oates - Marianne Helm

Three-piece. (Photo: Marianne Helm)

Inspired by an article about the Winnipeg Jets on Arctic Ice Hockey, Tyler Dellow, aka mc79hockey, created a nifty color-coded chart to visualize how puck possession is distributed throughout the Edmonton Oilers lineup.

Basically, you build a grid of players– with the forwards along the left and defensemen along the top, and you rank them by ice time. Then you list the possession percentage for each forward/defenseman pairing. Because coaches typically give better players more ice time, you’d expect to see the higher numbers at top left and the lower percentages at bottom right. There are exceptions, but that’s the basic idea.

It’s a novel way to get a quick, visual impression of a team’s makeup, plus it gives us a chance to reap some insight from how Adam Oates uses his roster.

Great possession percentages– above 55% —  are powderpuff pink (a Sunday favorite). Merely positive possession percentages– above 50% —  are canary yellow. Negative possession percentages– under 50%– are an ugly blue color that I regret using now. Basically, it looks like an Easter egg. (Because I’m nice, I didn’t pick a special color for possession scores under 40%, but there are a few.)

While I included Dmitry Orlov and John Erskine in the chart, they don’t get colored as they haven’t played ten games yet, so their numbers could be a little screwy. Same story with the empty cells; the forwards and defenders there haven’t played at least twenty minutes together.

The basic idea is that the top-left should be filled with good performances in yellow and pink, and the bottom-right should be filled with schlubs in fugly blue. Let’s see how it plays out.

Carlson Green Alzner Schmidt Oleksy Urbom Orlov Erskine
Ovechkin 52.7 54.4 55 49.4 47.4 48.1 49.3 55.6
Backstrom 51.7 56.8 55.6 48.9 40.9 45.4 50 55.6
Johansson 49.2 56.9 51.2 49.2 46.7 51.6 54.1 55.9
Brouwer 44 50.8 45.5 51.6 42.5 38.2 56.9 52.5
Laich 43 48.1 43.5 52.8 44.9 38.3 48.8
Ward 51.5 45.3 46.9 53.4 55.5 46.3 53.8
Grabovski 51.6 48.7 51.5 50 50 43.5 51.2 60
Erat 46.3 53.3 47.3 56 47.8 46.3 41.3
Chimera 52.2 41.6 49 50.6 52.7 43.9 46.2 52.5
Fehr 52.4 49.7 47 58.3 55.7 49.2 54.9 46.9
Beagle 42.6 46 35.1 55.4 50 63.3
Volpatti 37.6 39.5 33.3 48 41.8 33.8 40
Wilson 49 45.7 46.7 52.5 44.5 43.7 31.2

It kinda turned out like that.

The top left is chock full of good stuff. The Ovi-Backstrom-Mojo line is almost entirely above 50% when playing with the top-3 defenders (Carlson, Green, Alzner). Nobody– almost nobody– doubts the Caps’ top-end talent. 

Then, along both axes, there’s a big drop-off. Two supposed second-liners, Brooks Laich and Troy Brouwer, are almost completely in that nasty sub-50 blue color, though Brouwer is on the rise of late. Purported 4-D Nate Schmidt doesn’t so fare well with the top line, but — in a shocker–  he’s driving play very well with the third and fourth lines. Weird synergy. We’ve noticed before that Schmidt made Mike Green a better player; I’d like to know more about him.

The third line of Jason Chimera and Joel Ward have been startlingly strong, which is why the running joke for more than a week was that they were the actual second line. Their centers have been the excellent Mikhail Grabovski and Martin Erat, so it shouldn’t be all that surprising.

After that, we get into trouble– which was to be expected, but maybe not this dramatically. Defender Alex Urbom and the fourth line aren’t just getting outplayed on the ice; they’re getting dominated. We see an anomalous uptick when Beagle and Wilson kick it with Schmidt, but other than that it’s uniformly bad.

On the flip side, the success of the third line and Nate Schmidt make the Capitals surprisingly deep in places we wouldn’t have expected.

Most of the distribution is typical: good guys getting lots of ice at top left, less-great guys getting less ice time at bottom right. All of this does, however, advance the hypothesis that while Adam Oates has done a commendable job getting Alex Ovechkin to bloom again, the rest of the garden needs tending. It shouldn’t have taken twenty games to realize that Alex Urbom ought to take a seat, or that Mikhail Grabovski should be promoted, or that Laich and Brouwer shouldn’t be playing together.

The ongoing, macro discussion is why Washington sags in its depth, and this chart will help move that along. Why have we seen some players– ones with the initials A.O.– improve while others (like Brouwer and Green) decline? What effect has Adam Oates’ mandate for breakout tactics had on possession among puck-carrying blueliners? Is the team truly happy with its defensive roster, and will they remain that way after John Erskine ends his 9-game bid for the Norris Trophy? I’d like to hear your thoughts.

For context, Here are a few other teams’ charts shared by Dellow, which may be useful for comparing to the Caps:

Seem to me the Oilers are neither well constructed nor well deployed.

And that’s a pretty shallow Montreal squad.

And here’s a team that would get relegated if the league would allow it.

Cam Charron also did one for the Canucks over at CanucksArmy.com:

wowy

That’s probably the best-looking team so far, but I want to see more. A few teams I’m curious to see: Carolina, Minnesota, Anaheim, New Jersey.

This is a new and interesting tool, and I won’t pretend to have an absolute grasp of it. Let me know if you see something I might’ve missed.

Notes: These possession stats are Corsi%. That number represents the Caps’ share of total shot attempts during 5v5. The numbers were pulled from stats.hockeyanalysis.com. Players must have played 10 games to appear in the grid (Erskine and Orlov are exceptions), and each pairing must have been on-ice for at least 20 minutes.

No related posts.

 
  • Katth

    So simple, but it gives such an insight.

  • Tadd

    Wow, Peter. This just blew my mind. Fantastisc article! Thanks! This is why RMNB is the best hockeyblog I’ve ever read. This and Ian’s shenanigans.

  • billy

    love this, but am i missing erat on the grid? was he left off because he has been jumping around the lineup all season?

  • Joe

    This might be the coolest piece of advanced statistics I’ve seen yet. It really helps visualize what’s working and what’s not. Sidenote: any reason for Erat being left off of your chart? I’m pretty sure his numbers would be solid, although his play of late has not been as commendable. Either way, this is an awesome article!

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    ah transcription error— I’ll fix! Thanks

  • Topher Gee

    This is a great way to look at 5v5. I think it answers two of my questions: Grabo should always be 2C & we are pretty terrible on 5v5.

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    He’s in there now– sorry for the oversight and thanks for the heads-up.

  • X

    I’m an overall fan of #fancystats but don’t they seem to be a little biased towards offensive defenseman?

  • http://www.russianmachineneverbreaks.com Ian Oland

    That’s really all I’m here for. CAUSING TROUBLE. (And embarrassing the blog by losing a video game challenge to a 19-year-old on national TV)

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    Good Q. A raw possession number like Fenwick or Corsi *may* make a stay-at-home, shutdown D-man look bad, but not always.

    A good defenseman doesn’t just limit goals against, he limits shots against. His style of play limits the number of opponent shots, whereas an offensive defenseman would increase the number of Caps shots.

    Offensive defenseman: tons of Caps shots, lots of opponent shots

    Stay at home defenseman: limited number of opponent shots, hopefully more Caps shots

    So when we express possession as a %, they should look similar.

    But you’ll still see blueliners who get put up against the best forwards in the league, and while they get outpossessed, they’re still slowing down their opponents to some extent.

    That’s a great example of why we shouldn’t use possession stats alone to measure a player. Zone starts, ice time, match-ups, quality of linemates are all important in figuring out what kind of player a guy is and how he’s doing.

  • billy

    boom. great work, peter!

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    Ranked 24th overall in puck possession during 5v5. http://behindthenet.ca/fenwick_2013.php?sort=6&section=tied Worrying.

  • pixiestix

    Hi Hendy in a Caps 3rd sweater…I still really miss you.

  • Brendan Maltese

    Erat is on there. Peter, you missed Erat’s powderpuff pink with Schmidt

  • JH

    Very interesting. Would be more interested in how Caps fare vis a vis our nemeses, PIT, PHL, NYR.

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    ugh, you’re right. This is hard!

  • Red

    Nice tables but…..needs moar pie charts!!

  • Jack Conness

    So, statistically speaking, Schmidt ain’t half bad?

  • Derek Eklund

    Great article.

  • prut

    Beagle+Orlov=bromance? How many more games until we find out for sure?

  • dylan wheatley

    fehr is such a rockstar

  • CDizz

    WHY WOULD YOU SHOW HENDY IN THE MAIN PICTURE? MY WOUNDS HAVE NO HEALED.
    Also, solid article.

  • riggorules

    This chart confirms my doubts about Volpatti (four of six boxes under 40). Granted, that fourth line gets tiny minutes, but not sure why he keeps playing over the likes of Fehr, Erat, Latta and Beagle, who have all spent extended time in the press box or Hershey. Because he punches faces? Or am I missing something?

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    There’s one caveat. Because he’s a fourth liner, he might be expected to play a defensive role against the opposition’s best scoring lines. Except that’s not true– Volpatti’s opposition spends the least amount of time on ice (25.8%), so they’re not the scoring lines.

    And while Volpatti starts a lot of non-neutral shifts in the DZ (51.6%), that’s not very dramatic.

    He’s been dragging down team possession pretty significantly since the middle of November (Sun. Nov. 17, 2013).

    It’s a problem. Like I’ve said: I’m not sure this team is very good at talent evaluation.

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    They’ve only skated together for 21:01 in 8 games. I don’t think it’s a sturdy number yet.

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    BOO

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    Yeah. I’ll look into more, but it looks like he’s at least a good 3rd pairing D-man. And based on his synergy with Green, maybe even more.

  • riggorules

    I suspect Oates has no interest in #fancystats and places an unusually high value on obscure characteristics like making what he deems to be the right “read,” or being able to receive passes with your backhand, ect. At least that’s how I explain some of his lineup decisions to myself.

    I also think he values shot quality over quantity and as a result, I feel like the Caps get some good shifts in the offensive zone that don’t produce any SOG.

  • http://www.russianmachineneverbreaks.com/ Peter Hassett

    The team has its own analytics effort led by video coach Brett Leonhardt. They track a microevent called “touches” but I don’t know exactly what that means yet. It could be better than our publicly available data. It could be leading them down a rabbit hole.

    If they’re tracking zone time (and I’m pretty sure they are), they know the Caps spend most of the time in their own zone. If they track zone time by individual player, they’ll get data that correlates pretty close to what we have via Fenwick and Corsi.

    I think you’re positively correct about Oates’ opinions on tactics like backhands and shooting orientation. I don’t know if those are “obscure” as we’ve seen how much tactical and positional improvements have helped Ovi, but I do think it’s not helping the whole team uniformly– like we’d expect an effective system to do.

  • riggorules

    Brett Leonhardt, you mean our emergency goalie!

    I didn’t know that about their analytics, thanks.

  • X

    Thanks dawg

  • JPA

    “While I included Dmitry Orlov and John Erskine in the chart, they don’t
    get colored as they haven’t played ten games yet, so their numbers could
    be a little screwy.”

    Translation: There is no way Erskine is a possession monster. Sample size must be too small.

  • http://www.russianmachineneverbreaks.com/ Peter Hassett
  • Red

    What kind of a monster boos pie charts?! They’re like the statistical equivalent of puppies; dumb, pointless and adorable.