In hockey, the scoreboard lies more than in almost any other sport. A team can dominate possession, generate 40 shot attempts, and still lose 1-0 on a fluke deflection. Advanced stats cut through this noise.
Corsi: The Possession Proxy
Hockey has no possession statistic like football. Instead, analysts use shot attempts as a proxy. Corsi counts every directed shot attempt — on goal, missed, or blocked.
Corsi For% (CF%) = Team shot attempts / (Team + Opponent shot attempts) at 5v5
A CF% of 55% means your team generates 55% of all shot attempts when both teams are at even strength. Over 30+ games, CF% is one of the strongest predictors of future goal share.
The 2024-25 NHL average is 50% by definition. Elite teams sustain CF% above 53%; struggling teams fall below 47%.
Fenwick: The Cleaner Version
Fenwick removes blocked shots from the equation. The logic: a blocked shot may mean the defending team positioned well rather than the attacking team generating poorly. In practice, CF% and FF% produce nearly identical rankings.
PDO: The Luck Detector
PDO = team shooting percentage + team save percentage (at 5v5). The league average is always 100.0.
A team with PDO of 103 is benefiting from its shooters converting at above-average rates AND/OR its goaltender making saves at above-average rates. History shows that extreme PDO values are overwhelmingly driven by luck and regress toward 100 within weeks.
Betting Application
- High PDO team (103+) with mediocre CF%: Overperforming. Their record will worsen. Bet against them.
- Low PDO team (97-) with strong CF%: Underperforming. Their record will improve. Bet on them before the market adjusts.
A £10 moneyline bet at 2.40 on a low-PDO, high-Corsi team returns £24. If the true probability is 50% rather than the implied 42%, the expected value is £2 per bet.
Expected Goals (xG)
The next evolution beyond Corsi. xG assigns a goal probability to each shot attempt based on location, shot type, and game state. A shot from the slot carries 0.15 xG; a shot from the blue line carries 0.02.
xG models produce more accurate predictions than raw Corsi because they account for shot quality, not just quantity. A team with fewer shot attempts but higher xG is generating better chances.
Practical Framework
Before betting an NHL game: (1) compare 5v5 CF% for both teams over the last 20 games, (2) check both teams' PDO for regression signals, (3) verify the starting goaltenders and their recent save percentages, (4) account for back-to-back scheduling and travel. This workflow catches value the market overlooks in the early weeks of each season.