Sex Differences And Predictors Of Jump Height And Injury Risk: XGBoost Predictive Modeling

It has been widely accepted that countermovement vertical jump (CMVJ) is correlated with strength and performance when assessing lower-body performance. Given the growing availability of commercial force plates, important measures like reactive strength index (RSI) are more readily accessible. PURPOSE: Sparta Force Plates supplies multifaceted and proprietary performance metrics for in-depth analyses of athletes’ strength, explosiveness, balance, and injury risk (IR). However, few studies have evaluated the importance of these performance metrics in predicting CMVJ or IR. Therefore, this study aims to evaluate and differentiate the importance of Sparta’s jump metrics in predicting CMVJ and IR using XGBoost machine learning (ML) algorithm. METHODS: Data from 233 (56.7% male, 43.3% female) collegiate athletes from 11 different sports were collected. After a warm-up trial, participants performed 3 CMVJs on the Sparta Force plate. The best trial was retained for the analysis. Sex differences across 28 different jump measures were analyzed using the Kruskal-Wallis test. The XGBoost Classifier and the XGBoost Regressor were used to determine feature importance for IR classification and CMVJ, respectively. Using 10-fold cross-validation, models were trained on 50% of the data and selected via parameter optimization. The IR model was evaluated via ROC, PRC, and F1 scores. The CMVJ model was evaluated via MSE and R2. RESULTS: There were differences in feature selection and importance between males and females for IR and CMVJ (Figure 1). The CMVJ model in males had an R2 of 93% and a MSE of 0.05. The CMVJ model in females had an R2 of 98% and a MSE of 0.01. mRSI and RSI were top predictors of CMVJ in females only. CONCLUSIONS: XGBoost ML algorithm can assist in selecting important predictors of CMVJ and IR when many performance metrics are available, allowing sport performance coaches to develop highly specific training programs.