Friday, February 06, 2015
Footnotes
I’ve been interviewed a number of times in the past six or
so months about the metric that bears my name, the history behind it, and
miscellaneous things about hockey analytics. And I feel like I haven’t done a
great job of explaining some things.
Until now, though, I’ve resisted the urge to write more
about it. I don’t want to be accused of, or even perceived to be, re-writing
history to flatter myself. It really is all there in the blog archives (at least the ones whose owners haven't taken them offline). But
I do feel like there’s a bit more to be said that might otherwise be forgotten
or misunderstood, so consider this post to be footnotes.
ONE. The short story of the invention of the Fenwick number
is:
Vic Ferrari invented the Corsi number. The first time he
showed his results was for the Oilers players at that time, in an Excel table
embedded as an image in his blog post on Irreverent Oiler Fans. His method counted all recorded shot
attempts from the NHL game sheets (goals, saves, goalposts, misses, and shots
blocked). I either initiated or pushed a discussion about whether it was right
to include blocks with those other four things.
Meanwhile, Vic was fiddling with things such that he could
show his work better, and that others could replicate it or try their own
thing. It was a bit of programming code that read the NHL gamesheets and
presented the results in table form on a webpage. It wasn’t the most accessible
interface for noobs but suffice it to say that timeonice.com is one of my top
10 favourite websites ever.
Anyway, we (mainly me and "RiversQ" (@ThomsonCam)) were having these discussions in a couple of comment threads on Vic's blog. One day, timeonice.com had an additional column: instead of deciding one way or another whether Corsi should include blocks, he just added a column called "Fenwick" that excluded blocks, and we pretty much moved on.
TWO. In Bob McKenzie's recent book Hockey Confidential, he had a nice interview with Vic Ferrari, now known to be a fellow named Tim Barnes. Tim said some very kind things about me, and for that I'm extremely grateful. I bring this up mainly to point out that my 2nd best serious answer to the question, "Why is it called the Fenwick number?", after the whole long explanation above, is "Because I was nice to people on the internet."
THREE. An old discussion that seems long forgotten but that had a significant impact as I recall things was Goalposts. NHL gamesheets track posts and crossbars; they're still categorized as a MISS, but it does note Post or Crossbar. Tim looked at these, using the same tools he used to assemble & present the timeonice pages, and concluded that it was all awfully random.
- No players consistently hit the post more or less frequently than average over time
- No goalies consistently allowed more or fewer posts behind them than average over time
- At least at a glance, it seemed like when a player or team hit an abnormally high number of posts for a stretch, they came at the expense of Goals. (Conversely, a goalie who heard a lot of TINGs behind him for a while seemed to have a higher SV% go hand-in-hand.)
This general relationship helped lead us towards a couple of concepts which are now very familiar:
- If posts come partly (or significantly!) at the expense of goals, then maybe the coach-y approach -- you can't score if you don't hit the net -- is not the right approach for the analyst who's trying to guess how things will go in the future.
- The hockey gods are hilarious and cruel and all that. Strange things, including those that might have an attractive subjective explanation, often regress.
FOUR. Corsi was invented to look at players, not at teams. Since that time (barely 7 years ago), the perspective on it and its application seems to have revolved almost 360 degrees around the sun. But at the beginning, it was about Tom Gilbert and Ales Hemsky and Daymond Langkow and David Moss being underappreciated. It was about exposing an ailing player before his injury became public (or acute). Soon enough, we started realizing the effects a player's usage can have on his Corsi, and we started looking at teams more closely (wow, Detroit is good at this) which led to the doomed Minnesota Wild, and the 2012 8th Place Stanley Cup champions, and Oh Those Leafs. And now we are onto the significant effect that coaching can have, and the idea that your Corsi is your Destiny has fallen out of favour somewhat.
I watch it all with great interest, and I'm looking forward to the next leap. My own view is that there are certain skills that make a player "good at Corsi" that overlap but are not identical to the skills that make him or her "good at hockey". The chief of these skills is being able to disrupt what the other team is trying to do with the puck, while not allowing them to disrupt what you're trying to do. I think of a younger Robyn Regehr always making a calm, quick first pass to exit his own zone, and David Moss knocking down pucks with his stick, hands, legs, skates, whatever. These players will be good for you no matter what system you coach them up into because being good at hockey is much more "real" than any coach or system. For the same reason, every megastar in the last 40 (70? 100?) years of the NHL has been famous by his 14th birthday, if not earlier.
FIVE. This isn't a confession exactly, but a clarification, or a reminder: I'm not an analytics guy. I'm lousy at the math and I have no coding capabilities. Once my spreadsheet gets past about column J, I can't use it effectively anymore.
It's not false modesty, because I do think I was a pretty good blogger. I did try to explore objective realities, and to use logic, and avoid doing the opposite. But I'm not an analytics guy and really I never was. Thank you to everyone who has told me that they enjoyed the blog back in the day, and that they miss it. Go Flames.
# posted by Matt @ 2:51 pm 
