Statistical Modeling Final Project

For this project, I analyzed professional League of Legends match data to model and predict game duration using statistical learning methods. After cleaning and structuring a large dataset of international matches, I explored early- and mid-game performance metrics and built multiple predictive models, including OLS, LASSO, and Principal Components Regression. I compared these models using a train–test split and found that LASSO achieved the lowest RMSE by selecting the most informative predictors and reducing multicollinearity. The results highlight how a small set of tempo-driving variables, such as kill pace and objective control, can reliably forecast match length.