This research bridges the gap between virtual simulations and real-world accident analysis through a comprehensive study using the CARLA Autonomous Driving Simulator. It employed 128 distinct driving scenarios varying five critical parameters: weather (Clear/Rain), road geometry (Straight/Intersection), obstacle type (Pedestrian, Vehicle, Log, None), target speed (20–140 km/h), and obstacle distance (10 m, 40 m). Each configuration was executed in three trials, yielding 384 simulation runs and a dataset of 95,938 data points, capturing vehicle kinematics, proximity and time-to-collision (TTC). Analysis of eight graphs revealed that rainy intersections posed the highest risk, with collision probabilities reaching 68%, and pedestrian obstacles were the most hazardous. Validation against traffic databases from MoRTH (2023) and NHTSA (2022) showed an 87% correlation, confirming strong external validity and simulation accuracy.(calculations reported in section 2.3). These findings demonstrate, simulation-based risk models can reliably predict real-world collision trends, offering scalable framework for proactive safety testing and  foundation for AI-enhanced traffic management and ADAS training. The study  highlights the potential for integrating these insights into urban planning, insurance modeling, and educational platforms . It demonstrates that simulation-based models can achieve up to 87% real-world correlation, establishing CARLA( virtual stimulator) as a viable platform for scalable, risk-focused traffic research