November 2, 2011
Angles and Caroms
More Than Just PK%

by Jonathan Willis

The Toronto Maple Leafs have a pretty ironclad hold on the title of "Worst Penalty-Killers In Hockey" in the post-Lockout NHL.

This is the club, after all, that has finished 24th, 27th, 29th, 30th, 30th, and 28th in the six seasons since hockey's work stoppage. The 2009-10 edition is the worst penalty kill, in terms of overall percentage, that the NHL has seen in that span. If that weren't enough, Toronto's 2008-09 penalty kill comes in at second-worst, and the 2010-11 group makes the bottom 10.

A funny thing happens when we rearrange those teams by goal differential—that is, by goals scored minus goals allowed—those terrible Toronto teams lose their grip on the league lead. Here are the ten worst teams since the Lockout, by goal differential:

Worst Penalty-Killing Teams, by Goal Differential

Team		Season		GF	GA	GD
Washington	2005-06		15	116	-101
Pittsburgh	2005-06		13	113	-100
N.Y. Islanders	2005-06		5	99	-94
Los Angeles	2005-06		15	104	-89
Atlanta		2005-06		13	102	-89
Phoenix		2005-06		9	98	-89
Phoenix		2006-07		5	92	-87
Toronto		2006-07		3	90	-87
Vancouver	2005-06		6	93	-87
Florida		2005-06		5	91	-86

There is a definite trend to this list, and one that probably jumps out at the reader immediately: eight of the 10 worst clubs by goal differential come from the 2005-06 season, with the other two coming from the 2006-07 season. But since 2005-06 wasn't a particularly bad year in terms of penalty-killing percentage, why is this? 57-69-72-73-87

There is a simple explanation: the league called way more penalties in 2005-06. New Jersey was the most disciplined team in the league in 2005-06; they played short-handed 348 times. That would have been the worst total in the NHL in 2010-11. That's why the list above is dominated by teams from 2005-06—no matter how good a penalty-kill they ran, they couldn't possibly manage better goal differential totals than later teams, because they were on the penalty kill so much.

It all points to a flaw in the way we evaluate penalty-killing efficiency. As it stands, NHL rankings are based on a single variable: the percentage of power plays in which the opposition scores. In reality, there are three separate variables that impact how penalty-killing affects any given team:

1. The number of penalties taken
2. The percentage of penalties taken successfully killed (what the NHL currently tracks)
3. Short-handed goals

Take, for example, the 2010-11 Ottawa Senators and New York Rangers. Under the NHL's current system, they put in identical 83.7% success rates. Yet, the Senators finished with a goal differential of minus-42 while the Rangers finished with a goal differential of minus-31, 11 goals better.

Half of that total is short-handed goals—the Rangers scored 11 to the Senators' six. The other difference was the number of penalties taken; the Rangers took 12.6% fewer penalties, which over a whole season added up to 37 fewer penalties to kill. That saved the team six goals over the season, approximately equivalent to an entire win.

When a team wants to improve its record by cutting down goals against, the penalty kill is typically a prime target. Defensive specialists may be brought in, and coaching staffs turned over in search of an answer. Is that an easier approach than cutting down on penalties taken?

Let's start by looking at three penalty kills—league worst, league average, and league best. For comparison's sake, we'll pretend each of these penalty kills was short-handed 291 times, the average from last season. Because we want to catch the entire penalty-killing effect of these kinds of moves—which includes short-handed goals—we'll use an adjusted metric. Instead of dividing successful kills by times short-handed, we will divide successful kills plus short-handed goals by times short-handed—in effect, cancelling out a goal against for each goal the team scored while on the kill.

The best penalty kill by these numbers since the NHL Lockout comes from the 2008-09 Minnesota Wild—they finished with a paltry minus-27 goal differential despite being short-handed 291 times (90.7%). The worst penalty kill belongs to the 2009-10 Maple Leafs—they finished minus-69 on 288 total kills (76.0%). The league average over this span is 84.2%.

Team			Adj. Success Rate	Goal Differential
Post-Lockout best		90.7%			-27
Post-Lockout average		84.2%			-46
Post-Lockout worst		76.0%			-70

A team going from the worst performance in the league to the very best, while being short-handed an average amount of times, could improve to the tune of 43 goals—the equivalent of seven wins.

Now, we're going to try these numbers the other way—we're going to pretend that each team had an 84.2% adjusted success rate, and tinker with the number of times short-handed. Doing so is made slightly more complicated by the fact that penalties called have dropped since 2005-06, but we can get around that.

Since the NHL Lockout, the worst percentage differential between league-average penalties taken and most penalties taken is 18.8%—in other words, the least disciplined team in the league took 18.8% more penalties than the average team in that year. The most-disciplined team managed to take 31.8% fewer penalties than league average. Let's run those numbers, using the 291 penalties taken by an average 2010-11 team, and compare goal differential.

Team			Times Short-handed	Goal Differential
Most disciplined		198			-31
Average				291			-46
Least disciplined		346			-55

In a single year, the gap in terms of goal differential between the most and least disciplined clubs in the league is relatively small—24 goals, roughly equivalent to four wins. That's a little over half of the gap we saw when we compared the most successful percentage kills to the least successful.

It seems that the best route to addressing goal differential problems on the penalty kill remains to fix the penalty kill. Still, a combination approach must be regarded as a good idea—a team that combined the title of most efficient and most disciplined could conceivably end the year with just a minus-18 goal differential. It would be a mistake to think that a good percentage penalty kill takes all the damage away from taking too many penalties.

Jonathan Willis is an author of Hockey Prospectus. You can contact Jonathan by clicking here or click here to see Jonathan's other articles.