Yesterday, we took a look at what three seasons of CSSI analysis has taught us about the Red Wings' offensive output and came to the conclusion that it gave decent data on how goals are created, but not much else. Today we'll take a look at the other side of the CSSI system.
When I started doing CSSI for the 2010-11 season, I did so with a general unease for some traditional NHL stats and a worry about the some shortcomings of the newer possession stats which have grown both more complex in their analytical uses and more accepted in casual hockey conversations since then. I knew that such stats were useful if used correctly, but I wondered if one could create an even better system through the best-possible method of hockey analysis: watching the game and reading the scoresheet.
After more than 250 games worth of using such a system on the Red Wings, I believe that what I've done is to create a good system which adds additional clarity to the current method, but not to an extreme degree. For the most part, CSSI has created an interesting index which coincides nicely with the expanded statistical analysis popular today, but still has many of the same shortfalls.
One of the hardest things to study via statistics in the NHL is just how defensively good a player has been. For officially-tracked stats, there's merely a collection of hints which may or may not actually tell you anything useful. Plus/Minus is a long-derided stat which counts up the differences between goals for and allowed with a player on the ice and can mean any number of things. Takeaways credit a player with such random variance between arenas that even if it were the a more useful stat on its face, it couldn't be trusted very well. Time on ice while shorthanded tells us which players a coach trusts with defensive responsibility, but only really hints at whether he does a good job overall.
The extended stats universe does a much better job of the concept, but still essentially defines defense as the kind of thing which has happened when there's a notable lack of offense. To the NHL, defense is not a light; it is a darkness. By that very nature, this makes it harder to study.
The problem is that there's not a very good way to say what exactly "defense" is. Taking the puck away from your opposition is certainly defensive, but is it more or less defensive than blocking a shot? Creating more shooting attempts at your opponent's end of the ice is an indication that strong defense is being played, but doesn't quite explain how. The analytic community is working very hard on answering this. I feel that an additional useful set of stats will aid in this. That's basically what the CSSI has been trying to create.
In looking at three years' worth of CSSI Plus/Minus accounting, I can say that it hasn't simply inflated stats to eliminate some gray areas like the points system did. I can say though that numbers have been inflated and that it certainly seems as though even the defensive side of this extra analysis system was heavily influenced on how much offense a person could create as a factor of his defensive prowess.
The following chart breaks down all the players scored over the last three regular seasons, organized by their total adjusted plus/minus scores.
If you're like me, you didn't bother looking at the sea of grey at the bottom of this chart before realizing the problem at the top. Nicklas Lidstrom did less defensively than four players on the Red Wings, one of whom played fewer games than he did during that time period and all of whom had a better per-game adjustment rate?
I don't think so.
The immediate obvious issue is that these aren't all necessarily defensively equal considerations and some aren't even directly defensive. While the purpose of a goal-scored plus was to award a player for an inherently defensive play to either gain or maintain possession in a way that was defensively sound, the pure fact of the matter is that this is something forwards just do more often than defenders. Lidstrom had fewer GS+ events than Justin Abdelkader, despite having nearly twice as many points scored. This is just the nature of the beast.
The other consideration is the minus columns for mistakes that led directly to goals. You'll notice that not a single forward had as many coverage minuses which led to goals as Lidstrom had. You'll also notice that centers have more of them than wingers. Again, this is a consideration of the job. Put simply, a shot attempt that was stopped and turned into transition the other way wouldn't cost a player, but a failure to recognize the correct coverage on an oncoming rush would. Since those things just so happen to take place at differing intervals for forwards and defenseman, the blueliners would get punished more often as a factor of their own responsibilities.
Fortunately, if we do separate by position, we actually do get a clearer picture of overall defensive responsibility that does seem to play along with the eye test very well.
This separation still has a few contextual oddities (Ian White benefited greatly from having Nicklas Lidstrom as a playing partner, for one), but it does a better job of describing the players we would indicate as both defensively sound and good drivers of puck possession. Part of what gets corrected in here is that, while centers tend to grab more blame for defensive breakdowns, they also earn more credit for defensive transition.
You can also pretty clearly see why the fanbase has been so hard on Kyle Quincey and Jonathan Ericsson. While I've personally defended Ericsson thinking that he was closer to not being a horrible screwup, the numbers bear out that he had two very bad seasons that couldn't be overcome by being the highest-ranked defenseman on the Wings for adjusted plus/minus among guys with more than 15 games played. While Ericsson's overall defensive miscue goofs (turnover and coverage minuses) are lower than Kronwall's, he wasn't close to as good overall and took far more penalties.
The overall adjustments fit fairly well with the people who were simply on the ice more and for whom more events happened. Datsyuk and Zetterberg had more coverage and turnover minuses by the nature of them playing center against the very toughest competition. I do like how penalty killers Miller and Helm show up decently high in the rankings, but would be interested in finding a way to better credit penalty killing defensemen for more of the good they do to counteract their increased likelihood to get a minus on a goal.
I plan to do a bit more analysis on these numbers and a bit more soul-searching about what's going to happen with CSSI going forward. I think the plus/minus considerations are worth continuing to some degree, but may need to be pared down and/or separated to give us a more useful study tool in the future.