Martin Toudjarski

An AIC-Based Analysis of Novel Features for Predicting FIS World Cup Ski Performance

Started Nov. 4, 2024

Abstract or project description

This study aimed to identify which technical features most effectively predict athlete performance rankings in competitive Alpine Skiing, using data from the 2023 FIS World Cup downhill race in Kitzbühel. We proposed, manually annotated, and evaluated four movement-based metrics–upper body stability, pressure control, turn initiation timing, and jump landing efficiency–to determine which aspects of technique are most strongly associated with athlete rank. The primary research question was: Which individual or combined technical features best explain variability in skier ranking? We hypothesized that features reflecting upper-body stability and jump landing efficiency would significantly improve model performance. To test this, we evaluated all 16 possible combinations of the four features using linear regression and compared them pairwise using Akaike Information Criterion (AIC). Results showed that upper body stability had the strongest influence, improving model performance in all eight comparisons it was involved in, followed by jump landing efficiency (7/8). The full model including all four features yielded an R2 of 0.796 and the lowest AIC was 146.2. These findings suggest that upper body stability and jump-landing efficiency are critical determinants of competitive performance in elite-level skiing. Coaches may benefit from emphasizing upper-body coordination and core stability in training programs. Overall, this study bridges the gap between data modeling and applied coaching by identifying the technical features most predictive of race-day success.