Daily Hab-it: Advanced statistics 101
Published Wednesday, November 3, 2010 11:47PM EDT
In recent years, as the NHL has begun providing detailed play by play reports and shift charts for every game, a relatively new way of seeing the game through numbers has emerged.
Advanced statistics, as they're most commonly called, has been generally accepted by the blogosphere as a more sophisticated way to judge a player's impact on the game, or a way to explain why a player is struggling.
I find them quite fascinating, because I'm a bit of a hockey dork, but I thought I would start a semi-regular look at the Canadiens through the use of advanced statistics to try and shed further light on the team's performance for my readers. Because I know there are many other hockey dorks out there among you.
So this is the first one, but before I get into it we should start with a little primer, some disclaimers and a big acknowledgment.
The source for all my statistics will be behindthenet.ca, a site started by Gabriel Desjardins several years ago that has become the authoritative source of this kind of data. He came up with the idea of using all the new information being provided by the league and compiling it to come up with a brand new set of statistics that bring us deeper into the game than simple goals, assists, and especially basic plus-minus ratings ever can.
The statistics are not without fault. There is some human error involved in the gathering of the information used as a basis for the numbers, and sometimes we do see some glitches in the actual stats, but overall I feel they provide a good deal of insight and allow us to debunk certain myths borne of perception more than reality.
I've decided to focus on only four out of the multitude of new statistical categories Desjardins' work has provided us, but I invite you to explore his website a little further. I think you'll be interested and surprised at some of the stuff you can find.
I hope you're not prone to headaches
So the four I've decided to focus on, for now, are quality of competition, zone starts, relative Corsi ratings and penalties drawn and taken.
Here is where Desjardins himself explains his stats, but I'm going to give it a shot.
Before I get into an explanation of each term (though a couple are pretty obvious), I want to explain a basic principle guiding some of this data.
Almost every statistic tracked at behindthenet.ca looks at a player's performance in two ways. First, it breaks it down by minute and then multiplies that figure by 60, so for instance Mathieu Darche leads the Canadiens this season with 1.88 goals per 60 minutes of ice time, even though he only has two goals. But he accomplished that in far fewer minutes than it took Benoit Pouliot to get the same amount (Pouliot, by the way, is fifth with 1.01 goals per 60. Go figure).
Secondly, the statistics are tracked not only by a player's performance when he is on the ice, but also his team's performance when he is off the ice. This allows a player to distinguish himself when playing on a bad team or unmasks a bad player who is made to look good because he is on a good team.
This is how the site comes up with what's called a "relative plus-minus rating" which takes both on and off ice situations into account. Let's take Darche as an example again to illustrate the point, seeing as he's tied with Jeff Halpern for the team lead among forwards in this category with a plus-2.63 rating. Darche has been on the ice for five Canadiens goals and two opposing goals at even strength this season, which works out to a plus-2.83 per 60 minutes of ice time rating. When Darche is off the ice, the Canadiens click at only a plus-0.19 per 60 minutes rate. So if you subtract the team's performance when he is off the ice to Darche's performance when he is on, you get a plus-2.63 (yes, I know you actually get a plus-2.64, but the discrepancy is due to figures being rounded off).
In Halpern's case, he has an on ice rating of plus-2.50, but the Canadiens have a minus-0.14 rating when he is off the ice. So that boosts Halpern's rating to plus-2.64 because that's how much better the team is with him on the ice than when he is off.
Still with me?
Quality of competition
So, the first category I plan on tracking this season is quality of competition. It's a great tool to see how confident Jacques Martin is in his player's defensive capabilities, or to see what kind of players opposing coaches are sending to face the Canadiens top guns.
The statistic is derived using NHL shift charts to see which players are on the ice at the same time. An average is taken of the relative plus-minus of each player on the ice, and that figure is your quality of competition number.
For each of these categories, note I haven't included either Andrei Markov or Ryan O'Byrne because of their limited number of games played.
1. Halpern 0.131
2. Pouliot 0.118
3. Lapierre 0.115
4. Eller 0.063
5. Gill 0.044
6. Pyatt 0.039
7. Gorges 0.035
8. Darche 0.022
9. Boyd 0.020
10. Hamrlik 0.005
11. Picard 0.003
12. Spacek -0.001
13. Subban -0.003
14. Cammaleri -0.012
15. Gomez -0.027
16. Plekanec -0.030
17. Gionta -0.034
18. Kostitsyn -0.037
19. Moen -0.049
It's not surprising to see the current top-6 forwards make up the bottom-6 of this list because opposing coaches will try to match their defensive lines against them, and those players tend to have a lower relative plus-minus rating. However, last year's top players in terms of quality of competition were, in order, Gionta, Pouliot, Markov, Gomez, Plekanec and Cammalleri, so this is a shift we're seeing thus far this season.
We can also infer that the acquisition of Halpern was indeed made to lighten the defensive load for Plekanec going forward, who had a high quality of competition figure last year.
I'm very surprised to see Picard's number so high. I've been saying that he's benefitted from playing easier minutes than Spacek and therefore should have been the one to give up his spot upon Markov's return. These numbers suggest otherwise, at least at even strength.
This category is far simpler and is exactly what the name implies, it tracks where on the ice a player starts his shift: in the offensive, neutral or defensive zones.
This is another stat that is telling in how Martin is managing his bench, and also tells a story of how hard a player needs to work to get into scoring areas. Last year, Plekanec's breakout season offensively was made even more impressive because he only began 44.3 per cent of his shifts in the offensive zone.
These figures are all offensive zone start percentages.
1. Darche 72.4
2. Eller 62.1
3. Boyd 58.6
4. Subban 57.5
5. Picard 54.2
6. Pyatt 53.7
7. Hamrlik 53.5
8. Spacek 53.1
9. Cammalleri 52.3
10. Kostitsyn 49.4
11. Plekanec 47.1
12. Pouliot 47.0
13. Moen 46.9
14. Lapierre 44.4
15. Gomez 43.8
16. Gionta 41.6
17. Gorges 39.1
18. Gill 38.7
19. Halpern 34.8
Again, we see what Halpern was brought here to do, but it's clear that Gomez is being used differently this season. Last year he and Gionta began just over 55 per cent of their shifts in the offensive zone, but so far this season that number is down to 43.8 and 41.6 per cent, respectively. It might explain why they're having trouble putting up numbers. Another oddity is that the only top-6 forward to begin more than 50 per cent of his shifts in the offensive zone is Cammalleri.
Relative Corsi rating
Now we're going to be getting a little more complicated, but this is probably my favourite advanced statistic.
The Corsi rating was a system devised by Montrealer and longtime Buffalo Sabres goaltending coach Jim Corsi. Here's an article by the Edmonton Journal's David Staples where Corsi explains in his own words how he devised the system, but I'll give it a shot.
The basic figure used in coming up with a Corsi rating is shots at net, which means shots that actually reach the goalie or go in, plus the ones that are blocked or miss the net. Corsi devised it to understand how much work a goalie was getting because every shot attempt requires some sort of a reaction by the goalie, even if he doesn't need to save it. But it is actually a far better indicator of how often a skater is in an attacking position on the ice.
Essentially, most of the times a shot is released the team is in the attacking zone. So if you are on the ice for more shot attempts against than you are for shot attempts by your own team, then you have a negative Corsi rating, and vice versa.
Desjardins took it a step further at Behind the Net by creating the Relative Corsi rating, which uses the same principle I described above for relative plus-minus and applies it to a player's Corsi number on a per 60-minute basis.
1. Picard 25.4
2. Subban 22.2
3. Gomez 18.5
4. Gionta 14.3
5. Kostitsyn 11.2
6. Eller 8.9
7. Cammalleri 0.1
8. Plekanec -1.5
9. Spacek -2.8
10. Halpern -4.5
11. Pyatt -4.6
12. Pouliot -5.1
13. Moen -5.2
14. Lapierre -9.8
15. Gill -10.1
16. Gorges -13.4
17. Darche -15.9
18. Hamrlik -18.1
19. Boyd -31.2
Though Gomez always gets knocked for how seldom he shoots the puck, he's actually been a Corsi monster for years because of how he's able to set up his teammates for shot attempts. It's no different this year, but not too many of those attempts are finding the back of the net. It's interesting to find Eller so high on the list, making him the only bottom-6 guy in positive figures. And in fact, when I looked at these numbers during the Florida game on Saturday, Eller was sitting second on the team among forwards behind only Kostitsyn.
Also, when you look at these figures and compare them to the zone start figures, you notice that guys like Darche and Boyd aren't doing a whole lot with their offensive zone faceoffs.
Penalties Drawn/Taken +/-
This is a stat that actually isn't compiled in this plus-minus form at Behind the Net, but I find it interesting because it gives you an idea of a player's aggressiveness at both ends of the ice and whether it's ultimately hurting or helping the team. For instance, one would think Subban's reckless play at times has been a detriment (I did), but as you'll see it all evens out in the end.
1. Lapierre 8/3= +5
2. Gionta 4/1= +3
3. Halpern 2/0= +2
4. Gomez 2/0= +2
5. Subban 4/4= even
6. Plekanec 2/2= even
7. Cammalleri 2/2= even
8. Moen 2/2= even
9. Picard 0/0= even
10. Kostitsyn 1/2= -1
11. Gill 1/2= -1
12. Pyatt 0/1- -1
13. Gorges 1/3= -2
14. Eller 1/4= -3
15. Spacek 1/4= -3
16. Hamrlik 0/3= -3
You shouldn't be surprised to see defencemen at the bottom of the list because they are rarely in a position to draw a penalty and almost always in a position to take one. Eller's position at the bottom is normal for a rookie, though with his size and net drive you'd like to see him draw a few more on the opposition.
This is also a case that shows the limitations of Desjardins' web site because there was no data on penalties for Darche, Boyd or Pouliot.
So there you have it, the first of what should be several installments of this look at the Canadiens through the eyes of a math geek. How my Grade 11 math teacher would laugh if he read that one.
I should be producing these every 10-15 games, depending on my schedule, and future ones won't be quite as voluminous as this one because I won't need to explain the stats being used.