The Extra Attacker: Weekly Thoughts


Every week or two, I have a few thoughts. I’ve decided to call this the Extra Attacker. This week I look at whether the Jets’ size led to wins, re-consider the corsi data on Ron Hainsey discussed yesterday at the Nation, and consider the critique of Hainsey as too passive for his size. Below is a smattering of my synaptic output.

Being Big

Lindsey: "They hear the name Tobias, they think, ‘big, black guy.’"
Tobias: "Well, clearly I’m not a ‘big’ guy. I’m not a Carl Weathers par example."

Arrested Development Season 4 came out this week, so you’ll have to excuse the non sequiturs. It’s tough to think of anything else. On to hockey!

This year, James Mirtle recorded the size and age of every team’s starting lineup in January. (You can view he raw numbers here) The range isn’t huge – Winnipeg had the tallest average team at 74 inches, Montreal the shortest at 72.1" for example – but I was curious to know if size or age seemed to matter for Winnipeg’s rate of success. For reference, Winnipeg was the tallest team, the 4th heaviest team, and tied for the 26th oldest (i.e. 4th youngest).

There are problems with using the data the way I’m about to – the Jets didn’t play agianst every team’s opening night roster all year (or with their own), the sample size of games played is small, and it’s questionable as to whether aggregate numbers like this tell us anything about how teams use their size or age. But it’s a thoughts column, not a peer reviewed paper, so let’s start with the low hanging fruit.


See a pattern? If we won against small teams more than large, we’d expect a line going from bottom left (bigger teams, lower win %) to top right (smaller teams, higher win %). If we lost more to small teams (say, because of speed), we’d expect the opposite line. I see… a blob. Yep, it’s a blob. We lost every game to the tiny Canadiens, and we lost 4 of 5 to the monstrous Capitals. In between, there doesn’t appear to be a relationship at all.


Annnnd…!! It’s a blob. In actual fact, it’s a sort of mound shape, meaning that we did worst against teams at the extremes and best against those closer to average height. It would make more sense to our brains if it were a straight line relationship, but I see a parabola.  I’m not sure that fits with any narrative we might already expect, and it’s not enough data to start trying to conceive of an inductive model just yet. Okay, age!


Oldest opponents on the left, youngest on the right. This time we have a sort of bowl shape – the opposite of the graph above (if we ignore the Buffalo outlier at the top… I’m so happy we played Buffalo three times this year). So we did best against the oldest and youngest teams. We might imagine that teams ‘in their prime’ were the ones to beat us. That may be true. Florida tied with New Jersey for the oldest team on opening night. Remember when Florida signed Kovalev? (That could easily read, ‘remember when New Jersey signed Steve Sullivan?’ and be just as funny to me) Tampa Bay was 5th oldest. Carolina and Toronto tied us for 4th youngest. We did well against all of those teams. That they are young or old is part of them being bad teams – Carolina is developing, Florida tried to buy a team through Free Agency, New Jersey hasn’t drafted a quality player since Parise, Tampa missed their window, and Toronto… I’m going to avoid the hate mail on that one. At the same time, all teams have a mix, so we’ll have to dig a little more into this data as time moves forward.

Tobias Funke's headshots(Seriously, watch Arrested Development if you haven’t already)

Context and Corsi Data

One of the best pieces Tyler Dellow ever wrote in defence of advanced (or fancy) stats was this one. In it, he discusses the importance of considering context and expectations when using corsi (or any shot attempt) data. This in opposition to ‘high corsi good, low corsi bad’ thinking. Looking at the relative corsi data for defencemen who played tough opposition with poor zone starts yesterday, I was struck by how much a coach can affect our fancy stats through idiocy or stubbornness.

Carl Gunnarsson had a brutal assignment this past season – his zone start was sub-40%, his quality of competition extremely high. He was a shut down guy – a defenceman expected to have a negative shot-attempt differential, but also to reduce the number of high quality chances against the world’s best players. What struck me was that his relative corsi was an incredible -0.5. Given his assignment, that is astounding… Until you look at the whole defence group.


NAME TEAM POS GP TOI/60 Corsi Relative Corsi On
JAKEGARDINER TOR D 12 16.84 30.5 -1.19
MIKEKOSTKA TOR D 35 17.37 5.9 -6.22
CODYFRANSON TOR D 45 14.6 5.1 -10.41
MARKFRASER TOR D 45 14.78 2.8 -12.27
JOHN-MICHAELLILES TOR D 32 15.31 2.6 -10.53
CARLGUNNARSSON TOR D 37 17.12 -0.5 -14.21
DIONPHANEUF TOR D 48 17.9 -7.3 -18.16
RYANO’BYRNE TOR D 42 15.56 -15.7 -17.44


The point of Relative Corsi is to normalize for team-effect. It’s on-ice corsi minus off-ice corsi. I used Carl Gunnarsson to ask how good Ron Hainsey is yesterday, and was met with the answer ‘Ron Hainsey is a bum.’ When we look at the raw scores, however, we can see that Carlyle’s insistence on playing Holzer completely changed the dynamic of Relative Corsi for the Leafs. A sane coach would have benched Holzer or given him softer minutes long before he had this kind of affect on the team. In Toronto, though, the guy played 22 games against the league’s top players, and so their relative corsi has a ridiculous 55 point spread. LA has the next largest spread in its defence group at just under 40, with the huge majority of teams showing a spread of just 10 to 20 corsi events per 60 minutes.

Meanwhile, Ron Hainsey’s corsi on / 60 mins was a mild -6.53, or less than half as bad as Gunnarsson’s. We noted that Hainsey was 15th of 21 in Rel Corsi among his assignment-peers. He moves up slightly to 13th if we measure by raw Corsi-on per 60 minutes.

Ron Hainsey and Passive Hockey

Ron Hainsey can’t buy a friend on twitter these days. (That’s a joke – he obviously could if he wanted to – he’s so rich!). In reaction to my posts yesterday about Hainsey’s comparables, Ryan Blight of Arctic Ice hockey tweeted (jokingly):

I know Ryan was kidding – poking fun at an argument put forward by Mitch Kasprik (@WpgJetshocktalk). Still, it’s a real argument by many people. In fact, some crazy guy in Brandon has decided to ignore Hockey Canada’s ruling about hitting in Peewee hockey to start his own league… just to let 12 year olds run into each other. To some, hitting is not only part of the game, it’s a necessary part of the game.

So I’ve decided to answer a joking question with a serious table. Here is yesterday’s table updated to include (and sort by) hits as recorded by the NHL. I’ve also added Corsi On for reference. Hits isn’t an official stat – it’s counted by the teams – so it’s not hard science. But it’s almost the end of the article – why start with credible data now?

NAME TEAM Corsi Relative Corsi On Zone Start/Finish Diff P/60 A1% Hits
BRENDENDILLON DAL 4.5 0.14 6.4 0.49 25% 133
BROOKSORPIK PIT -18.1 -13.08 1.4 0.57 37% 119
SHEAWEBER NSH -2.1 -7.41 3.2 0.84 45% 112
DANGIRARDI NYR -6.7 -0.2 1.4 0.73 20% 102
ZDENOCHARA BOS 6 14.21 1.4 0.94 81% 101
TIMGLEASON CAR -11.5 -4.04 5.6 0.7 37% 90
JOHNNYBOYCHUK BOS 0.1 8.53 -1.9 0.48 60% 79
ZACHBOGOSIAN WPG -5 -3.29 4.2 1.06 17% 79
CARLGUNNARSSON TOR -0.5 -14.21 5.2 1.33 15% 78
BARRETJACKMAN STL -15.7 -6.97 5.6 0.82 57% 70
JACKJOHNSON CBJ -8.1 -9.45 8.3 0.74 50% 66
JUSTINFAULK CAR -5.4 -1.33 5.5 0.98 25% 63
TRAVISHAMONIC NYI -11.4 -7.23 -2.1 0.41 0% 59
MIKEWEAVER FLA -4.3 -5.31 3.5 1.23 38% 49
DANHAMHUIS VAN 6.6 7.66 2.8 1.11 42% 47
RONHAINSEY WPG -8.7 -6.53 10.3 0.75 45% 36
RYANSUTER MIN -2.3 -2.56 -5.6 0.85 30% 33
ANDREJSEKERA BUF -9 -16.18 3 0.94 37% 30
ANDREWMACDONALD NYI -10.8 -6.74 -1.3 0.61 28% 26
JONASBRODIN MIN 4.9 1.88 -2.4 0.42 17% 17
PAULMARTIN PIT -4.9 0.19 1.9 1.34 56% 17


So for the record, Ron Hainsey is as passive as Ryan Suter and Dan Hamhuis. I can live with that.



Don’t forget to sign up for StreakCred for round 3 of the playoffs. StreakCred is the the Nation Network’s new playoff pool game and for $10 you can compete for a trip to Octoberfest! Partial proceeds go to charity, so sign up here.