Many times in evaluating players on this site, in discussions or arguments or whatever, I bring up the advanced metrics. The reaction to those seems to be "That's great" and they are then shrugged off, only thought of again when I continue to annoy you by bringing them up again and again.
So yesterday I was having a discussion with BiB on Kyle Quincey and brought them up again. I realize there's a bit of a history with this board and some members of the Corsiatti, which has caused quite a bit of backlash against the advanced stats.
Since it's the offseason, and the feud has simmered since last season, I thought this might be a good time to bring up the discussion again.
One place where I feel the Corsiatti have done a completely poor job, especially on this board, is in explaining how they formed the opinions on Advanced stats that they have. I think part of what made those discussions so acrimonious was that there was an attitude of "We've looked at the data and you should just accept our conclusions". Many of these same people took a long time, and many discussions to change and form their opinions into what they have become. To expect someone else to just accept it, when they themselves would never just accept it, was a bit hypocritical. I don't think it helped the discussion at all.
On the other hand, there are a lot of incredibly smart people using their free time to do some incredibly smart detailed and fascinating work. To write it all off as "Well it still has it's problems" is a bit like throwing the baby out with the bath water in my opinion. I recognize there are limitations to the data out there, but I think it's good to understand exactly what those limitations are. There's never going to be a 100% certain statistic, but there are some stats that do a good job describing stuff.
What I wanted to do was present some of the research and blog posts that some various members have done that have influenced the way I have thought about these stats. These posts helped me form my own opinions on the advanced stats, rather than just relying on some math nerd's interpretation.
Why Corsi Matters to me :
The first post I've linked to is one of the more influential one's, for me at least, to understand why everyone likes Corsi so much. I don't think it's going out on a limb to say that better teams will have the puck in the other teams zone more than it will be in their own zone. But when you count shots (like Corsi does) there's always the cocern that you're missing something. Are some teams more prone to pass around looking for the perfect shot, thereby causing a discrepancy in their Corsi and their actual ability?
To me, the data presented in this post shows a very strong correlation for possession and Corsi. While there are some outliers here (Colorado being one of them) it's pretty clear to me that time of possession (and more importantly Time of possession in the offensive zone) is pretty accurately represented by Corsi (& also Fenwick).
Also, JLikens did some work on the correlation between Corsi, Fenwick, Shot Ratio and goals/winning, which, I think, backs up my conclusions above.
A great post for this season since there has been so much talk about defensemen this year, and they are incredibly hard to judge. (Think of all the differeing opinions on different defensemen over the year just on the Avs). I think Vic is correct in his conclusions here that defensemen can affect shot quality, but the affect is so tiny that it's nearly immeasurable, which means that the bigger affect (and focus) of a defenseman should be to keep the other player from shooting altogether.
This, to me, means that when a player has a terrible on-ice save percentage he isn't doing anything wrong other than being a victim of small sample size and bad luck.
Stuff like this is what leads me to believe that Kyle Quincey and Kyle Cumiskey, two guys raked over the coals here, didn't have nearly as bad a season as people want to believe. (There's more to it than this, but it's one of the stats I look at when judging a player things I look at). (I think Kyle Cumiskey was bad for different reasons, but he also had a healthy dose of bad luck to go with it too this season).
It's also interesting to watch people's opinions of a player change as his on-ice save% changes. When Hunwick first came over, his on-ice save % was REALLY bad (and I don't have a link to it, I just remember) and he was getting raked over the coals here. (I thought it was justified because his Corsi was low and his OZone% was high). Then his on-ice save % changed dramatically, but his Corsi and OZone% numbers stayed the same.
Everyone started saying "He played better" and "He's really playing better". But if you interpret Vic's data the same way, that the defender really doesn't affect save% all that much, then he didn't start playing all that much better. In fact he was playing almost the same, he went from being very unlucky, to very lucky and people perceived that as a dramatic increase in play.
(By the way, this is what I am predicting happens to Quincey next season. He starts getting some breaks and getting a bit luckier and everyone will say stuff like Johnson helped him and he's played with more Fire, and more tenacity. I wonder if the numbers will be all that different though).
This post was simply fascinating to me, mainly because it was one of my first into the field of advanced stats. Before reading this post I was a big believer in shot quality. But this post really made me think (and later conclude) that shot quality, on a large enough sample size, is a much less important factor than I thought. It makes a difference, but the difference over the course of a season is very small.
Now I disagree with Vic's conclusions (in the comments). I think Vic should have looked more at the standard deviation, because that's what would change more rather than the average. However, I think some of my comments were, in hindsight, an error (or an exaggeration). So while I think Vic's interpretation was incorrect, I think it was slightly incorrect. The chances that the Standard deviation of the two shows radically different results than what he displayed is pretty much nil. I'd like to see it, but I don't think it will change anything.
Shot quality, in essences, is less of a factor (over a significant sample size) than randomness (or luck). in my mind it's more likely that the stuff that affects shot quality is also heavily governed by luck. A defenseman slipping at the blue line (or the puck bouncing over his stick) resulting in a breakaway. A bad clear that goes into the stands causing a PK.
Kent Wilson also has a good article up today that gives more insight, and links to work, that help back many of the conclusions I cam to above. The one part that was most telling is that NHL GMs (and the good coaches) know that puck possession is key, and that's where Corsi comes in.