Why weaker penalty kill performers are still sometimes worth using

Recently we discussed one area where Jim Slater is a bit overrated, face offs. It is not Slater’s fault, but the impact of a face off win is not that large.

There is one area though that Slater may potentially provide value and that is on the penalty kill.

Let’s take a look.

First, here is the performance of the Jets most commonly used penalty killers this season.

Screen shot 2015-03-03 at 2.50.32 AM

The raw numbers show the player’s shot attempts against per sixty minutes and their differential in terms of a percentage. The impact on linemates is a WOWY (With or Without You) of either statistic weighted for time on ice.

Using the player rates and the average shooting percentage on the penalty kill, we can estimate the player’s impact in terms of goals. While skill may impact shooting percentage in one direction or the other, the shortened sample size in this case makes league average shooting percentages superior to non-regressed numbers.

Slater actually comes out the worst penalty killer of the bunch. This is not that surprising, as most hockey skills are transferable. The more skilled players tend to “outskill” others in most areas. While this is not always true and specialists do exist, they are often rare than what some may tend to believe.

This does not mean there is no value in having Slater on the penalty kill.

There is an opportunity cost. Players can only play so many minutes without fatigue taking an effect, especially late in the season where most players are dealing with minor injuries. Minutes that Slater takes on the penalty kill means more availability in skilled players taking even strength and power play situations.

So, to compare, we need to also estimate each player’s impact at even strength.

There is no highly representative goals above replacement model that includes all the current factors that we know exist in the game. I have constructed one previously using Corsi percentage and time on ice. 

Screen shot 2015-03-03 at 2.51.50 AM

While the model would grossly underestimate superstar percentage-drivers like Sidney Crosby or Steven Stamkos, it does pick up that there are differences in on-ice percentages experienced for different lines. 

It is simply an act of comparing the options.

Screen shot 2015-03-03 at 2.33.09 AM

Values are expected goal differential per 60 minutes of play.

At a strict 1:1 minute ratio, it is more effective to use the higher-skilled forwards than the options like Jim Slater.

In all likelihood the transfer is not a 1:1 ratio. Penalty kill minutes can be tiring, which is one of the reasons why most teams have three penalty kill forward units while only two power play units.

Here is an example if placing a skilled player on the penalty kill costs two minutes at even strength.

Screen shot 2015-03-03 at 10.39.54 AM

Values are same but 5v5 player and 5v5 Slater options have been doubled.

Despite being the weakest performer in both even strength and penalty kill, it could be beneficial to place Jim Slater on the penalty kill in place of certain players.

Now this was a pretty simplistic look at the variables, but shows why coaches make particular decisions at times. The caveat though is the simple model also shows how some coaches may assume incorrectly. 

Coaches generally have a good perspective on trends, but statistics should play a role with even the best of coaches. No coach will be able to see the impacts in a manner to appropriately compare impacts such as the above example.

The above example is also why effective penalty-killing skaters for the fourth line can be such a bonus. Again, statistics can aide in detecting which players actually impact the game effectively and sufficiently.

Raw numbers curtesy of stats.hockeyanalysis.com and behindthenet.ca