WASHINGTON, DC - FEBRUARY 13: Patrick Marleau #12 of the San Jose Sharks collides with Marcus Johansson #90 of the Washington Capitals at the Verizon Center on February 13, 2012 in Washington, DC. (Photo by Greg Fiume/Getty Images)
Last week we looked at T Mac's deployment and noted a few interesting things.
- T Mac likes to play his top players against top competition. (More on this later)
- Pavelski is having a killer year, both offensively and defensively. By all measures he is the most defensively responsible forward this year for the sharks. Joe Thornton is also playing extremely well at both ends of the ice.
- Brent Burns has been a tad sheltered in the offensive zone. I think this is starting to trend away as he is learning the system, and has been better defensively lately.
- Despite decent numbers A. Murray was sent down, probably to develop his defensive game.
- T Mac has still trusted D. Murray with tough minutes and zone starts despite a rough start to the year.
Now to give those numbers some context we're going to turn to league wide deployment. My hypothesis is that coaches tend to lean either toward matching competition (CORSI rel QOC) or matching zone (zone%). That is to say coaches decide they will role out specific lines either when certain competition is on the ice, or when the next faceoff is in a particular zone. There is probably significant blurring of those 2 camps, but let's see what this year's data suggests.
**Super nerding for the next few paragraphs, skip to the bottom for team discussions. A couple notes on methods. The data I used was limited to players with at least 20GP, and EV only. I generated standard deviations of zone% (y-axis) and CORSI rel QOC (x-axis) for each team. The Standard deviation of a range (in our case the zone% or CORSI rel QOC of each team) is a measure of how "far" the data points are from each other. A "tight" standard deviation means there isn't much variability in the data, ie. data points are close to the mean. A "wide" standard deviation suggests much higher variability, ie. data points are highly dispersed about the mean. In this case it tells us if each team is spreading out player deployment by either zone starts, competition (line matching), both, or neither. As I mentioned above, my hypothesis is that teams tend to choose one or the other. I'm also making a rather big assumption that zone% and CORSI rel QOC are not in themselves random. I think it seems intuitively true, and therefore unnecessary to analyze, but I leave that for you to ponder.
To determine if teams line match or zone match I decided to run a regression of the data. What we are looking for is a statistically significant negative slope which will answer our hypothesis outlined above. In essence we're interested in a negative correlation, which suggests that teams that opt to match by zone must give up the possibility of matching by lines, and vice versa. For forwards the regression equation is: y = -4.9676x + 10.16 p-value = 0.026, R² = 0.1644. For Defense, regression equation is: y = -0.8216x + 4.9123, p-value = 0.589, R² = 0.0106
Because p is < 0.05 for forwards, we can (reject the null hypothesis and) conclude that CORSI rel QOC and zone% are inversely related to each other. The evidence (p = 0.026) for forwards is much stronger than defense. It doesn't appear as though zone% and CORSI rel QOC are sufficiently different from each other for defensemen. Taking note of the qualifiers mentioned above. Unfortunately this leaves us with mixed results. While it seems clear that coaches try to match forwards by either line or zone, we can't tell with certainty that this is the case for defensemen. Clearly further analysis is needed, and I feel using home/away splits might tease out some more data. Trends do emerge in the data however, which are ripe for analysis.
Before analyzing the trees, it might be worthwhile to take a second and look at the forest. When I see these graphs I see outliers. Some of the teams adhere exclusively to zone matching or line matching, but most are jumbled around the mean for both measures. However, certain teams really jump off the page, setting an interesting scene in the NHL. I can't currently argue one system is better than the other, but I do find it interesting that such variability exists.
The Forward chart coincides with my expectations. Manny Malhotra skews the whole forward graph significantly north. He starts 87.3% of his (either offensive or defensive face-off) shifts in the defensive zone. The next closest non-Van player is Jim Slater with over double of that at 28.7%. This is Vignot's hyper zone matching style of coaching. He basically only plays Malhotra in the defensive zone, and only plays the twins in the offensive zone, relatively regardless of whose matching up against them. Joel Quenneville's Blackhawks are also up there, but not even close to VAN. On the other end of the spectrum, CGY, DET, and SJ seem to rely much more on line matching. CGY much more so than any other team. This confirms an early idea of mine that T Mac likes to line-match, usually going with his top line vs. the opponent's top line.
Some surprises came up for me (maybe not for you) on the D graph. I didn't know COL and TB zone matched their defense so much. TB has a well graded zone start with Brewer and Hedman dealing with some brutal zone starts. While Hejda and O'Byrne are carrying the load for COL. CGY and DET again look to be line matching on the D side of things. Nicklas Lidstrom, no surprises, and Ian White (gasp) get by far the toughest assignments for DET. Interestingly the Sharks D aren't line matching this year. This of course could be due to the fact that the top 4 was switched about half-way through the year during Murray's injury. CHI is very interesting in that the seem to zone match their forwards, while line match their defense, suggesting that it's possible (and potentially advantageous) to do both.
Next we will try to break down production by each zone, to see if all this matters.