GAR and Cap (2018-2019): Introduction

Welcome to this years’ installment of the annual evaluation of NHL contracts. Using Goals Above Replacement (GAR) values provided by Evolving-Hockey and cap hit numbers from CapFriendly, the purpose of this article is to determine which players performed above or below their contractual expectations.

Before diving into the numbers, first a quick primer on Goals Above Replacement (GAR). For the specifics on how Evolving Hockey calculates these values:

Goalie GAR/WAR

  • Fenwick shot-level regularized logistic regression. The target variable is 1 (Goal) or 0 (No Goal) and the predictor variables are:
    • all shooters (dummy variable, 1 = skater took the shot, 0 = skater did not take the shot)
    • all goalies (dummy variable, 1 = goalie facing the shot, 0 = goalie not facing the shot)
    • expected goal value of the fenwick shot (probability between 0 and 1)
    • strength state (8-level factor variable: 5v5, 4v4, 3v3, 5v4, 5v3, 4v3, Ev5, Ev4)
    • shooter score state (3-level factor variable: shooter team is leading, tied, or trailing when shot was taken)
    • shooter venue (home or away team)
    • shooter position (forward or defenseman)
    • back-to-back (home team / away team playing with < 2 nights of rest)
  • The goalie coefficients that result from this regression are interpreted as each goalie’s impact on the odds of a goal being scored. These coefficients are converted to probabilities and multiplied by the total fenwick shots each goalie faced to arrive at “Goals Above Average” (GAA).
  • Goals Above Average is then converted to Goals Above Replacement (GAR) as such:
    • GAA + (FA * replacement level GAA per fenwick)

Shooter GAR

  • The shooter coefficients from the regression described in the Goalie GAR/WAR section.
  • The shooter coefficients that result from this regression are interpreted as each shooter’s impact on the odds of a goal being scored. These coefficients are converted to probabilities and multiplied by the total fenwick shots each shooter took to arrive at “Goals Above Average” (GAA).
  • Goals Above Average is then converted to Goals Above Replacement (GAR) as such (replacement level is set per forward and per defenseman using the same cutoff described in the Replacement Level (Skaters) section):
    • GAA + (iFF * replacement level GAA per fenwick)

Simply put, GAR is an attempt to quantify the contribution of individual players in all game situations (even strength, short-handed and power-play). In theory, GAR represents the number of more or less goals with said player relative to a replacement-level player. In other words, GAR is a quantification of the individual players’ contribution compared to replacement-level talent. The larger the number, the greater the individual contribution. Negative numbers indicate that the individual player is performing at levels below replacement level.

Starting with a general overview of GAR (limiting the scope to players with a minimum of 200 total minutes played) and organizing the data by team. Here’s how players performed for their respective teams:

GAR by Team (D)GAR by Team (F)

There are many different conclusions that one could draw from the graphs above. For brevity sake, I’ll highlight a few themes that stood out to me.

Montreal was exceptional at 5v5 hockey, more so than the standings would indicate. Special teams was a huge anchor that prevented this team from realizing its true potential. This was even more pronounced when examining their special teams GAR values for defensemen.

Tampa Bay was off the charts in nearly all game situations (which is not surprising considering they tied the all-time NHL record with 62 wins). Their first-round sweep at the hands of Columbus was shocking, although the Blue Jackets did have the 3rd best even strength GAR in the league…and Bobrovsky.

Chicago is a very interesting team to keep an eye on next year. Based on their performance last season, they boast an above-average group of forwards at even strength and the power play. On the other hand, the defense was bad last year, and even more atrocious at special teams. Stan Bowman is already making moves to remedy that situation, which should put this team back in playoff contention.

I’m calling my shot now: The Florida Panthers will be in the playoffs next year with the addition of Connolly, Stralman and Bobrovsky. Florida already had the 3rd and 2nd highest power play GAR from their forwards and defense, respectively. Hoffman is entering his second season in Florida and the new free agent additions should propel this team into the playoffs.

Columbus is going to be in a tough spot next year after going for broke at the deadline. This isn’t really a hot take; Anyone will tell you that losing Panarin, Duchene and (possibly) Dzingel is going to hurt your playoff chances. At even strength, there is some cause for optimism even with the loses of Panarin (13.5), Duchene (10.4) and Dzingel (5.4), Columbus is still above average. The greater cause for concern is the power play, which looks benign without Duchene and Panarin.

This was merely the introduction to a multiple-part series that will explore GAR, player valuation and cap hits. The next part of this series will focus on defensemen and GAR. If you want to explore the source data that I used, you can find it here.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s