Loading...
Every shot has a probability of going in. xG measures that probability. It's one of the best tools we have for separating luck from skill in hockey.
TL;DR
xG assigns a scoring probability (0 to 1) to every shot based on location, type, traffic, and game state. A team's total xG tells you how many goals they should have scored based on shot quality — regardless of whether the puck actually went in.
Closer to the net = higher xG. A shot from the crease scores way more often than one from the point.
e.g. Slot one-timer
One-timers and deflections are harder for goalies to track. They carry higher xG than standard wrist shots.
e.g. Deflection
Shots with bodies in front of the goalie are harder to stop. Models weight screened shots higher.
e.g. Screened wrister
Power plays produce better looks. xG models separate 5v5, PP, PK, and empty net situations.
e.g. 5v4 power play
The gap between expected goals and actual goals tells you whether a team is running hot, getting unlucky, or performing at their true level.
Generating quality chances but the puck isn't going in. Usually bad luck — shooting percentage or goalie variance is suppressing results. Expect positive regression.
Buy low candidate
Winning but outperforming shot quality. Shooting percentage is unsustainably high. Expect negative regression — this record is fragile.
Sell high candidate
Results match underlying quality. This team is performing at their true level — what you see is what you get.
Trustworthy record
Shot quality is baked directly into the prediction model. Rather than just counting goals, PuckCast evaluates whether a team's recent results reflect their underlying shot quality — or whether variance is distorting the picture.
Teams with a big gap between xG and actual goals get flagged as regression candidates. The model adjusts their win probability to account for the expected correction, which is one of the reasons it outperforms raw standings-based predictions.
xG in the model
Western Conference
Elite cycle game and cross-ice passing generate consistently high-danger chances. Among the best xGF at 5v5 in the West — the model rates them as one of the most complete teams in the league.
Eastern Conference
Structured defense suppresses opponent xGA while relentless forechecking creates quality looks. Regularly lead the league in xGF% — a model favorite most nights.
xG is a statistical model that assigns a probability to every shot based on its quality: shot location, type, traffic, and game situation. A slot one-timer from the high-danger area might carry an xG of 0.30 (30% chance of scoring), while a point shot through traffic might be 0.05. Adding up all those probabilities gives you a team's expected goal total.
Actual goals are the final score; xG is what the score should have been based on shot quality. A team can out-xG their opponent but still lose if goaltending or shooting variance tips things the other way. Over a full season, xG is a more reliable predictor of future performance than raw goal totals.
PuckCast's model currently rates the Dallas Stars and Carolina Hurricanes as two of the strongest xG teams in the league. Both generate high shot quality consistently at 5v5 and suppress opponent shot quality through structured defensive play.