A (Minor) Re-Calculation of Corsi: Adjusting For Zone Starts
I get that there’s a lot to be said for leaving statistics as raw as we possibly can, to make it more accessible and relatable to the actual watching the game. There’s also a lot to be said about making our stats better, and pushing the frontiers of what analytics can tell us. So hear me out! I have a better way of calculating corsi.
We put a player’s shifts into 3 buckets: shifts that started with an offensive zone faceoff, a neutral zone faceoff and a defensive zone faceoff. Then the offensive and defensive zone corsi for percentages are weighted equally, with the percentage of neutral zone faceoffs (and therefore the weight of it’s statistic) remaining the same. Let’s use Dion Phaneuf as an example.
In the first pie graph, Phaneuf’s abysmal corsi of 30.5 in defensive zone starts outweighs his corsi in offensive zone starts by 15%. Once adjusted, the defensive zone started shifts have the same amount of weight as offensive zone starts in the calculation.
For most players, the effect is marginal.
On the very extremes we see a couple of percentage points in difference between this corsi and the normal way of calculating it, but the overwhelming majority of players have less then a percentage of difference in their numbers. I think this gels well with other recent research on how much zone starts affect a players numbers.
There have been a lot of great attempts at adjusting for zone starts in the past (Driving Play, Vic Ferrari, NHL Numbers) but they all were approximations of the effect based on league averages. With these numbers, we don’t adjust a players corsi by how OZ% effects a players corsi on average, we adjust for how it affects that player. All of these numbers are still the players, we’re just adjusting the weight distribution.
These numbers are now available at Hockey Prospectus’ own stats page. Team relative stats are also adjusted on the off ice part of the equation as well. You can still find unadjusted CF%, FF% and GF% in the back.
I decided against actually listing a statistic for the percent zone start effect, but if anyone would like to see it or see me write more about it I’d be happy to.