Agents tracking a live game keep needing the same number: given the score right now and the time left, who's going to win, and by how much? win-prob answers that in one call.

Give it the current scores, how much of the game is left (a fraction, or minutes-remaining plus total minutes), and the sport. It returns each team's win probability, the projected final margin, and the size of the swing still possible — with human-readable drivers explaining the call.

It's a probabilistic model built on three inputs: the current margin, the time left, and how much scoring each sport typically packs into that remaining time. From those it projects a likely final margin and the uncertainty around it, then turns that into each team's win probability. Less time left and a wider lead both push the probability toward a settled result.

Scoring volatility is built in for NFL, NCAAF, NBA, NCAAB, WNBA, MLB, NHL, and soccer; pass scoring_std to calibrate any other sport. An optional pregame_spread folds in who was favored at kickoff, so an early lead by an underdog is weighted correctly.

Every call settles in USDC over x402. It's a statistical estimate, not betting advice and not affiliated with any sportsbook.

win-prob opens the statline cluster — sports and statistics endpoints an agent can call as it needs them. More to follow.