Expected Shots-For and Shots-Against at Different Zone Starts

Here's something I've been farting around with a little bit…there's a good number of articles on Zone Starts and Corsi and Balanced Corsi (which takes into account Zone Starts).  Eric T. over at Broad Street Hockey's done some excellent things with it, worth looking at here.  JLikens has also done a similar exercise at his excellent site, Objective NHL, as has Matt Fenwick at The Battle of Alberta.  In all these cases, the data creates an abstract number that is a bit disconnected from how a large portion of the NHL audience views the results of a game.  I like to look at things a little differently, less in terms of the Corsi Number and more in terms of shots-for and shots-against per 60 minutes, because it fits my intuitive mind a bit better.  When I see a Corsi Number, or the difference of shots-for and blocked shots-against minus shots-against and blocked shots-for, it doesn't click quite the same as if I saw that, when a player's performance was stretched over a total game, the team would have 30 shots to the opposing team's 25. 

What we know for sure (as Eric T. was bearing out with the above post) is that a player’s Zone Start% has a pretty good correlation with their Corsi Number at 5v5, with a sizable chunk of the variance due to player skill and luck. It also has a good correlation with shots-for/against differential at 5v5 (0.44), in fact very close to the “best” indicator used by a lot of the stats analysts, the Fenwick number (0.45, though really they use the away score-tied Fenwick nowadays, which could be higher). It helps to understand just how steep of a hill our guys like Jim Slater (26% offensive zone start percentage in October) have to climb versus Evander Kane (71.2%)(be sure to follow that link for the rest of the Jets), and so I aim to give us a sense of the league averages at key Zone Start markers culled from the last four NHL seasons (2007-08 through 2010-11). I only included players that had played at least 20 games, which provided me with 2,689 player performance to derive these numbers.

Be sure to have a look at Eric T.'s comment below for some nifty Corsi articles from the statistical blog Driving Play.

All data was taken from behindthenet.ca, Gabe's great resource if you're looking to do similar work.

The data gave me these predictive equations for 5-on-5 play (make sure Offensive Zone Start is expressed XX.X and not 0.XXX):

Shots-for/60 = (Offensive Zone Start Percentage * 0.1668) + 18.07

Shots-against/60 = (Offensive Zone Start Percentage * -0.137) + 33.643

Or, if you'd prefer to eyeball it:

One thing to keep in mind: I'm not adjusting for the quality of competition or teammates, though I'm not so sure that it detracts from what I'm doing too much.