Note: for this post to be worth reading you need to have an understanding of what Corsi is. If you want a resource to figure that out, try this.
Stephen Burtch of Pension Plan Puppets has been working on a statistic for individual player possession that takes into account outside factors: zone starts, team effects, faceoff wins, and more. dCorsi, the statistic, measures the difference between a player’s Corsi and their ‘expected Corsi’ – the result if that player’s minutes had been played by “a perfectly average player in an average season if he was handed the same minutes with the same players against the same opposition”. Expected Corsi is calculated using a multivariate linear regression of the previously mentioned factors and others, including position, age, and TOI. Simply, a player is driving possession if their Corsi is higher than their Expected Corsi. It is split up further into dCorsi For (offense) and dCorsi Against (defense). Burtch explains the intuition and the mechanics of dCorsi much better than I have here.
I thought it would be interesting to have a look at some of the dCorsi results for the Canucks. Interesting not only in evaluating the roster but also to look at the differences in past player performances under two very different coaches. AV, of course, employed some of the more extreme usage strategies in the league to great effect with the Sedins for years. Tortorella, on the other hand, used the Sedins in a much more comprehensive role. Killing penalties, playing more minutes each night and with tougher competition, the strategy worked well – for about 20 games – and then it all fell apart. This post has ended up being a lot longer than I expected it to and so I’m going to split it up with one post for forwards and another for defenceman.
Above are last year’s dCorsi results (per 20mins) of all current Canucks forwards. Linden Vey doesn’t make the list because he’s only played 18 games and Shawn Mathias isn’t included because he played for two different teams last year (complicates the expected Corsi regression). A few immediate impressions. Dorsett had a hell of a year in New York last year, Henrik’s numbers were clearly affected by playing with his injured wrist, and Tom Sestito is still Tom Sestito. Below I’ve graphed Corsi and Expected Corsi to show the difference in contextual factors for some of the Canucks play drivers (ranked by decreasing Corsi20).
The main thing that stuck out to me is how high the Expected Corsi is for the Sedins even playing more and playing tougher minutes under Tortorella than they have in their careers. They stand out especially compared to a guy like Dorsett, who excelled even as Alain Vigneault did everything but make him fly the team plane and drive the zamboni between periods.
To extend the sample beyond just last year, I’ve calculated the Average dCorsi Impact over the last 3 seasons for the same 11 players (skaters with less than 3 seasons of dCorsi data are marked with an asterisk). In Burtch’s words, “As dCorsi is a normalized stat per 20 minutes of 5v5 TOI, dCorsi impact is determined by multiplying the player’s yearly dCorsi by the number of 20 minute intervals they have played in each season. To determine the player’s 3 year total, the yearly dCorsi impact values are then combined. The Average dCorsi Impact is determined by dividing the total by the number of seasons played by the skater in the data set”
dCorsi Impact is useful over a multi-season sample in showing players who have thrived in larger roles: more value is given to Daniel Sedin playing 20 minutes a night than Derek Dorsett playing 10.
Extending the sample puts right one anomaly in the single season dCorsi data: Henrik Sedin is a positive possession player. I also wouldn’t count on Derek Dorsett to play anywhere near as well as he did last season as one of the top dCorsi players in the league, but he still looks like a very useful bottom-six piece going forward. Past this point I’m going to discuss a couple instances of Canuck forwards for whom dCorsi is showing things that we might not have been missing using Corsi and CorsiRel.
Vrbata’s counting stats stand a very good chance of rising next season. Not only because he will likely be playing with the best linemates he’s had in his career in the Sedins, but also because he can expect his contextual factors to soften significantly. Compare his Expected Corsi to that of the Sedins: he actually had a negative Expected Corsi last year, which he outperformed significantly. There is a new coach in Vancouver and nothing is certain, but Vrbata is likely going to see usage totally different from what he saw in Arizona. He’s a good hockey player.
Nick Bonino/Zack Kassian
No need to write these two up separately. Both have been pretty much possession disasters every season until the last (significantly negative dCorsi Impacts). Both improved hugely to only slightly negative CorsiRel players this year. Both produce points at a surprisingly high level when adjusted to P/60min. Most importantly, both actually produced as positive dCorsi players this last season. Both were slightly below expectation for dCorsi Against but produced above expectation for dCorsi For (below is Bonino’s dCorsi visualization). This may suggest that given sheltered usage they could form the basis for an effective secondary scoring line. Each of them took a big step forward last season and there is reason to be (relatively) high on both going into next year.
Last year Richardson posted a CorsiRel of -4.9% (really really bad). I’ve seen him referred to as an anchor more than once around the blogosphere this summer, and that probably isn’t an unfair commentary. However, his Expected Corsi is lower than any Canuck forward not named Sestito. While he did underperform his Expected Corsi (dCorsi of -0.3) he was an absolute pack horse for Tortorella. Historically Richardson has only been about an average defender, and his good results have typically come from dCorsi For production. Given his exceptional results in LA the three years before last, I don’t think it’s unreasonable to hope for a bounceback year as a legitimate bottom-six forward next season. However, it may be that he isn’t well suited to carrying the toughest defensive assignments. He will be helped by the acquisition of Dorsett, who thrived in an even tougher role than Richardson last year.
I have a couple thoughts on Henrik and Daniel. One is that it’s important not to stress too much about Henrik’s season last year. He and his brother were clearly overplayed early in the year, and Henrik played with a seriously injured wrist later in the season. I’ve included Henrik’s dCorsi visualization below (visualizations accessed at this link) to make a point about Henrik, Daniel and Burrows’ dCorsi from 2010-2011 to 2011-2012.
All three of them posted negative dCorsi totals in both of those years. This was a period over which Daniel won the Art Ross and Ted Lindsay awards, their average CorsiRel was 6.18% and none of them ever posted a CF% below 54%. None of this necessarily precludes a negative dCorsi, but I do wonder if Vigneault’s usage during the period skewed their expected corsi disproportionately. It was the trio’s dCorsi For totals that were lagging behind their Expected Corsi in those two years, despite producing counting offense and overall Corsi at an elite level. In 2012-2013, when the Sedins were used in a more expansive role, all three of them returned to positive dCorsi totals.
If there is one Canuck forward who I am more concerned about after looking through the dCorsi results than I was before, it’s Higgins. His last two years he has tanked: I thought last year was more of a bounceback year than it really was. His dCorsi For has held – actually producing above Expected Corsi For last year – so it’s really his defence that has suffered. He produced well through counting stats last year, too, with 17 goals and 22 assists. His low dCorsi Against could be something systemic that hasn’t worked well with his game, but he is signed for three more seasons and so one of Desjardins priorities should be working on his defense.
Last year’s dCorsi totals compared to average dCorsi Impact totals suggest that there are several Canuck forwards who have a good chance at improving with a coaching change. Henrik Sedin and Jannik Hansen should have bounceback possession years. Overall, the dCorsi results support what everyone knows about the Canucks: there’s a pretty good first line, a group of decent bottom six players, and some big question marks on the second line. Burrows is a very good top-six player, and Higgins, Bonino, and Kassian all show signs of possibly being able to at least provide top-six offence fill that role from the dCorsi data.
If I were advising a coach from this data, one tactical option I would want to look at would be putting together a really strong third line to handle tough defensive assignments. Dorsett, Richardson and Burrows could be good choices (Richardson could be pushed down to the fourth line if Horvat makes the club this year). I would bury that line with tough assignments as much as possible so as to allow for a second line much better equipped to produce offence than defend effectively: some combination of Bonino, Kassian, Higgins, and Vey. Those first three players all have driven offensive production well in the past, and given complimentary usage could potentially produce at a legitimate second-line rate. I’ve included Vey because he’s produced points at a high rate in the AHL and would probably be benefitted by easy minutes in his first full year in the league, but if he looks capable he could also take Richardson’s spot on the third line. Sheltering the second line means that the Sedins would likely be exposed to more difficult usage along with that third line, but it would play to the strengths of the roster the Canucks have better than a traditional line 1/line 2/line 3 setup.
Of course that isn’t the only option the Canucks have, or even necessarily the best one. I’ve previously written about Alex Burrows driving possession for the second line. One of the greatest benefits of these dCorsi visualizations is being able to better break down Corsi into individual offensive and defensive contributions, and I think that the conclusions we can draw from that data could have really useful implications in usage strategies and tactics.
Coming up next will be a similar post on the Canucks defencemen, and then I’m thinking of writing something expanding on how the Canucks can structure their lineup to get the most production out of the forward pieces they have. That may wait until we have @ShutdownLine‘s neutral zone data from last season.