In a game I was watching not too long ago a series of events transpired to make me question the validity, or at least accuracy, of a fundamental NHL statistic.  During the game a minor penalty was taken.  However, on the ensuing faceoff, the team with the advantage committed a minor erasing the power play.  So what you ask?   Well the official score sheet will show that the original offending team was 1 for 1 on the penalty kill (and this would repeat on the back end for the 2nd penalty for the other team) by simply killing off a measly 2 seconds of clock time.

This got me to thinking.  How often does this happen?  A successful penalty kill lasting less than the standard 2 minutes?  Does this occur enough over the course of a season where it would misrepresent a team’s true penalty killing ability? 

I wanted to see if I could find an answer to these questions.  As a result I began tracking every penalty kill for the 2009-10 season.  I’ve used the official score sheets from NHL.com to see how every team is faring while down in manpower. 

My initial theory is that the current measurement does not accurately reflect a team’s true capacity for killing penalties thanks to shortened penalties skewing the success rates.  If the average time of a successful kill can be shown to be impacted by scenarios akin to the opening paragraph I’d like to propose a new measurement.  I’d then like to compare this measurement against the current statistics to see if there is any difference in the rankings.

First there is the existing data.  As we enjoy the Olympic break there have been 921 games played in the NHL this season.  During these 921 contests there have been 6961 officially recorded manpower advantages.  The total time spent in the box, league-wide, is 11,275 minutes and 22 seconds or nearly 188 hours.  Currently the NHL calculates PK success as goals allowed per times short handed with no regard for the amount of time killed.  According to NHL.com the official rankings for penalty killing effectiveness:

Chart

Now that we know who the NHL thinks are the most effective killers of penalties I’d like to start to look at numbers that interest me.

So.  How long is the average successful penalty kill? 

Unfortunately, since I was taking high level numbers from the score sheets, I do not have a 100% sample of penalty kills.  What I do have, however, is the ability to parse out those games where teams successfully killed all penalties in a game and the time served.  Using this method I have captured 2928 short handed situations out of the 6960 overall. This 42% sample size at a 99% confidence level gives us an average penalty kill time within a 2% margin of error

These 2928 advantages were successfully erased in a total time of 5,204 minutes for an average of 1:46 per kill.

I find it interesting to note that the needle moved nearly a full 15 seconds off the standard 2 minute kill.  On average the penalty killers only have to successfully combat 88% (14/120) of a called pen.

More data is required to see if this is consistent year over year but for the purposes of killing time during the Olympic break I am going to use this as my standard.

My next contention is that a penalty kill unit that gets scored on 1:59 into an advantage was more effective than a unit that allowed a goal within the first 10 seconds.  Not rocket science here.  The longer you go without allowing a goal while a man down the better you are at killing penalties.

I make this statement to pave the way for introducing a new way of measuring penalty kill effectiveness. 

Where am I going with this?  Something akin to an earned run average in baseball.  Baseball pitchers are held to a standard that puts the runs they allow against a standard over innings pitched.  I’d like to do the same for hockey.  Take the goals allowed, against a standard, over minutes served.  Speaking formulaically

Baseball:  ERA is calculated as Earned runs*9 / innings pitched.

Hockey:  I’d like to propose:  PP goals allowed*1:46 / minutes short handed

If we plug in the numbers farmed from the score sheets we get:

TSH = Number of short handed opportunities

PPGA = goals allowed while down a man

PK% = penalty kill percentage – league measurement

Time SH = actual amount of clock time with numerical disadvantage

PKA = Penalty Kill Average – New Overmars measurement

NHL Rank = PK rankings as determined by NHL currently

PKA Rank = PK ranking as determined by goals allowed and total time served

Well.  Now that the math is done we see that there is no real change at the top or the bottom of the rankings.  The existing and the proposed measurements both recognize who is the best and the worst of the lot.  What I’m liking about the new measurement is that it allows for further definition for the middle of the pack:

We can clearly see that the inclusion of time into the measurement does allow us a deeper dive into determining the worth of a team’s penalty kill.  Take, for instance, the numbers for Pittsburgh and Calgary.  They have both allowed 41 goals with a manpower disadvantage.  Since Pittsburgh has faced one more penalty kill the NHL says they are .1% better than Calgary in this category.  However, Calgary has faced nearly 10 more minutes of PK time, and under our new format is rewarded for allowing fewer goals per time served. 

I concede that there is not nearly enough historical data available to ensure that the standard used in the PKA is valid, however, seeing the average time for a penalty kill drop so significantly and positions changing so dynamically makes me believe that this measurement is worthy of further research.  To be continued?