Why all shot attempts aren't created equal

No, this isn't a treatise on shot quality. This is a post on the difference between a shot, and missed shot, and shot attempt for that is blocked.

When we look at Corsi, we treat all shot attempts as equal. A shot on goal is the same as a shot that goes 3 feet over the goalie's head. Now in terms of measuring possession (and Corsi/Fenwick after all are proxies for possession), this is an ok assumption. If a team gets 2 shots on goal and 1 missed shot in a shift, or 2 missed shots and 1 shot on goal in a shift, the line probably had equal time of possession on that shift.

However, I would argue that those two above possessions aren't equal. I think most of us would prefer the possession with 2 shots on goal over 1 shot on goal.

So the question is, can teams control the proportion of corsi events for or against that end up as shots for or against. The answer . . .


For team data, I collected all team data from even strength 5v5 from each of the last 6 seasons. I then calculated both shots for per 60 minutes per corsi for per 60 minutes (SF60/CF60) and shots against per 60 minutes per corsi against per 60 minutes (SA60/CA60). This basically is a percentage of how many corsi events (for or against) were shots (for or against). Note that both shots for/against and corsi for/against are on ice stats, rather than solely individual stats.

I then ran the correlations for the 2007-2008, 2009-2010, and 2011-2012 seasons to , respectively, the 2008-2009, 2010-2011, and 2012-2013 seasons. Basically, I wanted to see if a team's percentage of corsi events (for or against) that were shots (for or against) was a repeatable skill year to year. In other words, does knowing one year's SF60/CF60 tell us anything about the following year's SF60/CF60.


First, the offense.

r 0.702324711

Wow, that's a pretty strong correlation we have there. And with a p-value of less than 0.01, it is statistically significant as well. With an R^2 of about .49, nearly 50% of the variance in SF60/CF60 is explained via the previous years SF60/CF60.

Now to defense

r 0.735710541

Again, another strong correlation. And again, our p-value is less than 0.01.

It seems like a team's percentage of corsi events, both for and against, that end up as shots is a repeatable talent year to year, both offensively and defensively.

So now the question is, does it matter. To figure this out, I calculated the standard deviation and average for SF60/CF60 for both offense and defense, as well as the average total Corsi for and Corsi against in a full season and the league wide shooting percentage for those same seasons.

If a team was able to generate a SF60/CF60 at 1 standard deviation above the league average on offense and 1 standard deviation below the league average on defense (which would be a good thing, since this would prevent shots), than our imaginary team would come out with a goal differential of about +11.6, which is worth almost two wins.

I think this is something to look into a bit closer. I did run some numbers for forwards, using only players who played at least 200 minutes in both the 2010-2011 and 2011-2012 seasons (sample size of about 350 players). I found no repeatably talent at the individual level for SF60/CF60. If this is true, we are seeing something that can only be repeated at the team level.

This item was written by a member of this community and is not necessarily endorsed by <em>Broad Street Hockey</em>.