Every number on the live board comes out of the same four steps, run fresh every possession. No step ever reads the posted line — the book's number is only compared at the end. New to a term? The glossary has plain definitions.
Before tip, kickoff or first pitch, the model builds a baseline for each team from its recent games: points scored and points allowed, weighted so last week counts more than last month. Home field is added per sport. That baseline — not the posted spread — is the starting expectation for the game. If you want the market's opinion blended in, there's a slider for it; it defaults to zero, because the whole point is a number that doesn't owe anything to the book.
Once the game starts, the scoreboard is fact, not estimate. The current margin and current total go straight into the fair line. Around them, the model tracks pace: possessions used in football, scoring rate per second in basketball, half-innings in baseball. A team scoring faster than its baseline pulls its rate up — but the pull is capped, so one hot quarter can't rewrite who a team is. In football, a quarter where nothing happens matters too: we call it a no-show quarter, and each one drags the fair total down by a fixed penalty the market tends to apply too slowly.
The remaining clock is turned into expected remaining points for each team, scaled by who has the ball and — in baseball — by base-out state and the walk-off rule. Fair spread is the current margin plus expected remaining margin. Fair total is the current total plus expected remaining points. Win probability comes from putting that expectation through a normal distribution with a per-sport sigma, so a three-point lead means something very different with 40 seconds left than with a quarter to go. Late in games, when there's little time left to price, the board stops pretending to know more than it does and verdicts shut off.
Only now does the book's number enter the picture. The gap between book and fair is turned into a cover probability, and that probability is set against the break-even the odds demand — 52.4% at standard -110. Beat break-even by at least four points and the board says Bet. Sit inside the vig and it says Wait, with the exact thing that needs to happen for the edge to clear. Otherwise it's a Pass — the book's number is at or better than fair.
Say Green Bay leads Detroit 24-20 with 6:42 left in the fourth and the ball. The baseline said these teams should combine for about 48 points and play near pick'em. Two possessions remain, maybe three. Green Bay's rates say their remaining edge is worth roughly another point and a half on average, so the fair spread lands at -5.1. The book is hanging -3.5. Fair is 1.6 points stronger than the book, the model's cover probability clears break-even plus the four-point threshold, and the board prints BET GB -3.5. If the book had hung -6, the same game state would be a pass. Same teams, same score — the call lives entirely in the gap.
The steps are the same everywhere; the inputs change. NFL and college football price by possession and quarter, NBA and college basketball by the clock and the bonus, and MLB by half-innings and base-out state. Each sport page goes deeper on what's priced in.