• No magic bullet: faceoffs and the penalty kill

    by  • July 10, 2011 • Post, Uncategorized • 20 Comments

    Just over two years ago, I got into the topic of faceoffs and how they affect your penalty kill. It’s a topic of some interest to me, given the abysmal penalty kill and power play that the Oilers have iced over the past few years. This has coincided with the Andrew Cogliano era, faceoffs in general have been a sore point and the Oilers’ inability to win a faceoff while killing penalties has convinced a lot of smart people that the Oilers need a beast of a player who can win draws on the PK.


    This is the sort of issue with which data is quite useful in terms of framing the discussion. As such, I did my best to sort through all of the power plays awarded last year, sorting them by whether the shorthanded team won or lost the initial faceoff. I should caution that there are likely to be a few errors in here – it’s more difficult to do this in Excel than you’d think – although I believe the data to be close enough that it’s not worth me investing the time to sort out any power plays that are missing. It should be almost bang-on. I came up with 7996 power plays started by a faceoff, an average of 266.5 per team, which is within 10 or so of the actual average; when you allow for powerplays that start when one player leaves the box, I suspect that this is right on.

    Having sorted the power plays by whether the team on the penalty kill won or lost the draw, I can then basically perform a comparison between two types of team: one team that wins 100% of its faceoffs when killing penalties and another that loses 100% of its faceoffs while killing penalties.

    As you’d expect, the data does seem to show some difference based on whether or not teams win the initial faceoff. When teams won the initial faceoff, they killed the penalty at an 82.5% rate; when they lost it, they killed the penalty at a 78.9% rate. At first blush, you might think that there’s something to spending money in order to ensure that you have players who can win faceoffs on the power play or when you’re shorthanded.

    The problem, as I see it, with this line of thinking, is that the choice at the NHL isn’t between 100% or 0%. The bounds are much more tightly circumscribed. I did the math to calculate the expected penalty kill percentage of a team that wins 65% of the initial faceoffs when killing penalties and a team that wins 35% of them. I am, if anything, being generous with those bounds. I come up with the difference in faceoff percentage resulting in the 65% team killing penalties at an 81.2% rate and the 35% team killing them at an 80.1% rate. Over 266 opportunities, we’re talking about 2 or 3 goals. I note that this meshes well with the answer I came up with when I went at this question a different way:

    That might be significant but I suspect when I really drill into things I’ll find that being an atrocious faceoff team on the PK will cost you, at the absolute top end, a win over the course of a season. I suspect that it’s probably more like half a win but we’ll see.

    I’ve put some thought into why the difference might be so small and the answer that I’ve come up with is that, when you’re facing an elite PP team, whether you win the draw or lose the draw, you’re going to have to taste the poison. If you lose the draw, they get to set up and throw their best at you for a minute and a little more, subject to any clears. If you win the draw and ice the puck, what happens? Generally, a defenceman hurries back while the forwards take a leisurely skate back to the red line or far blue line. They aren’t hard seconds to play. It would be interesting to take a look at the average shift length on the PP for stars in shifts where they won the initial faceoff versus those where they lost. I strongly suspect we’d find that the shifts tend to be longer when they’ve lost the initial faceoff.

    All of this is a sort of long way of saying that I’m not sure a veteran centre who can take faceoffs is the panacea some, like Lowetide, have suggested because I lean towards thinking that winning or losing faceoffs doesn’t make enough of a difference in the outcome of hockey games to matter. If Belanger is a good addition – and he’s the only addition that I think is likely to make a positive difference, so I’m a bit baffled by all the praise for Tambo’s July 1 – it’s because he’s a good hockey player during the time between the faceoffs.

    There’s another angle to this that I think warrants some discussion too. If I was coaching the Oilers, the only true centre on the team who’d see much PP time next year is Sam Gagner. Shawn Horcoff can certainly play on the PP better than he’s given credit for – he’s a pretty average PP player, which is more than you’d think given his reputation in Edmonton. I see a lot of guys on the Oilers who could provide Horcoff’s skill on the PP though and, when they’re competitive, he’ll hopefully have been bumped on merit. (I am assuming that common sense prevails and that the Nuge goes back to the WHL.)

    If winning or losing faceoffs doesn’t matter a ton, it seems to me that you want to get your difference makers out on the PP. A first PP line of Hall in between Smyth and Hemsky doesn’t seem terrible to me. Smyth has a skill set in front of the net that isn’t found elsewhere on the team. You could have Hemsky running the thing off the half wall, with Smyth moving back and forth between the front of the net and the corner and Hall circling in the slot. Or maybe you organize it differently, I don’t know. The key, as I see it, is not getting bogged down in faceoff success rate as a major determinant of one of the three spots.


    20 Responses to No magic bullet: faceoffs and the penalty kill

    1. Trentent
      July 10, 2011 at

      Are most good faceoff men good penalty killers? Maybe there is a correlation between good faceoff man and being a good PK’er that makes people think if they get one they automatically get the other, thus a good faceoff man will fix the PK (even if it’s not necessarily true).

    2. RiversQ
      July 10, 2011 at

      Probably shouldn’t get bogged down in thinking there are only three spots for PP forwards either.

    3. dawgtoy
      July 10, 2011 at

      Assume…very dangerous thing to do in regards to Oiler management.
      “I am assuming that common sense prevails and that the Nuge goes back to the WHL.”
      Now, how many 1st overall players since 1989 have been sent down? I fear your assumption is incorrect, and common sense (another dangerous thing to say in regards to current Oiler management) will not prevail. For argument sake, he and Lander were the best forwards on the ice by far this week. All things being equal, Nuge might just be what the doctor ordered for the power play.

    4. Tyler Dellow
      July 11, 2011 at

      I harp on this every so often but the empirical data is overwhelming. Young players suck on the PP. The Nuge might be the answer a couple years down the line; he won’t be next year.

    5. Eetu Huisman
      July 11, 2011 at

      Would it be possible to do the same thing for just 3-on-5 penalty killing? Common sense says faceoffs are a lot more important when you’re down by two skaters, but then again, are there enough of those situations for it to make a difference.

    6. Coach Pb
      July 11, 2011 at

      My view is that they need help on faceoffs AND the penalty kill, not necessarily both, but both is good.

    7. bone
      July 11, 2011 at

      @Eetu Huisman

      You’re probably right, but the Oilers were surprisingly decent at killing 5-3 PPs last year. I think I remember at some point late in the year, that there hadn’t been a single goal scored in an Oilers game 5 on 3 either for or against. MPS or Omark finally got one later in the year against the Canucks if I remember correctly. All in all I think it again supports Tyler’s notion that Faceoff percentages don’t make a huge difference on results.

    8. July 11, 2011 at

      To be clear… are you only counting here the first faceoff of a power play? That seems wrong, but that’s what I seem to be getting from your post.

      If so, that doesn’t seem very intuitive – we already know that faceoff results only impact the play for ~15 seconds, so making a judgement on a full 2 minutes of play based on the first faceoff is going to naturally diminish the difference between the two extremes. In addition, power plays frequently feature multiple faceoffs over their duration. If a faceoff is relevant for only 15 seconds, and the average PP has, say, 3.5 faceoffs, then you’re talking about a possible 52.5 seconds in which faceoff results are relevant, versus only 15 seconds in the original methodology. If that’s what you did.

    9. ranford4life
      July 11, 2011 at

      I like the rationale for this analysis, but this particular methodology seems to have a rather large hole. You have only considered the effect of initial faceoffs on the PK, and have not accounted for subsequent faceoffs. If a team loses 65% of their initial PK faceoffs followed by losing 65% of all subsequent PK faceoffs during the same penalty, then it stands to reason that the effects of that loss would be compounded. I am not sure how to determine the average number of defensive-zone faceoffs encountered on a given PK, but I suspect it would be in the range of 2.5 (including the initial faceoff). However, I suppose that 2xnot much is still only marginally more than not much.

      It’s interesting how your analysis meshes with Gabe’s analysis from a few years back. He stated that “one out of every 40 lost face-offs [in the defensive zone] resulted in a goal in the next 20 seconds”. His analysis was an overview of all situations (ES/PK/PP). On this basis, using the bounds that you set of 35% faceoff winning versus 65% winning, if we assume 2.5 dzone faceoffs per PK and 266 PK opportunities (your #), then the difference between winning 35% of faceoffs versus winning 65% of faceoffs would be five goals, or not quite one win. Granted, that analysis is ROUGH and the 1/40 ratio would likely be higher on the PK, but even if it is 1/20, a HUGE swing in dzone faceoffs on the PK still only accounts for a differential of three points over a season.

      It’s interesting how counter-intuitive some of this stuff seems. We place a large emphasis on the importance of puck possession with Corsi and Fenwick measures, and it seems logical that winning a faceoff and controlling the puck would materially contribute to possession. Given this analysis, that doesn’t necessarily seem to be the case.

      Far more material to increasing the Oilers chance of success would be acquiring another league-average goaltender…

    10. Tach
      July 11, 2011 at

      Your analysis is based on just one faceoff per powerplay, but I looked at the Oilers numbers for last year and they have 538 faceoffs in 518:37 of penalty kill time. (ok, I get 537.75 draws when I add up all the 4 v 5 draws and divide by 4 so if I botched this up, let me know) 400 of those were defensive zone.

      That’s one faceoff every 57 seconds and a defensive zone faceoff every 81 seconds which is two per complete penalty kill (one at the start, one at the 1:21 mark). This rate probably increases when you consider that some of the penalty kills end early with goals against.

      I am also willing to bet that if someone looked, a lost faceoff at the start of a penalty kill leads to an increase in the odds of there being a second faceoff as a shot on net might be covered or a subsequent penalty taken.

      So your effect is at least twice what you describe, and possibly higher depending on if there is a multiplier effect.

    11. BRIdub
      July 11, 2011 at

      I agree with the premise faceoffs are less important than they are made out to be, but do think in 5v3 situations they are much more important. Generally speaking a 5 on 3 is significantly shorter than 2 minutes so 15-20 seconds of extra possession in the attacking zone is significant, and further, the conversion rate on a 5 on 3 should also be significantly higher.

    12. July 11, 2011 at

      Average of 127 initial faceoffs per team, the difference between a 60% faceoff man and a 55% faceoff man is 12-13 faceoffs.

      I was pretty on board with ‘situational faceoffs’ belief, particularly because I’m a Canuck fan and Manny Malhotra is boss, but I think I’m going to restate this somehow. Malhotra is a terrific player because of what he does between the faceoffs, not necessarily what happens on the faceoff.

    13. Hawerchuk
      July 11, 2011 at

      A faceoff win on the PK should be worth about six seconds, or 5% of the PK length. That’s a 1% difference in PK rate, no? Interesting that the delta is larger.

      Generally the difference between the best and worst faceoff takers is less than one win per season. But I think a Pavelski/Nichol PK in late-game high-leverage situations brings you a bit more.

    14. David Staples
      July 11, 2011 at

      I’ve gone at this same issue a different way, just charting the Oilers on faceoffs and seeing how many wins and losses are part of scoring plays.

      After doing this a few seasons, it strikes me that what I’m finding lines up with what you’ve found here, that faceoff wins and losses aren’t that important to the scoring of goals and the winning of hockey games.

      Based on how often this is mentioned by commentators and fans, the faceoff is likely the most over-rated aspect of a hockey game.

      In the end, it’s just one lost battle among hundreds of battles in a game, and it’s not as significant a loss as many, as the defensive team is organized and ready to defend when the battle is lost.

    15. slipper
      July 12, 2011 at

      I think there would be more value in seeing whether players can win possession in the seconds after a draw and exit their zone. Faceoff data isn’t very clean to begin with. There seems to be far more than two outcomes to a faceoff and often there is more than the two players involved in arriving at that outcome. Yet only the centers get credited with either a W or an L, and that barely communicates anything at all.

    16. Good Muckin' Tonite
      July 12, 2011 at

      Great discussion and stats. I’m more a flow guy (that’s game flow, not hair)than a stats guy, but to me it breaks down as follows: your scoring chances increase with the more quality shots you take. A SH face-off win means taking away possession and quality scoring chances, and eliminating time for the opponent to have both. Same goes for on the PP. Face-off win = better scoring chance = +PP%. I see it as a significant puck battle every draw.

    17. Vic Ferrari
      July 15, 2011 at

      Good post and terrific commentary all around. When William H. Internet invented the world wide web … I think this what what he had in mind.

      Just generally, I agree with the sentiment that fans and commentators overvalue faceoffs. I also agree with the sentiment that our math undervalues them. ‘Wins’ and ‘Losses’ are hard to define, and clean wins are very different from scrums after a draw was tied up. Winning board side matters more than winning to the middle. Winning against Crosby/Malkin matters much more than winning against whoever is playing on their second PP unit.

      Just generally, you see a lot of very good coaches obsess on faceoffs in this league. And selection bias tells us this has probably been a successful practice, i.e. coaches who dwell on this have survived.

      I know you’re quick to write otheres off as fools, Tyler. I lean towards giving them the benefit of the doubt. This knowing that they have watched thousands of hours of video, and I have no plans to do the same.

    18. speeds
      July 15, 2011 at


      I just wonder about something like pulling the goalie, and how coaches handle that. The studies I recall suggest that teams pull their goalies markedly later than they optimally should, but that trend continue. In American football, supposedly teams gamble on 4th down less frequently than they should.

      I know these aren’t exactly the same thing as the value of faceoffs, but could it be that coaches have been conditioned to think that faceoffs are really important from a young age, and that underlying premise isn’t often challenged?

      I wonder the same thing about shooting at the empty net late in the 3rd, even though it potentially causes an icing? Has there been a careful study on whether the costs outweight the benefits, or do teams just not shoot at the empty net because of the potential criticism, and that’s just the way things have always been done? That’s not rhetorical by the way, I truly have no idea, maybe it really is a terrible idea to shoot at the empty net from your end and risk an icing, up a goal with 43 seconds remaining?

    19. Vic Ferrari
      July 16, 2011 at


      My guess is that coaches are a bit more conservative than they should be. Because the conservative play is less likely to be criticized.

      Having said that, in the case of pulling the goalie, I think our math is far too aggressive. pull the goalie too early and you’ll end up with a matchup of Zetterberg/Datsyuk vs Fraser/Jacques … and that will probably end badly.

      As a junior coach, Andy Murray was famous for pulling the goalie very early. He once played 25 minutes of a game with the goalie pulled when he coached in the Manitoba equivalent of the AJHL. When asked why he didn’t bring the same philosophy to the NHL, he said that the other teams have too many good players. There’s truth in that I think.

    20. Name
      July 18, 2011 at

      Can someone please clarify for me?

      The author states that the difference between a team winning 65% of faceoffs and 35% of faceoffs is only 1% efficacy on the PP or PK, but looking at the table, it says a team like the Sharks, for instance, score on the PP (or get scored against on the PK) 13% more often when they win the faceoff on the PP, or lose it on the PK. How many goals is that? 255 PPs, 255 PKs in a season, that’s 510 total. If they won 100% of their faceoffs, that would be .13 x 510, a 66 goal difference. Taking it down to winning 60% of their special teams faceoffs, which seems much more accurate, that’s .6 x 66 = 40.

      So it seems like special teams faceoffs, not even counting even strength ones, gave the Sharks an extra 40 goal difference last season. Unless the 255 number was powerplays and PKs combined, in which case it’s 20, and that’s still very, very significant.

      I just went and checked the stats, and I don’t know where the 255 or 265 numbers are coming from. Not one team had that many powerplay opportunities last season. The Sharks had 277 PPs and PKs combined last year. That number x .6 to account for a 60% faceoff winning percentage = 166. 166 x .13 from the table above = 21.5 goal difference for the Sharks on special teams due to faceoffs.

      Am I going wrong here somewhere? That seems about right to me, actually, from watching them play last year. But let me know if I’m reading the table wrong or something.


    Leave a Reply

    Your email address will not be published. Required fields are marked *