Hockey InsightsExpected Goals

Expected goals for every NHL shot

Deserved standings, top 10, 2026-27 through November 16, 2026
RankClubGPPointsDeservedGap
1Carolina Hurricanes213433.2+0.8
2Utah Mammoth222330.6-7.6
3New Jersey Devils213029.4+0.6
4Vegas Golden Knights222928.6+0.4
5Colorado Avalanche212727.6-0.6
6Florida Panthers222726.8+0.2
7New York Rangers222725.9+1.1
8Tampa Bay Lightning202825.9+2.1
9Minnesota Wild223425.6+8.4
10Dallas Stars202624.9+1.1
Where the goals come fromxG Goals · G / xG
Tap for the replay
Shot or miss Goal Size = chance
This season's findings
Standings by expected goals

The deserved standings

Points banked against points earned by the chances. Tap a club for its xG page.

Win, more xG Win, less xG Loss, less xG Loss, more xG Bar height = how much of the game they deserved
Shot locations

Where every club shoots and allows shots

xG a game by patch of ice, against the league.

Where they create

More than leagueLess

Where they allow

More than leagueLess
Finishers

Goals below and above fxG

Goals against fxG, our xG adjusted for each shooter's past finishing.

Furthest below fxG Our fxG
Furthest above fxG Our fxG
GoalsfxG: xG adjusted for his own finishingTap a player for his game log.
The chance makers · most expected goals so far
The crease

Goals the goalie took away

Expected goals against minus goals allowed.

Goals in motion

Watch how the chances became goals

NHL tracking of every skater and the puck, with our xG for the shot.

Replays

Did the winner deserve it?

Running xG for both clubs, goals marked.

Biggest wins with less xG
Every number
How it works

Expected goals (xG) is the chance an unblocked shot becomes a goal. Ours is film-xg-v2.2, trained on the unblocked shots of the 2024-25 and 2025-26 regular seasons. It reads the shot's distance, angle and type, the skaters on each side, the period, the time left and home or away. It also reads what came just before: the previous event, which club made it and how long ago. It reads the time since the last attempt, a rebound within 3 seconds and how far the puck moved side to side. It does not read the score. The recipe was frozen on October 6, 2026. Each night a calibration shift is refitted on earlier 2026-27 games only, so no shot is scored with its own result. Empty-net shots are left out.

So far in 2026-27, through November 16: 27,540 unblocked shots, 1,973 expected goals and 1,961 goals. Log loss 0.221 against 0.257 for a model with no information; AUC 0.776, where 0.5 is a coin flip and 1.0 is perfect. The nightly calibration keeps total expected goals near total goals, so log loss and AUC are the measures of skill.

fxG multiplies a shooter's xG by his finishing factor, built from earlier seasons only. Finishing is about half repeatable from year to year, so the factor is pulled toward average.

Deserved standings. Each game is replayed thousands of times from its chances; a club's deserved share is how often it wins those replays. Deserved points are twice that share per game, scaled so the league total matches the real total (overtime losses included).

Goals taken away uses shots on goal only. Shot locations count every unblocked shot per game played, on a 6-foot grid, lightly smoothed.

NHL EDGE numbers (shot speed, the NHL's save percentages, goal tracking) come from the league's own tracking. They sit beside our numbers and are never mixed into them.

Game logs list the games with at least one unblocked attempt (a goalie's, every game he faced a shot). Each row opens that game's page. A club's xG page is linked as /xg/?club=TOR and a player's as /xg/?p= and his NHL id.

Early in a season the samples are small. Big gaps can come from a few games.