Buffalo Sabres

xG-Elo moves game by game with the model’s expected-goals margin, never with the score or the standings, which is why it can disagree with a club’s points. That disagreement is the sentence above. Under 7 spots apart the two are read as in line: a display rule for the wording, not a test of whether a gap is real.

This club’s rating is 1507. Every club starts from 1500 and is pulled back toward it between seasons, so above 1500 is stronger than an average team and below it weaker.

Every shift each player took in the game is lined up against everyone else’s, using the NHL’s own shift charts. Only five-on-five time with both goalies in net counts, so power plays and penalty kills do not split a line. Line 1 is the three forwards who spent the most of that time on the ice together, line 2 the most of the rest, and so on; defence pairs the same way.

A group that never took a full shift together, about a minute, is not called a line: its players are listed as also played. A game is used only when every skater’s time in the shift chart is within 30 seconds of the box score, so a chart the NHL published incomplete is skipped rather than read.

A power-play unit is the same five skaters on the ice together while the club had five skaters to the other side’s four or three; a penalty-kill unit is four skaters together while short a man. The second unit is the best group that shares at most one player with the first, since one player often plays on both. Power-play and short-handed time follow the penalties, not an empty net: a goalie pulled for an extra attacker at even strength is still even strength.

The box score is the NHL’s own, except four figures it does not publish. Short-handed goals and faceoffs won and lost come from the play-by-play, which is used only when it reproduces the box score’s goals, power-play goals and faceoff percentages exactly; power-play and short-handed ice time come from the shift chart.

Each player is listed under the club the NHL had them on when we last checked, and clubs are re-checked once a week: a player traded in the last few days can still appear under their previous club, while the rates beside them span every season since 2021-22, wherever they played them. Each rate is a recency-weighted per-game average, and Sample is the number of games it rests on.

Skaters are ordered by a points rate shrunk toward the league average in proportion to how little a player has played, so a one-game sample does not out-rank a career. That rate is Adj. P; PTS and the other columns are the recorded rates, and for anyone with a real sample Adj. P and PTS are nearly the same number. Save percentage is total saves over total shots, weighted, not the average of per-game save percentages, which would count a two-shot relief appearance as heavily as a forty-shot start.