February 21 2012 01:27PM
Alain Vigneault's extreme situational deployment strategies have been a hot topic among hockey stats nerds this season. The topic even went mainstream this weekend when Malhotra's O-zone start rate was mentioned by Elliotte Friedman on Hockey Night in Canada during Saturday evening's game.
I'm of the opinion that zone-starts have a major impact on a skaters point production and possession stats. In my view, they also tell us a lot about which skaters a coach thinks are his team's best at either end of the ice. Most people wouldn't dispute any of the above, but David Johnson of hockeyanalysis.com isn't most people.
Read on past the break.
I'm confident that David wouldn't dispute the latter two points in the above sentence: quite clearly zone starts impact possession numbers, and tell you a fair bit about which players a coach trusts in certain situations. What David Johnson is skeptical about, is the impact zone-starts have on point production.
First some background, David Johnson is something of a heretic in the world of hockey analytics. He's one of the few hockey-math folks who believe that players can suppress shooting percentage, or put another way, he's a believer in the impact of shot-quality. In general, I find David's work extremely interesting, and I've found both his site, and him personally to be extremely helpful in the past.
In his post today, titled "Zone Starts: Why we Shouldn't Care" Johnson chose to use Daniel Sedin's zone-start rate and production as a test-case. First, David tallied up the number of points Daniel Sedin has this season coming directly off of "offensive zone-starts" and found that he had 7. Put another way, 11.5% of Daniel Sedin's total points on the season have come on the same shift as an offensive zone draw. Then Johnson used this bit of math to repudiate the idea that zone-starts have a dramatic, inflationary effect on point production:
For the sake of argument, let’s say we can directly tie all 7 of [Daniel's] points to being a result of offensive zone face-offs. Also, for the sake of easy math, let’s assume his OZone% is 70% (it’s actually closer to 80%). So, on 70% OZone starts he scored 7 goals. If we reduce his Ozone% to 50% you’d naturally think you’d lose an equivalent portion of points so he’d end up with 5 points instead of 7. Net result, Daniel Sedin’s offensive zone start bias has accounted for just 2 additional points so far this season.
This is convincing math, but it's not necessarily convincing logic.
The major issue here, as I see it, is that David has made the argument that "it doesn't matter much" all the while his math shows that it does in fact matter! Maybe the margins aren't mammoth, but by exploiting the percentages, Alain Vigneault is -according to David's own math - putting "Daniel Sedin on the ice for 3-4 extra goals per season." 3-4 extra goals may seem like a drop in the bucket, but over the course of a full season it's worth more than a point in the standings according to Pythagorean win expectation - and that's only one skater.
Further, David puts some arbitrary qualifications that limit the number of goals he counts (I'd also prefer he count more events than simply goals scored). Others, like Driving Play, have done similar inquiries in the past and have found a significantly larger effect.
Or as Cam Charron put it in the Province this weekend:
It's apparent that Vigneault is exploiting a mathematical edge that other NHL teams don't. He puts his best defensive players in the best place to succeed defensively, and puts his best offensive players in the best place to succeed on offense.
Even if the "mathematical edge" isn't massive (which, is a matter of some dispute), over an 82 game season exploiting even extremely slim margins can have a tangible benefit. Professional sports teams are always looking for that extra 2% for a reason, just ask Tampa Bay's baseball team.