Leverage & close-game scoring

What we measure

Leverage asks a simple question: when a player's offense arrived, was the game still in the balance? A point that ties a game late is worth remembering. A point scored up four goals in garbage time is not. The box score treats them the same. We tag each goal and each point with the leverage of the moment it came in. Then we report the share of a player's production that landed in the tightest moments. We call it his close-game share. We show it beside the league's share for that season.

How we approximate leverage

True leverage would weight each point by the swing in win probability at the instant it was scored. The public feed has no live win-probability model. So we estimate it from something the feed does give us cleanly: the score margin when each goal went in. A goal in a tied or one-goal game counts as high-leverage. So does any goal in overtime. A goal into an empty net is the exception. Its points count toward his total but in no tier. A two-goal game is medium, and a margin of three or more is low. This is judged goal by goal, not from the final score. Say a team trailing by one ties the game, then goes on to win 6-1. That tying goal still counts as high-leverage. Shootout deciders are set aside, since they aren't scored in the run of play. It's a stand-in, and we label it as one. But it is grounded in the state of the game as each goal happened, not in how the night ended.

Graded against the league, not a round number

The temptation with a close-game metric is to pick a round cutoff and let it stand. Say, 55% of a player's offense in tight games. That would mislead here. Most goals in hockey are scored in tied or one-goal games to begin with. A clear majority of all league points land in high-leverage moments. That's a property of the sport, not a mark of any one player. A fixed cutoff would flag almost everyone.

So we compare every player with that season's league share instead. Above league means his close-game share runs at least six points above the league's share that season. Below league means at least six points under it, with more of his production in lopsided games. Near league is everything in between. Six points either way splits off roughly the top and bottom quarters of qualified players. The comparison is made within each season. So above league means his share beat his peers' in the same scoring climate. It does not mean he cleared a fixed number. A fixed number would drift with each era's style of play. It describes this season, not a trait.

Coverage and tier distribution

We publish it for 2023-24 through 2025-26. Over those seasons, 1,721 qualified player-seasons from 752 players carry a reading. A player needs at least 10 total points in a season to get one. So a handful of well-timed goals can't produce a reading. The counts below are what those boards contain right now. Under the season-by-season rule, the three readings split like this:

We checked whether this share repeats for the same player. It does not. So it describes a season and says nothing about the next one.

We also show game-winning goals, overtime goals and games played for context. Those come straight from the box score. They do not go through the leverage stand-in. They round out the picture and do not feed the reading.