NHL Advanced Statistics for Betting: Corsi, Fenwick, and xG

For decades, hockey bettors relied on the same handful of numbers: goals, assists, save percentage, goals against average. These stats told you who scored, who stopped the puck, and not much else. Then the analytics movement arrived, and suddenly a sport that had resisted quantification longer than any other major league found itself drowning in metrics with names that sounded like European furniture brands. Corsi. Fenwick. Expected goals. High-danger scoring chances. The language changed, and so did the edge available to anyone willing to learn it.

The adoption curve in hockey analytics is still steeper than in baseball or basketball, which means the gap between casual bettors and analytically informed bettors remains meaningful. Sportsbooks have integrated advanced stats into their models, but public betting money still moves primarily on goals, points, and name recognition. That gap between how the public prices teams and how the data prices them is where advanced stats create genuine value for hockey bettors who do the work.

See also special teams analysis.

Corsi: The Shot Attempt Revolution

Corsi is the foundational metric of modern hockey analytics, and understanding it unlocks nearly every other advanced stat in the sport. At its simplest, Corsi measures all shot attempts — goals, shots on goal, missed shots, and blocked shots — directed at the net during five-on-five play. A team’s Corsi For percentage (CF%) represents their share of total shot attempts when they’re on the ice at even strength.

The logic behind Corsi is straightforward: teams that generate more shot attempts than they allow are, over time, more likely to outscore their opponents. Scoring a goal in hockey requires getting the puck to the net, and Corsi captures the volume of those attempts regardless of whether they’re saved, blocked, or miss the net entirely. A team with a 54% CF% is directing 54% of all shot attempts toward the opposing goal, which means they’re controlling the flow of play more often than not.

For betting purposes, Corsi becomes valuable when it diverges from a team’s actual results. A team posting a 54% CF% but only winning 45% of its games is likely experiencing bad luck — poor shooting percentage, weak goaltending, or both. Historical data shows these teams tend to regress toward their underlying shot metrics, meaning their win rate should improve. Conversely, a team winning 58% of its games with a 47% CF% is overperforming its process and is a candidate for regression downward. Betting on the direction of regression — backing underperforming Corsi teams and fading overperforming ones — is one of the most robust analytical strategies in hockey betting.

One important caveat: Corsi doesn’t account for shot quality. A team might generate high Corsi numbers by firing long-range shots from the perimeter, which are easy to produce but rarely result in goals. This limitation is real, and it’s why Corsi alone isn’t sufficient. But as a measure of territorial dominance and puck possession, it remains the single most accessible entry point into analytical hockey betting.

Fenwick: Corsi’s More Refined Cousin

Fenwick is identical to Corsi except it excludes blocked shots. The reasoning is that blocked shots are partially a function of the defensive team’s skill — good shot-blocking teams can suppress Corsi numbers without actually being outplayed — so removing them produces a cleaner measure of unimpeded offensive generation.

In practice, the difference between Corsi and Fenwick is usually small. Teams with high Corsi numbers tend to have high Fenwick numbers, and the rankings rarely diverge by more than a few positions. Where Fenwick becomes specifically useful is in evaluating teams that face opponents with extreme shot-blocking tendencies. If a team’s Corsi looks mediocre but their Fenwick is strong, it suggests they’re generating plenty of offensive chances that happen to be getting blocked at an unusual rate — a situation that may correct itself as they face different opponents.

For betting, Fenwick’s primary value is as a confirmatory metric. If both Corsi and Fenwick agree that a team is controlling play, the signal is stronger than if they diverge. When they do diverge, investigating why — shot-blocking tendencies, coaching system, specific opponents — can reveal whether the underlying performance is sustainable or likely to shift. Most serious hockey bettors track both metrics but weight Corsi more heavily in their models simply because the larger sample of shot attempts (including blocks) provides more statistical stability.

Expected Goals: The Gold Standard

If Corsi tells you who’s controlling play and Fenwick refines that picture, expected goals (xG) answers the question everyone actually cares about: how many goals should a team be scoring and allowing based on the quality of chances they create and concede? Expected goals models assign a probability to every shot attempt based on its location, angle, shot type, whether it came off a rush or a rebound, and various other contextual factors. A shot from the slot on a two-on-one might carry an xG of 0.35, meaning it scores about 35% of the time historically. A point shot through traffic might carry an xG of 0.03.

By summing all these individual shot probabilities across a game or a season, you get a team’s expected goals for and against. A team generating 3.2 xG per game while allowing 2.5 xG is performing at a level that should, over time, produce a strong goal differential and a winning record. If that same team is actually scoring only 2.4 goals per game, their shooting percentage is below expected, and positive regression is likely.

For bettors, the comparison between actual goals and expected goals is one of the most powerful predictive tools available. Teams that significantly outperform their xG are riding unsustainable shooting luck, while teams that underperform are due for improvement. This isn’t theoretical — studies of multi-season NHL data consistently show that xG-based models outperform models based on actual goal differentials at predicting future results. The implication for betting is direct: trust expected goals over actual goals when they disagree, especially early in the season when actual goal samples are small and noisy.

The best publicly available xG models — maintained by sites like MoneyPuck and Natural Stat Trick — are free to access and update daily. Evolving Hockey offers additional advanced metrics, though some features require a subscription. Incorporating their outputs into your betting process doesn’t require building your own model from scratch. Simply comparing a team’s xG differential to the sportsbook’s implied probability can identify overvalued and undervalued teams on any given night.

High-Danger Chances and Scoring Location

Expected goals models already account for shot location, but the concept of high-danger scoring chances isolates the most valuable offensive opportunities: shots from the inner slot, net-front deflections, and rebounds. These chances convert at rates of 15-25%, dramatically higher than the 3-6% conversion rate on low-danger perimeter shots.

Tracking high-danger chance differentials provides a more focused view of a team’s offensive and defensive quality than raw shot volume metrics. A team might have a mediocre CF% of 49% but generate 55% of all high-danger chances at five-on-five, meaning they’re getting outshot overall but winning the battle for the most dangerous scoring opportunities. From a wagering perspective, the high-danger chance share often explains why a team with middling Corsi numbers is winning more games than the shot data would predict.

Defensively, high-danger chances against is one of the most telling metrics for evaluating team structure. Teams that allow few high-danger chances tend to have strong defensive systems — good gap control, effective net-front clearing, and disciplined positioning. These teams are more likely to sustain strong goaltending numbers because their goaltenders face easier workloads, regardless of overall shot volume. When a goaltender posts a .925 save percentage behind a team that suppresses high-danger chances, that performance is more sustainable than a .925 behind a team that gives up breakaways and cross-crease plays every night.

Putting It All Together: An Analytical Workflow

The practical challenge for bettors isn’t learning what these metrics mean — it’s integrating them into a workflow that produces actionable betting decisions without requiring a PhD in statistics. A streamlined approach works best for most people.

Start each day by checking the xG differential standings on a site like MoneyPuck or Natural Stat Trick. Identify teams whose actual record significantly diverges from their expected record — teams winning more or fewer games than their underlying metrics suggest. These are your primary candidates for value bets, either backing the underperformers or fading the overperformers.

Next, check the specific matchup’s Corsi and high-danger chance profiles for both teams. If Team A dominates five-on-five possession and Team B struggles to generate quality chances, Team A has a structural advantage that may not be fully captured in the moneyline if Team B’s record is inflated by lucky shooting or hot goaltending.

Finally, layer in the contextual factors that analytics don’t fully capture: goaltender matchup, rest days, travel schedule, and injury status. Advanced stats tell you what a team’s true talent level is. Context tells you whether tonight’s game will reflect that talent or whether situational factors push the probability in a different direction.

The Numbers Don’t Watch Hockey

Here’s the tension at the heart of analytics-based betting: the numbers are always incomplete. Corsi doesn’t capture line chemistry. xG models don’t account for a player’s confidence after a four-game pointless drought. High-danger chance rates can’t measure whether a team is mentally checked out after a coaching change or energized by one.

The bettors who wield analytics most effectively are the ones who use the numbers as a foundation and their own hockey knowledge as the finishing layer. The stats identify where the market is likely wrong. Your understanding of the game — the coaching systems, the player dynamics, the intangibles that don’t fit into a regression model — helps you decide whether the statistical signal is actionable tonight or whether context overrides it. Analytics without hockey sense produces a rigid model that misses game-to-game variance. Hockey sense without analytics produces confident opinions that lack a track record. The edge lives in the overlap, and it belongs to bettors who respect both the spreadsheet and the sport it’s trying to describe.

Advanced NHL statistics for betting at ice hockey bets.