For most of hockey’s history, a team’s quality got judged almost entirely by the scoreboard — wins, losses, and maybe goal differential if you wanted to get fancy. That approach has always had an obvious flaw: a single game’s result can hinge on a lucky bounce, a hot goaltender, or a couple of officiating calls that could have gone either way. Advanced analytics exist precisely to cut through that noise and answer a more useful question — not just who won tonight, but which team is actually playing better hockey over a meaningful stretch of games.
Corsi and Fenwick: counting the shots that matter
The starting point for most modern hockey analytics is shot attempt data, commonly tracked through two related metrics: Corsi and Fenwick. Corsi counts every shot attempt a team generates or allows — shots on goal, missed shots, and blocked shots all included — while Fenwick strips out blocked shots, working on the theory that a shot good enough to get blocked still says something different about play quality than one that sails wide or gets stopped clean. Both metrics rest on a simple premise: a team that consistently generates more shot attempts than its opponents is controlling the run of play more often than not, even on nights when the final score doesn’t reflect it.
The reason these numbers caught on is that they’re far more stable over time than goals alone. Goal totals bounce around heavily from game to game because scoring is relatively rare and heavily influenced by shooting luck and goaltending variance. A team’s shot-attempt share, by contrast, tends to be a much steadier signal — which makes it a better predictor of how a team is likely to perform going forward than its record over the last ten games.
Expected goals adds a layer of context
Where Corsi and Fenwick treat every shot attempt roughly equally, expected goals (xG) models try to account for shot quality rather than just shot volume. A point-blank one-timer from the slot and a low-percentage attempt from the blue line both count as one shot attempt in a Corsi tally, but they’re obviously not equally dangerous. xG models assign each shot a probability of becoming a goal based on factors like location, shot type, and the situation it was taken in, then sum those probabilities across a game or season to produce a more nuanced picture of scoring chances created and allowed.
This distinction matters because a team can post strong shot-attempt numbers while generating mostly low-danger looks from the perimeter, or conversely post modest shot totals while consistently creating high-quality chances close to the net. Neither Corsi alone nor raw shot totals would catch that difference — xG is built specifically to surface it.
Why these numbers are more accessible than ever
A decade ago, digging into this kind of data required either scraping play-by-play logs yourself or relying on a small handful of independent analysts publishing spreadsheets on personal blogs. That’s changed considerably. NHL teams now employ full analytics departments, mainstream broadcasts routinely cite shot-share and xG figures during game coverage, and public-facing dashboards make it possible for any fan to pull up a team’s underlying numbers in seconds rather than needing a background in data science to interpret raw shot logs.
That accessibility has changed how fans argue about hockey, too. A debate about whether a team’s playoff push is sustainable used to rest almost entirely on record and gut feeling; now it’s just as likely to involve a discussion of whether that record is backed up by strong underlying shot-share and expected-goals numbers, or propped up by an unsustainable hot streak from a goaltender or a shooting percentage that’s unlikely to hold.
The limits worth keeping in mind
None of this means the old-fashioned eye test is obsolete. Advanced stats are excellent at identifying broad patterns over a large sample of games, but they’re less useful for capturing things like a defenseman’s positioning on a specific shift, the value of a strong penalty-killing system, or a goaltender’s ability to read a specific opponent’s tendencies. The best hockey analysis today tends to blend both approaches — using shot-share and expected-goals data to establish the broad trend, then watching the actual games to understand why that trend is happening and whether it’s likely to continue.
For a sport that used to be judged almost entirely by the final score, that shift toward richer, more predictive numbers has made following hockey closely a genuinely different experience than it was even ten years ago — one where the scoreboard tells you what happened, but the underlying numbers are far more likely to tell you what’s actually going on.