55%: Best close-game control (TOR). 51→55: TOR overall vs close. +4: MTL lead-padded inflation. tied: The state to trust Every shot this season carries the score at the moment it was taken, so we can ask a question the standings cannot: how does a team drive play when it is ahead, even, or chasing? Do it and the swings are enormous. Some teams own the puck with a lead and get buried when behind; others are the reverse. It is tempting to call that character, front-runners and gamers. Mostly it is not. Here is the trap. A team's score state is not handed out at random. Good teams spend most of their time leading, often against weaker opponents, so their leading minutes look dominant for the simple reason that they are the better team that night. When they trail, it is usually against someone good or on an off night, so those minutes look ugly. The huge ahead-to-behind swing is largely that selection, not a coachable trait. So we do not rank teams by it. We strip it out. The clean read: tied hockey When the score is even, neither team is sitting on anything and neither is throwing caution away. Tied 5-on-5 expected-goals share is the closest this data comes to a distortion-free measure of who is actually better at controlling play, and there is plenty of it (most of a season is spent within a goal). The bars below are tied-game control; the tick on each is the team's all-situations number for comparison. TOR controls close games best in the league at 55% tied xGF%. The more interesting names are the mismatches. TOR drives play in tied hockey (55%) better than its overall number (51%) and far better than its record, which is exactly the profile of a team that controls the run of close games and loses them anyway. That is not a coincidence: it is the same team our regression read flagged as deeply unlucky and due to bounce back. Controlling close play and not winning it is what bad luck looks like, and it tends to correct. At the other end, MTL 's overall share (55%) sits well above its tied number (52%): a chunk of its season-long dominance is piled up after it already has the lead, which counts for less than control of an even game. Same number on the standings page, different team underneath. The full picture All eight teams by state, leading, tied, and trailing. Read the tied column as the real measure; treat the leading and trailing columns as descriptive, with the selection caveat above firmly in mind. TOR (tied 55%, leading 61%, trailing 38%, overall 51%). NY (tied 55%, leading 35%, trailing 64%, overall 53%). MIN (tied 53%, leading 62%, trailing 43%, overall 52%). BOS (tied 52%, leading 79%, trailing 25%, overall 53%). MTL (tied 52%, leading 78%, trailing 40%, overall 55%). SEA (tied 47%, leading 21%, trailing 60%, overall 44%). VAN (tied 45%, leading 31%, trailing 67%, overall 48%). OTT (tied 41%, leading 36%, trailing 56%, overall 44%) What this read isn't The lead/trail swing is not personality. It is mostly that a team's score state is earned: you lead because you are good, often against weaker opponents. We do not rank front-running or comeback grit from it, and neither should anyone else.. Tied is the clean cut, not a perfect one. Even tied minutes carry opponent quality and home/road that we do not adjust for here. It is the least distorted view available from score state, not a full strength-of-schedule model.. Control is not results. Driving expected goals in close games is what tends to translate, but it is not goals or wins. The gap between control and results is the luck (and goaltending) our regression and goalie posts are about.. 5-on-5, our xG. Special teams are excluded, garbage time (third period, three-plus-goal margin) is dropped, and the expected-goal values are our model's. Methodology Score state. Each 5-on-5 shot is tagged leading, tied, or trailing from the shooting team's view, using the score at the time. Team xGF% per state is expected goals for over for-plus-against in that state.. Garbage time. Third-period shots at a three-or-more-goal margin are dropped, where score effects and effort are least representative.. Why tied. Leading and trailing minutes are selected (a team is in them for a reason), so their xGF% blends play-driving with why the team got there. Tied minutes are the least selected, so tied xGF% is the cleanest single-number team read, in the spirit of score-adjusted shot metrics from the public-analytics literature.. Numbers regenerate on every build. A play-by-play re-sync or an xG recalibration reshapes the splits. This pairs with our regression read (who was lucky) and goaltending audit (who actually stopped the puck). Control, luck, and goaltending are the three pieces of a result; this is the control piece.