Terah Skillings

This project explores whether early-stage quadriceps strength can indicate rehabilitation plateaus after ACL reconstruction. Using longitudinal data from 18 athletes across 350+ rehab visits, we focused on the first 60 days post-op and defined plateau based on limb symmetry index (LSI) and stalled progression. A multivariate logistic regression incorporating early quadriceps torque (Nm/kg) and LSI improved in-sample discrimination (AUC 0.71) compared to LSI alone (AUC 0.59). Early absolute torque was the most influential predictor, suggesting that strength deficits, even in athletes who appear symmetrical, may signal risk for stagnation. These findings provide evidence that monitoring early quadriceps torque could help guide targeted interventions in ACL rehabilitation.

This practicum project uses ensemble and machine learning methods to predict which NHL teams will make playoffs using historical NHL data from the 2008 season to 2024. It can be categorized by team performance, roster investment and player availability. Focusing on early season indicators, the analysis used the first twenty games of each season. This project follows the data science workflow, including data cleaning, exploratory data analysis, feature engineering and model development. Multiple machine learning models were used and optimized. These models included both traditional machine learning models to ensemble models such as XGBoost and stacking. Model performance was evaluated on accuracy, ROC-AUC, F1 score, and cross-validation. Results showed the stacking model to have the overall best performance achieving an ROC-AUC score of 0.84. Feature importance showed that goal differential and points are key predictors. Using historical NHL data, this project demonstrates how data science methods can be applied to better understand team performance and playoff outcomes in the NHL.