Leverage share
What if he scores…
+0.0
Your win probability
—
Coming soon. There is no conversational endpoint here yet, and the brief above needs no model to be complete.
Bigger cell, bigger swing. A player still has room on this board while he can move your week — plenty of game left, in a matchup that is still close. Blowouts and finished games shrink away on their own.
Each cell is scored on how much your win probability would move if that player's afternoon went differently — his remaining uncertainty, weighted by how close the matchup still is.
leverage = φ(m ÷ s) × σᵢ ÷ s
where m is your expected final margin, s is how uncertain that margin is, and σᵢ is how much is still unknown about this player. The same model sizes a whole league on the front board.
Two things it gets wrong. It treats players as independent, so a quarterback and his own top receiver are counted twice over — a stack looks more important than it is. And it uses an average spread for each position rather than each player's own, so a steady runner and a boom-or-bust one are treated alike.
Read cell size as where to look, never as a prediction.