pitch-prediction

Per-pitcher Binary Classifier Analysis as part of STA4241 Final Project

Setup

Clone the repository then move into the pitch-prediction directory

git clone https://github.com/ZackAllen1/pitch-prediction.git
cd ./pitch-prediction

Open the main.R file and load the necessary packages and helper files in lines 1-8. To properly install the baseballr package you may have to run the following code to install it off of GitHub instead of CRAN:

if (!requireNamespace('pacman', quietly = TRUE)){
  install.packages('pacman')
}
pacman::p_load_current_gh("BillPetti/baseballr")

Insert the names of the pitchers you wish to analyze as "FirstName LastName" elements in the pitcher_names list on Line 11. If you wish to analyze the original five pitchers (Justin Verlander, Zach Eflin, Clayton Kershaw, Blake Snell, and Shohei Ohtani), the proceeding Data Gathering section of code can be skipped since the datasets for those pitchers is already collected.

In the Get Results section, finding the best hyperparameters for GBM using grid-search and 5-fold CV can take 20-30 minutes per pitcher. For the original five pitchers, the best hyperparameters found in the study have been saved in the best_gbm_params.csv file. To use these parameters instead of recalculating, set findBestGBM to FALSE on Line 36. Anytime a new pitcher has their hyperparameters calculated they will be automatically saved in best_gbm_params.csv so they do not have to be recalculated on future runs.

After running the Get Results section, all results will be saved in results.csv. Any plots or datasets generated will be saved in each pitcher's ./data/pitcherName/ directory.