Extra Base Analytics

Quantitative research applied to baseball betting markets.

Opening Day 2026
--
Days
:
--
Hours
:
--
Min

We build quantitative models for baseball pitcher props, batter props, and game-level markets. Every model is walk-forward backtested on real closing odds across multiple seasons — no hypotheticals, no overfitting to history.

We scanned 16 markets and tested dozens of model configurations. We killed more models than we shipped. What survived is a portfolio of models launching on Opening Day.

3
Seasons of real odds
16
Markets scanned
5M+
Odds observations
01
Walk-forward validation. Every model is tested on data it has never seen, in chronological order. No peeking, no leakage, no curve-fitting.
02
Real closing odds. Backtests use actual market prices from multiple sportsbooks. If a line didn't exist, the bet didn't happen.
03
Parsimony wins. Three to four features consistently outperformed larger models. We build the simplest model that captures the edge.
04
Kill what doesn't work. Dozens of features and model configurations tested and rejected. The feature cemetery is larger than the production codebase.

We're sharing our research process and findings in the weeks leading up to Opening Day. Follow along on X for market structure analysis, killed model write-ups, and what we've learned about where the books leave edges on the table.