How can coaches and players tell the outcome of a pitch in a baseball game based on observed data?

Abstract

Machine learning and the development of artificial intelligence continues to grow at a rapid pace with the age of technology and computer science jobs growing in demand. With this development the use of predictive technology becomes more realistic to make systems more efficient. For example, in the future of the medical field, a software program may diagnose patients given an input of symptoms and previous health issues more effectively than even humans can.
 Sports are an area with vast possibilities of outcomes, meaning a lot of predictions to be made in all sorts of sports. Humans don’t have the analytical ability to examine every aspect of each player to predict the next situation in a sports game. For example in baseball and softball, there are thousands of possible scenarios that can occur from a singular pitch. 
 How can coaches and players tell the outcome of a pitch based on observed data?
 A simple machine learning predictive model using a technique called gradient descent takes features and examples as input in order to predict one of a couple possible outcomes for a specific pitch with its various features. Results show that upon running such programs numerous times with different iterations the model actually grows stronger in accuracy.