Using GAR and Game Score to identify who is driving the bus for the Leafs

As hockey statistics continue to advance, they continue to use ideas first created for other sports. Right now the biggest focus in advanced hockey statistics is on creating a one number statistic which attempts to summarize a player’s total value to his team. This idea was first brought to light by baseball, of course, and it was named WAR (wins above replacement.) WAR has gained a lot of traction in the baseball world and there are now WAR models in basketball and, more recently, hockey.

Game Score

Dom Luszczyszyn’s (@domluszczyszyn) “Game Score” model is based on the basketball statistic which goes by the same name. Game Score is designed to be a rough summary of a player’s performance on a per game basis. It combines and weighs basic statistics: goals, primary assists, secondary assists, shots on goal, blocked shots, penalty differential, faceoffs, 5-on-5 corsi differential and 5-on-5 goal differential. You can read Dom’s more thorough explanation here, but the formula is this: 

Player Game Score = (0.75 * G) + (0.7 * A1) + (0.55 * A2) + (0.075 * SOG) + (0.05 * BLK) + (0.15 * PD) – (0.15 * PT) + (0.01 * FOW) – (0.01 * FOL) + (0.05 * CF) – (0.05 * CA) + (0.15 * GF) – (0.15* GA)

Here is the Leafs’ depth chart with each individual’s Game Score per game, as well as the team’s total Game Score per game average (top left) and each line/defence paring’s average combined Game Score per game. Via https://gamescorecharts.wordpress.com/atlantic/

Leafs Game Score 2016-17 to March

Up front, the Leafs look really good. Surprise: Auston Matthews is the best player on the team by a decent margin. The Leafs also have six forwards providing top line value, while Connor Brown is producing like a second liner. Hyman and Boyle rate very closely to Brown, but are producing at a high-end third line rate. Komarov also shows up as a third liner in Game Score. When combining each line’s Game Score, the top three lines look really good. The Leafs have two lines producing at a first line rate, while their third line is producing at a second line rate. The problem up front, not surprisingly, is the fourth line. Matt Martin, he of a four-year ten million dollar contract, shows up as “replacement level,” meaning he is not worth a roster spot, while Soshnikov is operating at a fourth line pace. The good news here is that Brian Boyle is a huge improvement as the fourth line centre. More good news for that line came today when the Leafs called up Kasperi Kapanen and are going to put him on the right wing in place of Soshnikov who is injured. All of a sudden that’s a decent looking fourth line, in theory.

The defence is a bit of a different story. Not surprisingly, Gardiner and Rielly show up as the best defencemen on the team, both producing at a “2D” rate. This means they’re both playing like your average top pairing, complimentary guy as opposed to the “number one defenceman,” teams so heavily covet. This is not surprising and it seems to confirm the notion the Leafs need to find a top-flight, number one guy, preferably on the right side. Connor Carrick and Nikita Zaitsev both show up as “second pairing defencemen,” which is no surprise. Again, like the bottom line up front, the problem is the bottom pairing. Hunwick is fine, rating as a third pairing guy, but Polak shows up as “replacement level,” which should surprise no one. I think a bottom pairing of Marincin-Marchenko would be ideal, but I digress.

The team as a whole averages out to a .49 Game Score, good for third in the division behind the Bruins (.59) and the Canadiens (.52). Most metrics seem to point to those being the best three teams in the division, but the Senators keep winning one goal games and the Bruins’ 97.7 PDO (second worst in the league, ahead of only the Avalanche) seems to point to them getting fairly unlucky.

Goals above replacement (GAR)

Dawson Sprigings (@DTMAboutHeart) has also created a single number statistic, using six inputs:

  1. Even-Strength Offense
  2. Even-Strength Defense
  3. Power Play Offense
  4. Drawing Penalties
  5. Taking Penalties
  6. Faceoffs

Sprigings model also uses ridge regression in an attempt to account for quality of competition and quality of teammate, which is another topic regularly discussed on hockey twitter. By “discussed” I mean “people yell at each other online.” You can read Sprigings’ five part series detailing everything you need to know about his model here. Today, Sprigings released his updated GAR data for this season which you can find here. Here are the Leafs results:

 

A couple of things of note right off the top. First, Auston Matthews again shows up as the Leafs best player by a decent margin. No surprise there. Second, it is important to note that this is total GAR, rather than rated to adjust for TOI, so keep that in mind.

GAR agrees with Game Score on a lot of players. They both really like Matthews, Kadri, JvR and Bozak, but GAR rates Nikita Zaitsev as the team’s most valuable defenceman this season. Gardiner ranks pretty close to Zaitsev when you adjust for TOI, but Rielly (the third highest ranked Leaf defenceman) ranks significantly lower than those two. This seems to be mostly due to Rielly ranking quite poorly on the defensive side of things, with a -1.8 even strength defence rating, compared to Gardiner’s 1.9 and Zaitsev’s .9. Rielly still ranks decently in GAR due to his positive offensive impacts, but GAR really dislikes his defensive abilities (which it probably should.)

Another interesting thing which stood out to me is the JvR-Bozak-Marner line. GAR loves their offensive abilities and hates their defensive abilities which, again, it probably should. GAR seems to think that their offensive strengths heavily outweigh their defensive deficiencies, so all three players rate well individually.

Sprigings’ model gives Komarov a lot more credit than Dom’s Game Score, strictly due to defensive impact. GAR has Komarov with more than double the defensive impact of the next best Leaf in that department which, when combined with a good penalty differential, results in Komarov ranking quite high, whereas he doesn’t show as well in Game Score.

GAR doesn’t love Nylander nearly as much as Game Score (or I) do but a big part of that seems to be his bad penalty differential. Since he’s a rookie it’s hard to say whether this is something we should expect to improve over time, but he’s never taken many penalties over his young career in the AHL or the SHL. On that note, it’s pretty impressive that Kadri ranks so high in this model despite having a negative penalty differential for the first time in his career. In fact, Kadri usually owns one of the league’s best penalty differentials due to his ability to draw a ton of them. The refs seem to have put their whistles away when they see 43 involved, but he’s still managing to be a big contributor.

Other than that, Game Score and GAR seem to be in agreement over the bottom of the lineup: it’s been bad.

Ultimately, both models also agree that the Leafs are good up front and not so good on the back end, but it doesn’t take a complicated statistical model to see that.

Leave a Reply

%d bloggers like this: