Navigation: Jump to content areas:

Pro Quality. Fan Perspective.
Around SBN: Long Beach Grand Prix 2011
Sprint-network-bar 01

What Drives Power-Play Success?

SAN JOSE, CA - MARCH 5: Ian White #9 of the San Jose Sharks on a Dallas Stars power play dives in front of Brenden Morrow #10 of the Stars to try and block a shot in the third period during an NHL hockey game at the HP Pavilion on March 5, 2011 in San Jose, California. The Stars won the game 3-2. (Photo by Thearon W. Henderson/Getty Images)

We've spilled what must by now be millions of words at this site on the subject of shot differential (aka - Fenwick or Corsi) and how it's the best predictor of future team success, primarily because it's the most persistent team-level talent we can identify.  It's even possible that we've made some headway convincing people that this is true.  But for some reason, it's hard for people to make the leap to seeing the other implications of this relationship.


For example, I looked at the Sharks' PP and PK performance.  By power-play and penalty-kill "efficiency", the Sharks are #3 and #26 in the league in those two categories, respectively.  But is power-play "efficiency" - the percentage of power-plays where a team scores - the best way to determine a team's true talent on the PP?  Let's look at the most persistent talents on the PP and how much they regress to the mean:


PP% 70


What I've done here is look at even and odd games over the last four seasons and run a regression between each PP metric in those two bins.  Shooting rate - shots for per 60 mins at 5-on-4 - is by far the most persistent talent.  PP efficiency and especially shooting percentage regress much more heavily to the mean.  A team's shooting percentage in one half of its games has almost no predictive value for the other half of its games.

If we want to know how well a team will do in the future on the PP, what should we look at?  Let's regress each of these quantities in the even half of each team's game against their PP efficiency in the odd half of their games:


Odd PP%
Even PP% 70
Even SF/60 61
Even Shot% 89


Again, this shows the regression to the mean of each variable.  Shooting percentage very little predictive data as we might have expected, but surprisingly (or not surprisingly to those who follow the line of reasoning behind shot differential metrics), the rate at which teams shoot on the PP is a better indicator of their future power-play "efficiency" than their past power-play efficiency.

A number of commenters have expressed discomfort that I don't believe that the team with the highest power-play "efficiency" is the team with the best power-play.  But remember that power-play "efficiency" is a flawed measure of talent - it's merely a sample of a team's ability.  And ability is not the same thing as performance.

Comment 14 comments  |  Add comment  |  1 recs  | 

Do you like this story?



But remember that power-play “efficiency” is a flawed measure of talent – it’s merely a sample of a team’s ability. And ability is not the same thing as performance.

It’s merely a sample of a team’s performance. (And ability is not the same thing as performance.) Don’t you think you’re confusing enough people without the typos? Yeesh Gabe.

by MattF on Apr 9, 2011 11:22 AM EDT reply actions  


No, observed performance (like PP efficiency) is a sample of a team’s talent.

by Hawerchuk on Apr 9, 2011 11:23 AM EDT up reply actions  

OK, so I don’t disagree, I got tripped by your sentence I guess. PP% isn’t a sample of performance, it is performance by definition, & a sample of ability. Check.

Anyway, stepping backward, I think this is one of the areas where teh statzis haven’t made (or sold) the case very well — how appropriate is it to credit actual performance?? As you’re well aware, one of the ways to really infuriate people is to insist that their team has been lucky, not good. And sometimes it’s very obvious just from watching, but sometimes not so much.

Take Jarome Iginla’s game against St. Louis last Friday as an example. He scored 2 goals and 1 assist, and all 3 were outstanding plays. Now in one sense, that’s obviously lucky, because he’s not scoring 160G/240pts in a season, and it’s not because he’s usually lazy and indifferent. On the other hand, insisting to someone that Jarome had a lucky game would be like tattooing ASK ME ABOUT MY SPREADSHEETS on the forehead.

Do you have a good blurb on how to reconcile this? (Someone must have written one w.r.t. baseball).

by MattF on Apr 9, 2011 11:52 AM EDT up reply actions  

I think Vic and Tyler have frequently made the point that you need to take a multi-year sample to assess true talent.

There are a bunch at

Here were my pieces on ‘always bet the under’ – players who score a lot of points will regress the following season:

I’m not really sure what would make the case better than that. Seems like pretty good evidence that performance exceeds talent and we need to be more conservative in our assessments.

by Hawerchuk on Apr 9, 2011 12:23 PM EDT up reply actions  

I understand all that… I guess what I’m asking is, do you know of a nice way of saying

Every great individual performance you ever saw was lucky

or are you happy just leaving it at that. To me, this is a major rift between “us” and even smart, open-minded types like say Ron Maclean and Bob McKenzie… if you’re insisting that there can be no “magic” in a wonderful moment that already happened (i.e. that it was just the hockey gods, or the loaded dice rolling, or what have you), then I don’t think that bridge will ever be joined.

For my part, I think there’s an actual difference between saying That’s Not Repeatable on the one hand, and That Wasn’t Real on the other. Even if it wasn’t.

by MattF on Apr 9, 2011 1:04 PM EDT up reply actions  

Well, Vic would tell us that the second anyone had some serious money on the line, they’d be much more open to this idea. (Back in November, a WSJ reporter bet me that Sidney Crosby would finish the season with more than 130 points. He came back the next day and stipulated that Crosby had to play 79+ games or the bet didn’t count.) And I think the hockey gods have had little patience for luck the last few years: the last four cup winners were the #1 or #2 teams in the league; Canada won the Olympic Gold Medal.

All of this comes out of forward-looking analysis – I have no quarrel with people enjoying what just happened; I want them to be realistic about the future. Do you think Ron Maclean would be hostile to the notion that Daniel Sedin and Corey Perry aren’t likely to continue scoring at this rate? I think they’re more rational than we might give them credit for being.

by Hawerchuk on Apr 9, 2011 1:38 PM EDT up reply actions  

I also think that looking at the top 20 scoring leaders in the following season is a great exercise for people. The data is accessible and the math is simple. And it justifies being conservative…

by Hawerchuk on Apr 9, 2011 1:39 PM EDT up reply actions  

Every incredibly bad individual performance was lucky, too…only it was bad luck instead of good. We’ve had this exact same argument time and again about the word “luck” and how it’s actually been part of the statistical lexicon for a long time. I mean, we could be patronizing and use the word “magic”, but not all good luck is magical either. You have a hell of a lot more goals produced by “un-magical” lucky bounces than you do Ovechkin weaving past Coyotes and finishing his shot with his arms practically behind his head.

We might use words like “magical” or “amazing” when recounting a play, or saying what it was like to see Teemu Selanne during his 76-goal, 132-point season, but when it comes to analysis you can’t assume that all luck is “amazing”, or even good. “Amazing” and “magical” plays are in the minority when it comes to luck.

Does he call it Luongo underwear?

Co-Manager at Behind the Net

by Bettman's Nightmare on Apr 9, 2011 1:41 PM EDT up reply actions  

The Canucks first unit is Sedin-Kesler-Sedin, with Ehrhoff and Edler. The second unit is much fuzzier, usually a combination of Raymond, Burrows, Higgins, Torres, Samuelsson, Salo, Bieksa, etc.

The on-ice shooting percentage when the first unit is on the ice ranges from 17.54% to 18.02%, but all 5 are within half a percentage point. The player immediately behind them is Samuelsson – likely because he took Edler’s spot at the point when Edler was injured. The second unit is all over the place, due to lines being jumbled and whatnot, but they range from 10 to 15%.

The first unit gets a lions share of the time; without checking the numbers I’d say 65%+ of the total PP time. Kesler and Daniel are both top-10 in PP goals, I’m pretty sure.

Are you suggesting that the first units PP efficiency, as measured by the NHL, will certainly drop, as their shooting percentage is unsustainable? Or is it not possible that a select group of players can maintain a consistently higher than average shooting percentage? The powerplay seems a logical place for that to happen, as they’ll frequently have possession in the offensive zone.

The difference in both NHL-measured efficiency, and shooting percentage, between both units has been large, and it’s been large since the first week of the season more or less; almost solely due to the first unit, the powerplay has been tops or top-3 in the league most if not all year.

I’m not sure if there’s anything worth looking into there, but the good news is that they’ll very likely have the same unit next year (all are signed except Ehrhoff who is quoted as wanting to stay).. So there will likely be a larger sample size available..

by Bort Sampson on Apr 9, 2011 1:40 PM EDT reply actions  

“is it not possible”

Anything is possible, but you have to prove the theory!

We get into this all the time, right? I say X with data, someone says they believe Y is possible…But the data associated with X says Y is exceedingly unlikely. And so a personal belief in something improbable is good enough. I can’t really win here.

by Hawerchuk on Apr 9, 2011 2:23 PM EDT up reply actions  

On a related note Gabe, the Sedins have had an on-ice SH% at EV at 10.5%+ for three straight seasons. Is this a statistically significant in showing true talent above league average or is there a realistic probability this is an anomaly? I don’t have the chops to figure this out on my own.

by Pitseleh on Apr 9, 2011 2:47 PM EDT up reply actions  

we bump up our talent estimate of the Sedins very slightly. We have a lot of history with these guys and we don’t expect them to develop new talents at age 30.

by Hawerchuk on Apr 9, 2011 7:12 PM EDT up reply actions  

Thanks, Gabe. I was the one who essentially asked the question (I asked if you put the ranking of teams by shots and PP% at mid point of the season, which one matches closer to the team’s second half PP%), and you showed me the numbers. Looking at the numbers, past PP% is not a bad predictor of future PP%, but not as good as shots for. And shots for is probably not be as good as Fenwick, and to me, Fenwick is probably not as good as puck zone time (if they ever track that).

by SJKel on Apr 11, 2011 4:33 PM EDT reply actions  

User Tools

Hockey Analysis

FanPosts Recent Stories

ANAHEIM, CA - APRIL 13:  Towels are displayed on seats prior to the start of the game between the Nashville Predators and the Anaheim Ducks in Game One of the Western Conference Quarterfinals during the 2011 NHL Stanley Cup Playoffs at Honda Center on April 13, 2011 in Anaheim, California.  (Photo by Jeff Gross/Getty Images) +6 updates

NHL Playoffs Schedule: Times, TV Information For Friday's Four Stanley Cup Playoff Games

GLENDALE AZ - OCTOBER 21:  Jarret Stoll #28 of the Los Angeles Kings warms up before the NHL game against the Phoenix Coyotes at Arena on October 21 2010 in Glendale Arizona. The Coyotes defeated the Kings 4-2.  (Photo by Christian Petersen/Getty Images) *** Local Caption *** Jarret Stoll

Jarret Stoll Suspended 1 Game For Hit On Sharks' Ian White

WASHINGTON, DC - APRIL 13:  Michal Neuvirth #30 of the Washington Capitals makes the save against the New York Rangers in Game One of the Eastern Conference Quarterfinals during the 2011 NHL Stanley Cup Playoffs at Verizon Center on April 13, 2011 in Washington, DC.  The Capitals defeated the Rangers 2-1 in overtime.  (Photo by Bruce Bennett/Getty Images)

Rangers Vs. Capitals, Game 2: New York Looks To Score More, Rebound From Tough Game 1 Loss

More from >


Hawerchuk_small Hawerchuk

Gary_bettman_bad_dreams_small Bettman's Nightmare

Resident Curmudgeon

Bens_grandpa_small Ben's Grandpa