We present a machine-learning-based system for automated soccer analytics, leveraging computer vision techniques to detect players, differentiate teams, determine ball possession, and generate graphical representations of player movement and influence. Traditional sports analytics rely on manual observation, which is time-consuming and prone to human error. To address this, a You Only Look Once (YOLO) based object detection model was created for identifying players, referees, and the ball, coupled with a multiple-object tracking (MOT) algorithm to maintain detection continuity across frames. For team differentiation, a novel method utilizing SigLIP embeddings, UMAP dimensionality reduction, and K-Means clustering ensures robust classification, even in the presence of variations in lighting and jersey colors. Additionally, a statistical analysis framework was developed to calculate ball possession between the teams. To enhance tactical insights, the system generates graphical outputs, including radar views, Voronoi diagrams, and heat maps, providing an intuitive visualization of game dynamics. Experimental results demonstrate an average team classification confidence of 87.54% across test videos, with ball detection accuracy varying based on occlusions and camera perspectives.