Title: Are The Sedins Really That Sheltered?
Date: June 26, 2013
Original Source: The On Deck Circle
Synopsis: Originally crafted for Canucks Army, this piece toyed out loud with an xCorsi measure to predict NHL player possession performance. They didn’t want to run it until it was more fine tuned, so I threw it up on The ODC until the kinks are worked out.
Based on some previous work by Michael Parkatti over at Boys on the Bus, I’ve been trying to improve on his initial try at an Expected Corsi of sorts.
Where I’m at right now, and there’s more coming on this next week when I’ve gone through more iterations, is that Corsi is a stronger predictor of Corsi than any combination of factors. That is, if you want to know a player’s Corsi for next year, the best place to look is their Corsi from this year, especially if they haven’t changed teams, and even more so if said player is a forward.
Of course, even Corsi doesn’t even explain 40% of thevariation in the next year’s Corsi, and I’m confident if I (and we as a stat commuity) keep working, we’ll find a combination of inputs that tell us a little more.
As it stands right now, though, I have a formula for fowards that I’m calling Sit (as shorthand for “situation”) that does slightly better than Michael’s in terms of correlation with Corsi. My original goal was to create a catch-all metric for “situation” that would allow us to reference one stat rather than Quality of Competition, Quality of Teammates and Zone Start Percentage. I’m not there yet, as I mentioned, but I did create an xCorsi for forwards that was a .6618 R-squared (explains about two thirds of the variance in Corsi, up slightly from Michael’s mark of 61%).
So for today, since all of the Excel sheets and math aren’t where I want them to be, I thought I’d use what I have right now to examine Henrik and Daniel Sedin. The Sedins, of course, are very good hockey players, but are sometimes smacked with an “easy deployment” label and some of their success is thus discredited.
By the way, it turns out QoC doesn’t matter a whole lot. It correlates really low with Corsi for forwards and does almost nothing to change the R-squared of an xCorsi. The SIT metric referenced below leaves QoC out, dropping the R-squared from .66 to .65.
So using my very raw SIT metric, I took a look at where the Sedins rank among the 2362 qualified player season (from 2007-08 to 2012-13, for forwards playing in at least 30 games that year).
The highest a Sedin ranked in all of those seasons was actually 22nd, Daniel Sedin’s 2011-12 season. The list is dominated by Detroit Red Wings from the 2007-08 and 2008-09 years, mostly because their top lines were just so stacked, raising their Quality of Teammates. For the Canucks, Henrik Sedin and Alex Burrows from 2011-12 also crack the top-50 individual seasons, but those are the only appearances of the Canucks.
So, the Sedins have had some favorable years, but it’s not like they’re constantly the most coddled forwards.
Next I wanted to see if they outperform their expected Corsi based on their situation. Corsi Gain, the final column, shows their actual Corsi minus their xCorsi, so a positive means they outperformed and a negative means they underperformed.
So, there’s a mixed bag there. The positive thing is that the Sedins did really well this past season, outperforming their situation better than they ever have. Perhaps Vigneault was on to something using them in more difficult situations (less O-Zone Starts). 2010-11 looks like a bad year, which I suppose makes sense given that it becomes increasingly difficult to outperform your expectations when your situation involves three quarters of your starts starting in the offensive zone.
Again, this is all really early, and I’m just trying to make marginal improvements on Michael’s excellent first crack. I’ll try and roll out more details and iterations in the near future, but from this first run it at least appears that the Sedins can handle a tougher role. We’ll see if rumored new coach John Tortorella agrees.