Abstract
Evaluating alpine skis on snow is pivotal for ski development and consumer decision-making, yet it is resource-intensive and hindered by subjective assessments. Leveraging recent extensive ski physical measurements and on-snow ski evaluation metrics, this study proposes an automated methodology that employs elastic net regression, bootstrap resampling, and intelligent feature selection to predict the on-snow performance using a minimal set of physical attributes. Results on 192 skis divided into 10 categories and 29 metrics indicate promising predictive capabilities, with models exhibiting an average Mean Absolute Error rank prediction of 15%. Importantly, the models utilize less than three physical attributes on average, underscoring their simplicity and effectiveness in identifying key performance-defining properties. These findings, to the authors' knowledge, represent the most comprehensive description of ski on-snow performance to date and hold implications for ski design and consumer guidance. Moreover, the automated methodology enables the easy integration of other evaluation sources, facilitating further refinement and validation, while promising to consider the diversity of opinions related to ski on-snow performance assessment. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s12283-025-00511-w.