Beyond the Fenwick: the Limitations of Current Hockey Analytics Pt II by @DXTraeger - PensInitiative | Pittsburgh Penguins Blog | Rumors | News

The Latest

Post Top Ad

Thursday, November 3, 2016

Beyond the Fenwick: the Limitations of Current Hockey Analytics Pt II by @DXTraeger

Phil Kessel is a Stanley Cup Champion

No, no, not the "Witches of" film starring Jack Nicholson, Susan Sarandon, Michelle Pfeiffer, Cher— hey, that's one hell of a cast— I'm talking about the hockey stat that bears more than a casual resemblance to its older, more popular brother, Corsi.

As you may recall, I explained a bit about the Corsi stat and what it statistically represents to coaches and analysts, and Fenwick tracks most of what Corsi does, albeit with one major difference.

Corsi FOR/AGAINST for a player is the sum of all shot attempts his team tried, including shots on goal, shots that were blocked by anyone on the ice, shots that missed the goal, shots that traveled through a Donnie Darko space-time continuum and subsequently crashed through a roof in California, etc. etc. while said hockey player was on the ice.

Name: "Basil Exposition"  Bad joke, true story.

Fenwick FOR/AGAINST simply subtracts any shots that were blocked.

Sounds simple enough, right?  So why the distinction?  In Austin Powers parlance, "Woopty do, but what does it all mean, Basil?"

Again, the point of analytics and event forecasting comes down to specificity and correlation.  While high Corsi FOR (henceforth CF in the name of brevity) carries a statistically significant correlation with possession and thus winning (operating under the generally true axiom that the opponent cannot score if he does not have the puck), Corsi reveals nothing about in-game context, and a shot from center ice carries the same weight and value, statistically, as a one-timer from the slot.

Fenwick attempts to peel back some of the nebulous void around Corsi by further identifying which players are able to get the most of their shot attempts through the defense.  While Fenwick is limited in that it doesn't discriminate between shot locations (basically, a legitimate scoring chance or blind shot in the general direction of the goal), Fenwick does begin to help discriminate between the Nerf machine guns on the ice (see: Kennedy, Tyler) and bona fide snipers like Steven Stamkos.

By the Way, My Photoshop Skills Are TOTALLY For Hire
The end result is an improvement on Corsi in regards to predicting future results: a player that generates a higher Fenwick FOR (FF) is thus more likely to get more shots toward the goal, thus making him more likely to put the 'ol proverbial biscuit in the basket.

Of course, FF does not account for shots that should have been blocked, or shots that skipped on the ice, or shots that were not knocked down because of a teammate's successful screen, so once again the value of Fenwick is quantitative, measuring quantity over quality.

As such, while Fenwick is an improvement on Corsi (which I regard as a glorified improvement to the old stand-guard of plus/minus), FF and FA do not provide analysts with a full interpretation of how a player, line, or team played on any given night, and definitely provides no insight into whether or not the shots being measured were of a scoring chance variety.

I'll delve into my personal preference- subjective analytics- in my next article, and I will break down the Penguins' performance over the first 10 games of the season and attempt to augment the HOW and WHY's of Pittsburgh's early success.

No comments:

Post a Comment

Post Top Ad