In this project focused on predicting pitcher's performance, I developed 27 predictive models utilizing linear and polynomial regression techniques. By incorporating statistical testing, data cleaning, and scaling techniques, I ensured the accuracy and reliability of the models. Demonstrating a 95% confidence level, I established that the best model outperformed the mean line, making it a robust tool for predicting player performance.
2022 MLB season player data is used to evaluate their goodness given specific stats and conditions. This prediction model aids in comparing and selecting the most suitable pitchers for teams, with the dataset comprising 2022 MLB player stats encompassing 1068 rows and 35 columns.
The project focused on relevant features, eliminating irrelevant columns and redundant stats, resulting in a reduced feature set of 15 stats and conditions crucial to a pitcher's performance. By examining a pitcher's historical stats and the conditions they are expected to play in, the research question centered around effectively judging their future performance.
Key factors such as age, games played, games finished, intentional bases on balls, batters faced, shutouts, saves, hits allowed, and runs allowed emerged as crucial indicators in predicting pitcher performance. Utilizing a second-order statistical model with these features, teams can make informed decisions when acquiring pitchers. Additionally, the model's predictions lie within a certain range with 95% confidence, providing a valuable range for evaluation.
The best model was identified based on its minimal error. This project empowers baseball teams with a reliable tool for predicting pitcher performance, enhancing their decision-making process in selecting the most impactful players for the upcoming season.
Check the presentation for the complete walkthrough of the project.
Dataset: https://www.kaggle.com/datasets/vivovinco/2022-mlb-player-stats
Code: https://github.com/SwaroopMeher/Baseball-pitcher-performance-predictor/blob/main/Baseball%20pitcher%20perf%20predictor/mini-project-solution.R