About NHL Player & Goalie WARs

The ScotiaStat NHL WARs are my take on player contribution during the course of a season. The model is inspired by the works of those before me, but probably most heavily influenced by the wonderfully descriptive work that Dom Luszczyszyn has been putting out since he introduced gameScore.

The ScotiaStat NHL Player & Goalie WARs use 100% public data from MoneyPuck, PuckIQ and NHL. The WARs are measured in Wins Above Replacement. Replacement is measured as the (number of teams * 12 + 1)th forward, (number of teams * 6 + 1)th defenseman and (number of teams * 2 + 1)th goalie. For example, in 2014 a replacement forward would be the 361st ranked forward by WAR, a replacement defenseman would be the 181st ranked defenseman by WAR and a replacement goalie would be the 61st ranked . In 2024, those numbers change to 385th, 193rd and 65th respectively.

WAR is accumulated throughout the course of the season and are displayed on player cards the same way, as opposed to a prorated value. Prorated value is important contextually as if a player were to miss part of the season due to injury, the prorated value would show their pace if they hadn't missed games. With that being said, prorated value can be very misleading without context. For example, Laurent Brossoit's 2014-15 season gives him the best single season prorated WAR since tracking began in that season. For those (like me) who thought there had to be some sort of mistake, it wasn't. Brossoit saved 3.86 goals above expected in his only game that season. If he didn't place first in prorated WAR, there would be a problem. I could have filtered him out of the dataset, but he's important for team wins, which I'll get into further down. 

Player WARs are the sum of Offensive and Defensive WARs. 

Offensive WARs are calculated using goals, primary assists, secondary assists, expected goals, powerplay impact, offensive impact, faceoff differential, penalties drawn and giveaways.

Defensive WARs are calculated using matchup strength, blocked shots, faceoff differential, penalties taken, penalty kill impact, defensive impact and takeaways.

Goalie WARs are calculated using high danger expected goals, high danger goals, high danger shots, medium danger expected goals, medium danger goals, medium danger shots, low danger expected goals, low danger goals and low dnager shots.

A dataset with WARs for the past 12 seasons can be found in the google sheet attached.

• One of the limitations of the MoneyPuck dataset, is that a players team, is only the team they finished the season on, even though their statistics will be from the whole season. This will affect players moving between good and bad teams midseason. ie can't distinguish between how many goals against came from the first vs second team they played for.

• Team/Line results might be impacting team stats too heavily.

• There is currently no measure for quality of teammates.

On the left side, there are 3 WAR values.  The outer circle around the WAR value shows the player's percentile rank among all players in the position. The circle split into 10 parts is to show you what percentile the player falls under. If the player's percentile falls in the first and brightest red section, than they are in the 0-10th percentile in WAR (which is not good). If the player's percentile falls in the last and brightest green section, than they are in the 90-100th percentile in WAR (which is very good). This concept applies to all three WARs on the left side of the card. For goalies, in place of an Offensive and Defensive WAR, are GSAX and GSAX at five on five. 

The values displayed inside the circles are the player's actual WAR values. Using Connor McDavid's card below, he has 7.08 Overall Wins Above Replacement, 6.80 Offensive Wins Above Replacement and 0.28 Defensive Wins Above Replacement. The average WAR for each position and WAR type, is shown below the WAR header. For example, in the 2022 season the average forward's Overall WAR was 0.5 and the average defenseman's Overall WAR was 0.34. 

The card also shows a bar chart to display the percentile ranks of the player in their position for each of the components that go into WAR, as shown on the chart. The first card shown below is Connor McDavid. It shows his percentile ranking beside each category with each bar sized to represent a value between one and ten percent. For counting categories, being closer to 100% means leading in that category. This is flipped for Penalties Taken and Giveaways.

Underneath the logo of the team the player plays for is how they are deployed by the average minutes they play per game. For goalies, this is based on games started.

Underneath the player's cap hit on the card, there is a number shown for Value, which is there to represent the player's worth at that point in time. Value is also shown in the table underneath the graph on a per season basis. Value is determined by a few different measuring techniques and for every season, is closely matches the total salary spent on players. This way of valuing players ties the value of the player to the WAR they provide. On a year to year basis, this can change drastically. If a player misses the majority of the season due to injury, their value will suffer. Intuitively this makes sense as a significant injury history impacts the way they are viewed.