Performance Prediction and Evaluation of Wind Effects in Ski Jumping using Machine Learning

This study demonstrates a data-driven approach to modeling ski-jump performance using machine learning algorithms trained on publicly available International Ski and Snowboard Federation (FIS) competition data. By modeling jump distance as a function of in-run speed, wind conditions, and hill size, we evaluate the capabilities of four distinct model classes to predict performance, explain the impact of wind, and critique wind point compensation in order to demonstrate a machine-learning-based approach to analyzing the sport. The analysis delineates the strengths of this approach in predicting jump distances and confirming the empirical influence of wind on flight distance. The study also identifies significant drawbacks inherent to using only public data, specifically determining how the absence of athlete-specific properties and high-resolution environmental data constrains the precision of these predictive models.