Breaking Down Corsi: Looking Into Shot Blocks and Misses

The willingness to block shots is one that’s often touted by the media as being the difference between winning and losing. Players who put their bodies on the line are held in the highest regard and with good reason: Fenwick, which excludes blocked shots tends to be a better predictor of future victories than Corsi, which includes them. The key question around blocked shots, as with anything that correlates well with winning, is whether they’re actually repeatable. We know that blocked shots are valuable, but we want to know whether they tend to happen at random or with some predictability.

A few years ago Sunny Mehta looked into shot blocks and misses over the 2008-2009 season and found that there seemed to be an element of skill in blocking shots, while both shot attempts for blocked, and shot attempts for and against missed seemed to be mostly luck. I wanted to update his study, looking at data from the 2008/09 through to the lockout shortened 2013 season and examining to what degree teams had control over blocked and missed shots. I also wanted to take a look at how each metric broke down between forwards and defenseman, to see whether rates varied the groups. Before we start looking at repeatability though let’s go through some high-level numbers.

Goals Shots Corsi MS BS Sh% FSh% BS% MS%*
F 20531 213655 368133 75971 78507 9.6% 7.1% 21.3% 26.2%
D 3087 72347 161722 34534 54841 4.3% 2.9% 33.9% 32.3%
All 23620 286009 529865 110508 133348 8.2% 5.9% 25.2% 27.8%

*(Fenwick-Shots)/Fenwick

Forwards score on roughly 5.5% more of their shots than defensemen, and roughly 4.2% more of their Fenwick attempts. This is an important fact to keep in mind-blocking a forward’s shot is much more valuable than blocking a defenseman’s. With that being said, defensemen tend to have more of their shots blocked, and tend to miss the net more often on the shots that don’t get blocked. This, of course, makes sense-forwards tend to take shots from closer in, leaving less of a chance to get a leg in the way, or to put it wide or high.

We can also look at the numbers another way, breaking down what percentage of each event comes from shot attempts taken by either forwards or defensemen. While forwards take 74.7% of the shots, they only record 68.7% of the missed shots and 58.9% of the blocked shots, while accounting for 86.9% of the goals scored.

Goals Shots Corsi Missed Shots Blocked Shots
F 86.9% 74.7% 69.5% 68.7% 58.9%
D 13.1% 25.3% 30.5% 31.3% 41.1%

With all that in mind, let’s get down to the more detailed stats: first off, let’s look at whose shots are being blocked and who’s doing the blocking.

Shooter Blocked By % of Shooter % Total
F F 21.3% 12.5%
F D 78.7% 46.3%
D F 56.9% 23.4%
D D 43.1% 17.7%
All F 35.9%
All D 64.0%

The results we see aren’t exactly surprising: the majority of the time when a forward takes a shot and it’s blocked it’s a defenseman doing the blocking (78.7% of the time). Similarly, when a defenseman’s shot is blocked, it’s more likely to be blocked by a forward (56.9%) than an opposing defenseman (43.1%). In total, defenseman are doing most of the blocking, accounting for 64.0% of the total shot blocks, with the majority of these blocks coming on shots taken by forwards.

Now let’s look at the repeatability of blocked and missed shots. What we want to get a sense of is whether the ability to block shot attempts or to force a shooter to miss the net is more ability or luck. Similarly, we want to know whether certain teams or players are better at hitting the net than others, and if this is a persistent ability, or random variation. To look into these, I’ve run split half correlations at the team level for each of our four variables of interest (SAF Blocked %, SAF Missed %, SAA Blocked % and SAA Missed %). Split half correlations simply look at to what degree the numbers a team puts up in the odd numbered games predict the numbers for that team in the even numbered games. A higher correlation implies that a particular number is more skill than variance, while a lower number implies the exact opposite. In each table I’ve looked at the overall correlation as well as the correlation for shot attempts taken by forwards and defenseman.

Correlation 2008-2013 Correlation 2008-2012
SAA Blocked (All) 0.71 0.72
SAA Blocked (F) 0.62 0.66
SAA Blocked (D) 0.64 0.61

We’ll start by looking at SAA Blocked %, the percentage of total shot attempts against that a team manages to block. Our results here show a fairly strong correlation, in agreement with what Sunny found around the skill involved in blocking shots. We also see that a team’s ability to block shots from opposing forwards appears to be slightly more persistent than their ability to block their shots from defensemen (note that when we include the lockout shortened 2013 season these numbers flip, but I’m more inclined to leave it out of split-half comparisons due to sample size constraints).

What about at the other end of the ice though? To what degree can teams control how many of their shot attempts for are blocked?

Correlation 2008-2013 Correlation 2008-2012
SAF Blocked (All) 0.57 0.59
SAF Blocked (F) 0.55 0.60
SAF Blocked (D) 0.34 0.34

Once again, we seem to see a fair amount of repeatability in the number, although it’s definitely less than for SAA Blocked. Our data here does seem to differ from Sunny’s conclusions, although the breakdown between forwards and defensemen is much more important in this case. While the split-half correlation is reasonably high for forwards SAF blocked, for defensemen it’s significantly lower.

What this means is that forwards do seem to have some influence over whether or not their shots are blocked, while for defensemen it’s much more random. At a glance, I think these numbers make a lot of sense intuitively: a defenseman is often taking shots from further out than a forward, and while they can obviously avoid firing the puck directly into a forwards shinpads, they don’t have a lot of control over whether a shot hits an opposing defenseman nearer to the net. On the other hand a forward will generally only have to get a shot by one or zero players, which leads to a lot less luck influencing their results.

Correlation 2008-2013 Correlation 2008-2012
SAA Missed (All) 0.46 0.46
SAA Missed (F) 0.39 0.40
SAA Missed (D) 0.26 0.22
SAF Missed (All) 0.48 0.49
SAF Missed (F) 0.38 0.40
SAF Missed (D) 0.31 0.31

Looking at missed shots we see that the repeatability of both SAA Missed and SAF Missed seem to be significantly lower. While the overall and forward missed shots percentages show a higher correlation than defenseman SAF blocked %, the numbers aren’t that encouraging. Knowing the missed shot percentages in the odd-numbered games helps us explain less than 25% of the variance in the even-numbered games. While it may make intuitive sense that some players would be better at hitting the net than others, the numbers don’t really back up that conclusion. Forwards may be able to avoid the maze of legs and bodies in front of them, but once the puck gets by that maze whether it hits the net or not is up to lady luck to decide.

One important thing to note with this analysis is that we’ve only looked at team level data, and only looked at numbers within seasons. What that means is that we can’t really make any conclusions about individual player’s ability to block shots or avoid blocks themselves. Without examining the individual level data, we can’t know whether the repeatability is due to the system a coach has put in place, or some innate ability of the players themselves. We can speculate of course: if I had to bet I’d lean towards blocking shots being a product of a team’s system with there being more of an individual focus on avoiding shot blocks. But without the data to back up the theories that’s all they are-until we can dig down into the details, we just can’t reach any conclusions.

Advertisements
Tagged with: , ,
Posted in Statistics

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: