Analytics in ice hockey

Lore & Background

Hockey Hall of Fame coach Roger Nielson is credited as an early pioneer of analytics, using measures of his own invention as early as the late 1960s during his tenure with the Peterborough Petes. In modern usage, analytics were traditionally the domain of hockey bloggers and amateur statisticians. They have been increasingly adopted by National Hockey League (NHL) organizations themselves, reaching mainstream usage when the NHL partnered with SAP SE to create an enhanced statistical package that coincided with the launch of a new website featuring analytical statistics during the 2014–15 season.

Corsi, named after coach Jim Corsi but developed by an Edmonton Oilers blogger and fan, was originally created to better measure the workload of a goaltender. Today it is used to approximate shot attempt differential and predict future goal differentials. Fenwick, a variant of Corsi, excludes blocked shots and is viewed as having a stronger correlation to scoring chances. PDO, derived from the online handle of Brian King, sums shooting percentage and save percentage and is often viewed as a proxy for luck, with most teams expected to regress toward a sum of 100.

Zone starts measure the ratio of offensive to defensive zone face-offs a player takes. A high zone start ratio often increases Corsi numbers, while a low ratio depresses them. However, use of zone starts in analysis has decreased, as it has been determined that on-the-fly shifts account for more than half (58%) of all shifts. Score effects and situational modifiers are also considered, with statistics like Corsi close and Corsi tied restricted to when one team leads by one goal or the game is tied. Corsi close was later scrutinized for not predicting future goals as well as unadjusted Corsi.

Reader's Guide

Analytics have reshaped how player and team performance is evaluated in ice hockey. Corsi and Fenwick provide approximations of puck possession by counting shot attempts, allowing analysts to predict future goal differentials. A player or team with a Corsi For percentage above 55% is often considered elite. PDO serves as a quick check on luck, with values over 102 suggesting a team may not be as good as they appear, and values below 98 indicating they may be better than their record shows. Newer statistics like RAPM and xG address limitations of earlier metrics. RAPM uses ridge regression to isolate a player's play-driving ability by accounting for factors such as shot volume, location, teammates, competition, score effects, zone starts, and even-strength state. xG assigns a quality value to each shot attempt based on location and type, solving the problem that Corsi treats all shots equally. These tools have moved from the domain of bloggers and amateur statisticians to official NHL adoption, with the league partnering with SAP SE to provide enhanced statistics. The legacy of analytics is a more nuanced understanding of performance, though debates continue over the best methods and adjustments for score effects and situational play.

Did You Know?

• PDO is not an acronym; it comes from the online handle of Brian King, who first proposed it for forums and Counter-Strike.
• Corsi was developed by an Edmonton Oilers blogger and fan, not by coach Jim Corsi after whom it is named.
• RAPM and xG were created recently for hockey but had been used in other sports before.