Purpose: In this study, we compared the influence of movement specificity during velocity–load jump testing to predict on-ice acceleration performance in elite speed skaters. Methods: Elite long-track speed skaters (N = 27) performed velocity–load testing with 3 external loads during unilateral horizontal jumping, lateral jumping, and bilateral vertical countermovement jumping. For the unilateral tests, external load conditions were set to 10 N, 7.5% and 15% of external load relative to body weight. For the countermovement jumping, load conditions were body weight and 30% and 60% of external load relative to body weight. On-ice performance measures were obtained during maximal 50-m accelerations from a standing start, including maximal skating speed, maximal acceleration capacity, and maximum horizontal power. The 100-m split time from a 500-m race was also obtained. Regularized regression models were used to identify the most important predictors of on-ice acceleration performance. In addition to regularized regression coefficients, Pearson correlation coefficients (r) were calculated for all variables retained by the model to assess interrelationships between single predictors and on-ice performance measures. Results: The countermovement jump with 30% of body mass demonstrated the strongest association with maximal skating speed, maximum horizontal power, and 100-m time (regularized regression coefficient = .16−.49, r = .84−.97, P < .001). Horizontal jump with 15% of body mass was the strongest predictor of maximal acceleration capacity performance (regularized regression coefficient = .08, r = .83, P < .001). Conclusions: The findings of this study suggest that mechanical specificity rather than movement specificity was more relevant for predicting on-ice acceleration performance.