Following framework from 538’s MLB ELO model, I developed my own
code base and model to analyze team strengths and generate pre-game predictions. Trained on data going back to 2016, my model looks at past game scores,
adjusting for home field advantage, year to year regression, and the starting pitcher of each game. At the end of the 2024 regular season,
it had a 57.4% moneyline accuracy and also had a strong log-loss performance. The code for the model and a more detailed explainer can be found here:
MLB ELO Model Code.