How hockey analytics works

Hockey has one of the lowest scoring rates of any major team sport, which means goals, the outcome everyone cares about, are too rare to reliably separate a good team from a lucky one over a short stretch of games. Hockey analytics grew out of a practical workaround: since shots are far more frequent than goals, and shot generation correlates with underlying team quality, analysts built metrics around shot attempts as a higher-volume, faster-stabilizing proxy for possession and territorial control.

The questions analytics is built to answer

Analysts and front offices use hockey analytics to answer questions the goal-and-assist box score is too sparse to settle on its own: is a team that's winning actually outplaying its opponents, or riding hot goaltending and finishing that is unlikely to continue? Which players drive play in their team's favor even on nights they don't register a point? And how much of a team's record is deserved, given how a small number of low-probability events — a bounce off a skate, a lucky deflection — can decide a one-goal game?

The data it runs on

Hockey analytics runs on two main layers of data.

• Shot and event data records every shot attempt, whether it hits the net, misses, or is blocked, along with faceoffs, penalties and other discrete events, tagged by team and game situation (even strength, power play, shorthanded). This data has been tracked and available for professional hockey for many years and underlies most of the sport's public advanced statistics.
• Optical and positional tracking data, increasingly deployed in professional leagues via puck- and player-tracking sensor systems, adds location, speed and zone-entry detail that shot data alone can't capture — where on the ice possession changes hands, how a team enters the offensive zone, and how quickly a player closes down space. This tracking layer is newer and less universally available across leagues and levels than shot data.

Measuring possession without the puck

[[corsi|Corsi]] counts all shot attempts — goals, shots on net, missed shots and blocked shots — for a team while a specific player is on the ice, as a proxy for which team controlled the puck and generated offensive pressure. The logic is straightforward: a team that consistently generates more shot attempts than it allows is usually controlling the run of play, even on nights the score doesn't reflect it yet, because shot generation is a much larger, faster-stabilizing sample than goals.

[[fenwick|Fenwick]] is a close variant of Corsi that excludes blocked shots, on the reasoning that a blocked shot reflects the defending team's positioning as much as the attacking team's pressure, and so is a slightly noisier signal of true offensive intent. In practice, Corsi and Fenwick tend to tell a similar story for most teams and players; analysts often report both and treat large disagreements between them as worth a closer look.

Measuring luck versus sustainable performance

[[pdo-hockey|PDO]] combines a team's shooting percentage and save percentage at even strength into a single number, typically centered around 100. It is used specifically as a regression indicator rather than a skill measure: a team with a PDO well above 100 is generally getting finishing and goaltending luck that is unlikely to continue at that rate, while a team well below 100 is often better than its record suggests and due for its results to improve as shooting and save percentages normalize. PDO is one of the clearer examples in any sport of a metric built explicitly to flag unsustainable performance rather than to praise or blame a team for it.

[[expected-goals|Expected goals]] in hockey, adapted from the same underlying idea as in soccer, estimates the probability each shot attempt should have resulted in a goal based on shot location, type and the situation it was taken in, giving a quality-adjusted alternative to simply counting shot attempts. A team generating a high volume of low-danger, perimeter shots looks different under expected goals than under raw Corsi, even though both would show similar shot-attempt totals.

How the work is done in practice

Analytics staffs use possession and expected-goals metrics together to evaluate line combinations and matchups — which forward line reliably outshoots its opposition, and against which opposing lines — and to build the case for or against a lineup change independent of how the scoreboard happened to look on a given night. Front offices use multi-season Corsi, Fenwick and expected-goals trends, rather than single-season point totals, when evaluating a player for a long-term contract, because point totals in a low-scoring sport are more subject to variance in teammates' finishing and a team's overall goaltending than the underlying possession numbers are.

Common mistakes and misreadings

• Reading a hot or cold PDO as a skill change rather than what it usually is: shooting or goaltending luck that tends to move back toward league-average over time.
• Using a single game or short stretch of games to judge possession numbers. Corsi and Fenwick need a reasonably large sample of shot attempts to stabilize; a handful of games can be dominated by score effects (teams that are already ahead tend to play more defensively, which suppresses their own shot attempts).
• Ignoring score effects generally. A team's shot-attempt rate shifts systematically depending on whether it is leading, tied or trailing, which can distort raw possession numbers if not adjusted for game state.
• Treating raw shot-attempt volume as equivalent to shot quality. A team generating many low-danger perimeter attempts can look strong on Corsi while creating little real scoring threat — expected goals exists specifically to catch this gap.
• Extrapolating a strong or weak start to a full season before enough games have accumulated for possession and finishing numbers to settle down; regression to the mean is a bigger factor early in any season than most box scores suggest.

For how teams combine possession data with tracking, video and injury-risk information across performance and roster decisions, see how sports teams use analytics and how front offices use analytics. Browse every tool in this category.