytics and demonstrate the feasibility of using machine learning to predict player performance metrics. Future work could involve integrating real-world data and exploring deep learning techniques to enhance model robustness and generalizability.