As promised, (and just in time for the post lunch slacking) I've compiled similar statistics for penalty kills as I did for power plays (note: the PP statistics will be be updated to include games through 10/26 soon).

Again, the idea is that keeping track of mere occurrences is a fairly coarse metric of penalty killing efficiencies (PK%).  Instead, I determine rankings by the number of goals against per 2 minute short handed situation (PK G/2; thanks F&B!).  I argue that this is a better metric because if two teams both allow goals on every power play, but one allows them in the first 30 seconds and the other in the last 30 seconds, the second team should have a higher ranking, even if the end result is the same.  Also, teams get more credit for killing 5 minute majors in my system, as a full kill essentially counts as 2.5 kills, while under the old system it only counts as 1.  Likewise, for my whipping boy play, the 9 minute PP the Caps had vs Calgary last year gave the Flames a perfect 3/3 for killing 7 minutes of penalties.  That is the same 3/3 they would have had if the Caps remained ineffective and Nylander had not taken a minor.  In my system they effectively killed 3.5 penalties in the game situation.  They would have had 4.5 had the Caps not evened up the teams.

When looking at the numbers from the 2008-2009 and 2009-2010 seasons, we see that the old metric (PK/G (NHL)) tends to overestimate the number of short handed situations before a goal is given up (PK/G) by around 1.  For teams with better PK units it overestimates by far more than one.  I have included the number of goals allowed per 60 minutes of short handed time as well (PK G/60) because, while it is a mere scaling of PK G/2, I find it more readable.

For comparison to a rather well understood season, here are the data for 2008-2009:

The data through October 26, 2009 for the 2009-2010 season is here:

In this data we see that while Ottawa has the highest ranked power play in the old metric, Minnesota is truely the dominant team.  And boy are they good.  Their penalty kills last a full 7 seconds longer, on average, than the next closest team, 14 seconds longer than the average, and 25 seconds (almost 1/4 of a PP) longer than the worst.  Additionally, while the old metric shows that they actually would allow a goal 30 seconds earlier than Ottawa, we see that they would actually last more than 30 seconds longer.  We also see how bad they Leafs penalty kill really is.  The Leafs are allowing goals at a rate of slightly less than 1 for every 2 full penalty kills.

As for the Caps (this being a Caps blog and me being a Caps fan, after all), they aren't too shabby given the relative lack of discipline.  They don't take a hard hit in the new metric (unlike the new power play metric).

I'll be updateing these stats as the season progresses.  When grad school gets out of the way for a few hours I will post everything on an external site, rather than cluttering fan posts with updates.

Next in the series will be a look at 5v3, 5v4, and 4v4 situations.

For quite a few years I have pondered the meaning of the percentages provided as official measure of power play and penalty killing success.  To me, they never seemed to be a fair assessment of a team's PP or PK ability because

1. A PP/PK of any length counts toward the total.  Team A, Player 1 takes a tripping call at 19:57 in the 3rd and it counts as a PK when it is inevitably killed off, helping Team B's PK ranking and hurting Team A's PP ranking.
2. If Team A scores two goals on a PP from a major penalty, are they 1 for 1 or 2 for 1.  They aren't 200% on the PP for that game, are they? They were successful on 100% of their PP's but showing them as 1 for 1 is also not really indicative of the game situation.
3. In last year's painful first period against Calgary, where Tyler Sloan earned the Caps a 9 minute power play,  Washington was able to do precisely nothing with it (well, they did manage to turn it into a 7 minute PP).  On the official score sheet, this counted as 0 for 3 on the PP (2 minors and 1 major), however the team managed to score no goals on 7 full minutes of PP time.

This last one really got me thinking about this issue and I occasionally dabbled with writing a script to parse the RTSS data and figure out how good a PP team the Caps really are.  I was even more motivated (yet still managed to do nothing) when the Caps finished the year ranked 2nd on the PP.  One part laziness + one part not knowing how to quantify things + 3 parts grad school forced this to the back burner.

A couple of weeks ago I realized that PP effectiveness might be measured just like the Goals Against Average, i.e., Goals (scored or allowed) per 60 minutes.  I've adapted my parser to handle the 2008-2009 and 2009-2010 seasons (2007-2008 will come, they changed the page formatting some), so we can look at the results.  Thanks to F&B for the suggestion to look at Goals per 2 min of PP and the number of PPs per Goal.  I've added these columns to the tables and rearranged things a little to make it more readable.

First, the 2008-2009 season, since that data is complete and we have a firm idea of where teams should be:

The important things to see in this chart are the 2nd and 3rd columns, which contain the PP effectiveness as measured by the league and my method.  The 8th and 9th columns have each teams rankings under the two metrics and the last column contains the difference.

So, what do we learn from this?  The answer is not much.  It appears that those rare arbitrary situations don't affect the final rankings too drastically.  But, as fans we like to see that the Caps moved into the #1 spot.  (Don't get your hopes up for the PK data, though.)  The Devil's drop of three spots is interesting though, I'm not sure what to draw from that. 

One thing that we learn is that the Capital's have really short power plays, on average, when compared to the rest of the league.  This is clearly due to good goal scoring but I'd be willing to wager some of this is due to taking bad penalties that end PPs early.

For completeness, here is this season's data (through Friday, October 23):

Interestingly, the Caps still have a far shorter average PP time than teams that are near them in the rankings.  I wish I could explain that somehow.  We'll have to see as the season goes on, I'll update my numbers and post them again sometime a month or two from now.

An interesting note from looking at the PP/G data is that the current method for computing PP efficiency seems to overestimate the number of PPs required to produce a goal by more than 1 on average.

Stay tuned for data on penalty kills.  That data is processed but I havn't had a chance to look at it in too much detail, that and I'm still figuring out what these data mea, if anything.  Additionally, if anyone is curious, I actually have the breakdowns for one- and two-man advantages as well as both 5-5 and 4-4 even strength times, so I can spit those into a table at any time.