The model's price for every game, next to the book's.
Fairline runs four models on free public data. None of them reads a book line. A card wears a chip when a model and a book disagree by enough, and that chip is not a pick.
Each league gets its own model.
It reads starting pitchers, park and weather, lineup quality and bullpen fatigue.
It publishes a win probability, fair moneyline odds and a projected run total.
It reads team scoring and save rates, goalie starts, special teams, home ice and rest.
It publishes a win probability, fair moneyline odds and a projected goal total.
It reads club attack and defence rates split by venue, recent form and the tau draw correction.
It publishes home, draw and away probabilities, fair odds and a projected goal total.
It reads play-by-play EPA, opponent strength and rest between games.
It publishes a win probability, fair odds, a model line and a projected point total.
How to read a card.
- 1The bar. It splits the model's win probability between the two teams.
- 2Fair odds. This is the price at which the model breaks even on each side.
- 3The book price under it. This is the best number the books we track are quoting right now.
- 4The chip. It names the largest disagreement on the game, with the side, the line, the book price and the fair price.
Every graded bet stays on the record.
Transparency →A sport publishes a skill number once it passes 500 graded games. Under that it prints its count and reads gathering.
The 2025 MLB backtest replayed the pipeline over 2,421 games with no lookahead. It reported 55.8% moneyline accuracy and +0.026 Brier skill against a coin flip. Fairline lists that study as pending data.
The limits.
The trial takes a card up front and you can cancel inside it. The pricing page lists what a plan opens.