What is a prop betting model? It's a system that projects player performance using historical data, then compares those projections to sportsbook lines to find mispriced odds — situations where the book is offering worse odds than the true probability.
How we build projections: We use XGBoost machine learning trained on 13.6M+ closing-line data points, combined with real-time Baseball Savant metrics (xBA, xSLG, xwOBA), pitcher splits, umpire tendencies, weather, ballpark factors, and recency-weighted player logs.
What edge means: Edge is the difference between what our model thinks the true probability is and what the sportsbook's odds imply. A +7% edge means our model thinks you have a 7 percentage-point advantage over the market price.
Closing Line Value (CLV): The gold standard for measuring a model. If you bet at +150 and the line closes at +130, you got better odds than the final market price. Consistently beating the closing line is mathematical proof of an edge.
Confidence %: The model's estimated probability that the pick wins. It factors in calibration — our model is deliberately conservative, capping at 58-65% because real win rates in sports rarely exceed that range over large samples.