F1oresight: Predicting Final Position and Race Winners in Formula 1 Using Random Forest and Historical Performance Data

Abstract

Formula One (F1) is a highly data-intensive motorsport where race outcomes are influenced by driver skill, car performance, team strategy, and track-specific factors. Predicting race results is challenging due to the inherent unpredictability of mechanical failures, weather conditions, and in-race decisions. F1oresight, a machine-learning model, models the relationship between qualifying performance, recent driver consistency, and race outcomes. Using historical race and qualifying data collected via the FastF1 API, features such as average position over the last three races and grid advantage were engineered to capture recent driver performance and track effects. Random Forest models were trained to predict finishing positions. By isolating stable performance indicators from unpredictable race-day variables, the models provide insights into the drivers’ raw pace and expected outcomes. Results demonstrate that machine learning can effectively estimate race rankings and winning probabilities.