“Where does our estimated distribution disagree with the line?”
The first release will expose documented research outputs. Model scores will always ship with context, sample size, and an honest uncertainty range.
03 / Price the matchup
Find the historical games that actually rhyme.
Represent each matchup by the conditions known at decision time, find comparable historical games, estimate a fair outcome range, and compare it with a vig-adjusted market price.
- Pregame team, player, opponent, and situation features
- Historical closing and time-stamped market lines
- Injuries, weather, rest, venue, and roster context
- Comparable-game evidence
- Fair spread, total, and probability ranges
- Tracked edge, closing-line value, and calibration
Build sequence
From foundation to live product.
- 01
Acquire licensed historical odds snapshots
- 02
Backtest similarity and outcome models chronologically
- 03
Publish only edges that clear uncertainty thresholds