Pitching Predictive Model

What does this model predict?

• This model predicts how many strikeouts a given starting pitcher will have against a specific on a specific day.

What is this model based on?

• This model learns patterns using historical box scores and contextual game features available before each game.
• This includes average recent strikeouts,average season strikeouts, days of rest, and opponent strikeout tendencies (including throwing hand).
• It also uses a mix of player performance, opponent behavior, and game context to predict the number of strikeouts a pitcher will get in a given game.

What is the model type?

• This model is a Random Forest Regressor -- A tree-based ensemble method that learns patterns from historical data to predict a continuous value (strikeouts).

What data is used to train the model?

• Game logs are grabbed for almost all starting pitchers in the MLB via the ESPN API to provide accurate information for this model to learn from.

How accurate is this model?

• This model was only recently created so it's effectiveness is still being studied. More information will be provided later to provide it's accuracy.
• As time goes on, this model should only become more accurate with more data (games played).

What flaws could affect this model?

• In the MLB, there are sometimes starters that are injured or sent to the minor leagues. THis means some games are entirely used and can affect the model.
• In addition, this model doesn't fully account for injuries on a team right away. For example, if a batter on a team suffers a major injury, it could take a few games for the model to see a trend on the opposing team if there is one.