The problem xG solves
For decades, shots on goal was the closest thing hockey had to a possession stat. The trouble is that a shot is a terrible unit of measurement. A harmless wrister from the blue line and a point-blank chance from the slot both count as one shot, yet one scores maybe three percent of the time and the other closer to twenty. Expected goals exists to fix that.
Expected goals, or xG, assigns every unblocked shot a scoring probability based on the things that actually matter:
- Distance from the net
- Angle to the net
- Whether it followed a pass across the slot, which forces the goalie to move
- Whether it came off the rush or a rebound
- Shot type
Add up those probabilities across a game and you get a team's expected goals total. It is, in plain terms, how many goals an average finisher would have scored from those exact chances.
Reading a game through xG
Say a team wins a game 2-1 but was out-chanced badly, posting 1.4 xG to the opponent's 3.2. The scoreboard says they won. The xG says they got away with one, thanks to a hot goalie or a lucky bounce. Over a single night, luck rules. Over twenty games, xG tends to be right and the scoreboard tends to catch up.
Expected goals will not tell you who won last night. It will tell you who is likely to win next month.
That is the core value. xG is a stickier, more predictive signal than goals themselves, because it strips out the enormous randomness of finishing and goaltending in small samples.
Where people go wrong
xG is powerful and abused in equal measure. A few honest cautions:
- It is a model, not a truth. Different public models weight chances differently and will disagree at the margins.
- Most public xG does not know about pre-shot passing or screens perfectly, so it undervalues elite playmaking setups.
- A single game's xG is noisy. Judging a player or team on one night of xG is exactly the mistake xG was invented to prevent.
How to actually use it
Use xG as a lie detector for the scoreboard. When results and underlying chances diverge over a stretch, bet on the chances. Use it to separate a team that is genuinely good from one that is riding save percentage. And use it at the player level with humility, paired with your eyes, because the number describes the quality of chances a player is involved in, not the intangibles around them.
Once xG clicks, you stop asking who scored and start asking who earned it. That is a better question, and it is usually the one that predicts tomorrow.