How it works

Sharp Juice turns four well-known public college football power ratings into one consensus point spread for every FBS game, then compares it with the betting market and Kalshi's prediction market.

The model line

For each rating system, the projected margin is the home team's rating minus the away team's rating, plus a home-field value for that specific home team (from Massey Ratings; zero at neutral sites). The model line is the negative of that margin, so a negative number means the home team is favored, the same convention sportsbooks use. The published line is the simple average of the four systems.

The edge

Edge = model line − market line. A positive edge points to the away team, a negative edge to the home team. Market lines are the median spread across U.S. sportsbooks. The model is set on Sunday, the opening snapshot is taken at the same time, and the market is refreshed every morning and hourly on Saturdays. Results are graded against the Sunday line.

Win probability

Projected margins are converted to win probabilities with the Winston formula: P(favorite) = 0.5 + 0.5 × (1 − e^{−0.0454·|S|1.12}). Kalshi prices are the midpoint of the bid and ask, normalized across both teams.

Closing line value

Closing line value measures whether the market moved toward the model's side between Sunday and kickoff. Over a few hundred games it is usually a steadier signal of real skill than wins and losses.

What this isn't

This is a research project, not betting advice. Six weeks of college football is a small sample, and even a good model loses often against the spread.