INTRODUCTION: Alpine skiers are faced with too many skis to choose from during shopping. The information about the skis provided by the manufacturers is also sparse, while reviews are subject to several biases (e.g., geographical, reviewer physical characteristics and skills, preferences, fatigue) and challenges (e.g., snow types, limited testing time, reviewers agreements). This limits the reliable comparison of thousands of new skis each year. As large databases of measurements are now becoming available [1,2], it is possible to envision the automated and deterministic categorization of alpine skis, as well as the prediction of their on snow performance. METHODS: To achieve that goal, different categorization and performance evaluations were reversed-engineered. On one hand, the SoothSki database of alpine ski measurements was used. This database includes detailed geometry, mass and bending/torsional stiffnesses measurements of more than 5000 commercially available skis. On the other hand, the categorization of 865 skis from evo.com (e.g., Powder, All-mountain, Carving) and the rankings from Blister Gear Review were used (36 best-to-worst rankings of over 300 skis). A procedure based on Decision Trees was used to predict categories [3], while a procedure based on Elastic Net was used to predict rankings [4]. RESULTS/DISCUSSION: Results show that simple and robust rules can be found based on mechanical measurements to categorize and predict the on-snow performance of alpine skis. The rules found generally include less than three mechanical parameters and reached classification accuracy and mean absolute rank errors of respectively 97.5% and 15%. These simple rules can thus be easily analyzed to obtain a better understanding of complex on-snow performance and the reviewers` language. Furthermore, to better suit each skier`s preferences, the rules can be easily fine-tuned. Finally, the automated processes could be repeated on other datasets to better represent different views (e.g., different skiing cultures). CONCLUSION: Novel methods were developed to classify and predict on-snow performance based on mechanical properties. The rules found can be used to describe skis more easily and uniformly across brands, educate skiers, and simplify their shopping experience.
© Copyright 2025 10th International Congress on Science and Skiing, January 28 - February 1, 2025, Val di Fiemme, Italy. Alle Rechte vorbehalten.