Major League Baseball
MLB predictions & analytics
Baseball is the most random of the major sports on a single-game basis: even strong teams lose around 40% of the time. Projections here are naturally closer to 50/50, and the useful signal comes from starting pitching, run differential and the park.
Upcoming MLB games
What the MLB model looks at
Run differential
Runs scored minus runs allowed is a steadier guide to team quality than win-loss record.
Starting pitchers
When probable starters are known, their recent effectiveness shifts the projection.
Ballpark
Some parks inflate or suppress scoring, which affects projected totals.
Bullpen workload
Heavy recent relief usage can weaken late-inning performance.
Known limitations
- Pitching changes are common and can make a projection outdated.
- A 162-game season means individual game probabilities rarely exceed 65%.
Measured results for every model version are published in the Model Lab. Projections are estimates, never guarantees.
Related reading
- MLB Analytics and the Challenge of Predicting BaseballBaseball has the richest statistics in sports and some of the least predictable single games. Why even great teams lose four in ten, and how that shapes MLB projections.
- How Sports Prediction Models WorkA plain-English tour of a sports prediction model: the data it learns from, the features it builds, how it turns them into a probability, and how it is tested before anyone relies on it.
- What Does a 70% Win Probability Actually Mean?A 70% win probability is not a prediction that a team will win. It is a statement about frequency. Here is how to read it, how to check it, and the mistakes people make with it.
- Understanding Brier Score in Sports PredictionsThe Brier score measures how good probability forecasts are, not just whether picks were right. Worked examples, benchmarks and the common traps when comparing models.