Canucks scoring chance totals: 2011-12

Thom Drance and I at Canucks Army, along with many other hockey analysts, like to use numbers like “Corsi” and “Fenwick”, which measure on-ice player shot differentials throughout the season. But critics of simple shot indicators like to say that these numbers don’t matter because they don’t take into account the quality of a shot.

So Thom and I seeked to reconcile this by sitting in front of the TV and watching every minute of every game, sober even, and recording every single quality shot. A web application helps us find out who was on the ice for every single scoring chance.

We’re given a sheet at the end of every game, we post it here, and then we count them up. In the absence of much else better to do today, I figure we’d share our data with the world.

Here are the definitions:

Corsi: All shot attempts, including saved shots, missed shots and blocked shots. This can be expressed as either a differential (where 0 is even) or a rate (where 50% is even).

Fenwick: Like Corsi, but blocked shots are removed.

Scoring chance: An unblocked shot attempt (dribblers excluded) from within the “home plate” area of the offensive zone. The home plate area is defined as the area between the faceoff dots, extending to the “Bowman line” at the top of the offensive zone circles.

Now, many (or some) readers will remember that I have already posted raw scoring chance totals. Using those raw numbers, we can also compare players via time-on-ice. I took every player’s even strength time on ice total and calculated their scoring chance rate per 60 minutes:

# Skater TOI EVF EVA CF/60 CA/60 CD/60
8 Chris Tanev 393.3 120 91 18.3 13.9 4.4
7 David Booth 743.1 235 188 19.0 15.2 3.8
26.1 Mikael Samuelsson 66.3 17 13 15.4 11.8 3.6
20 Chris Higgins 952.5 295 239 18.6 15.1 3.5
33 Henrik Sedin 1257 365 301 17.4 14.4 3.1
22 Daniel Sedin 1096.8 310 256 17.0 14.0 3.0
17 Ryan Kesler 1123.9 327 281 17.5 15.0 2.5
15 Marco Sturm 61 16 14 15.7 13.8 2.0
4 Keith Ballard 675.8 169 149 15.0 13.2 1.8
21 Mason Raymond 744.7 210 188 16.9 15.1 1.8
14 Alex Burrows 1155.9 319 285 16.6 14.8 1.8
2 Dan Hamhuis 1528 410 370 16.1 14.5 1.6
9.1 Cody Hodgson 686.2 163 156 14.3 13.6 0.6
3 Kevin Bieksa 1506.9 395 400 15.7 15.9 -0.2
29 Aaron Rome 602.1 161 164 16.0 16.3 -0.3
25 Andrew Ebbett 152.2 34 35 13.4 13.8 -0.4
36 Jannik Hansen 1033.9 256 263 14.9 15.3 -0.4
23 Alex Edler 1488.8 361 378 14.5 15.2 -0.7
6 Sami Salo 1060.4 258 275 14.6 15.6 -1.0
5 Marc-André Gragnani 193.7 52 56 16.1 17.3 -1.2
79 Mike Duco 48.2 10 11 12.4 13.7 -1.2
41 Andrew Alberts 575.4 120 141 12.5 14.7 -2.2
52 Alexander Sulzer 180.9 37 45 12.3 14.9 -2.7
26.2 Samuel Pahlsson 236.6 57 68 14.5 17.2 -2.8
54 Aaron Volpatti 203.9 32 46 9.4 13.5 -4.1
9.2 Zack Kassian 174.7 45 58 15.5 19.9 -4.5
40 Maxim Lapierre 821.4 151 213 11.0 15.6 -4.5
34 Byron Bitz 104.3 19 27 10.9 15.5 -4.6
24 Mark Mancari 49.9 6 10 7.2 12.0 -4.8
32 Dale Wiese 549 77 129 8.4 14.1 -5.7
27 Manny Malhotra 769 120 205 9.4 16.0 -6.6
42 Bill Sweatt 10.2 0 3 0.0 17.6 -17.6
38 Victor Oreskovich 5.9 0 4 0.0 40.7 -40.7

Oh, hey, look who’s right up at the top…

With Chris Tanev on the ice, the Canucks would get a little over 18 scoring chances per hour and concede a little under 14. That’s the biggest differential on the team. David Booth, a man who supposedly suppresses his teammates’ shooting percentage, also ranks highly. 55.6% of all chances when Booth was on the ice were in Vancouver’s favour, but unfortunately, just 50% of all goals. He’s still very snakebitten, but brings the puck to the right areas.

Note Cody Hodgson. Much better than Zack Kassian. That is how Zack Kassian will forever be judged, unfortunately.

Adjusting for zone starts, now. I only took the players who played 200 or more minutes:

# Skater CD/60 EDS/60 Adj/60
8 Chris Tanev 4.4 5.9 5.1
20 Chris Higgins 3.5 2.0 3.7
7 David Booth 3.8 -6.4 3.1
17 Ryan Kesler 2.5 1.1 2.6
4 Keith Ballard 1.8 2.9 2.1
2 Dan Hamhuis 1.6 2.2 1.8
21 Mason Raymond 1.8 -6.5 1.1
33 Henrik Sedin 3.1 -23.2 0.5
9.1 Cody Hodgson 0.6 -1.3 0.5
22 Daniel Sedin 3.0 -22.6 0.5
36 Jannik Hansen -0.4 6.0 0.3
29 Aaron Rome -0.3 4.5 0.2
3 Kevin Bieksa -0.2 1.0 -0.1
14 Alex Burrows 1.8 -19.8 -0.4
26.2 Samuel Pahlsson -2.8 18.3 -0.8
41 Andrew Alberts -2.2 7.0 -1.4
23 Alex Edler -0.7 -7.5 -1.5
6 Sami Salo -1.0 -6.5 -1.7
27 Manny Malhotra -6.6 44.7 -1.7
40 Maxim Lapierre -4.5 23.4 -1.9
54 Aaron Volpatti -4.1 13.2 -2.7
32 Dale Wiese -5.7 18.6 -3.6

CD/60 – Chance differential per 60 minutes

EDS/60 – Defensive zone starts minus offensive zone starts per 60 minutes

Adj/60 – Adjusted scoring chance differential. The Chris Tanev Index.

Point brought up in the comments of our Corsi/Fenwick post:


So the question that all this begs is what do all these numbers mean? Specifically, what do your adjusted totals tell us about specific players. When I read those numbers, and a variety of alarm bells go off about your analysis.

First of all, let’s just outline what the Sedins have done specifically over the last couple of years: combine for two league scoring titles, win a Ted Lindsay and a Hart trophy, and guide the franchise to two consecutive Presdident’s Trophies and a Western Conference Championship. In other words, they are elite talents. This fact is undeniable.

However, the adjusted Corsi and Fenwick totals do not indicate this at all. In fact, based on these adjusted reports, your numbers would predict more success for Malhotra, Mancari, Sturm, etc. than they would for Henrik and Daniel if O-zone start% was more or less equal. To me, this says one of two things: either Corsi and Fenwick are completely useless at indicating anything resembling success and should be ignored entirely, or the method of standardization is critically flawed in that it skews cases of “specialized deployments” too much one way or another. This second conclusion is particularly significant in this case because we know that Alain Vigneault radicalizes player deployment to a greater extent than any other coach in the NHL.

Now, I’m familiar enough with the concepts of possession to know that they are not completely useless, and in fact they are a pretty decent indicator of something (although my own research into team Corsi% and Fenwick% vs. actual team success has yielded only weak correlations at best), so I’m much more comfortable concluding that there is a problem with these particular adjustments to individual Corsi and Fenwick numbers. Granted, I haven’t looked in to adjusting for zone starts enough to suggest a more accurate standardization, however I can say with a fair degree of confidence that the numbers you have simply do not make sense in the context of what we know about these players.

Therefore, I would either seek a different method of adjusting Corsi and Fenwick for zone starts, or I would disregard the adjusted numbers entirely.


Good point. Either we look at these numbers and theorize that Aaron Volpatti is a better player than Henrik and Daniel Sedin, or what we were doing in the first place was wrong. I halved my original scoring chance adjustment (counting an extra offensive zone start as -.11 of a chance as opposed to -.22) and the results fit more with reality.

I remain convinced that Henrik and Daniel are poor defensive players. Because Alain Vigneault deploys them so dramatically in the offensive zone, we don’t know to what effect just yet. I am also quite convinced that the American trio of forwards are the three best two-way players on the Canucks, and that Chris Tanev is very effective when playing limited, optimized minutes against third line competition. Wonder how soon until he takes the next step.

Oh, and onto the whole “shot quality” thing? Check out the correlation between a player’s scoring chance rate and his Corsi, Fenwick, and goal rate:


R-squared = .81931


R-squared = .80548


R-squared = .40754

Fenwick and Corsi are much better than goals at predicting a player’s scoring chance differential. Fenwick is therefore then probably better at predicting a team’s total goals rate. This is why we disregard shot quality in many instances: it doesn’t predict the actual amount of goals a team will get.

Some players perform marginally better by Fenwick standards than scoring chances. Those players are Keith Ballard (53.1% chances, 49.4% Fenwick) and Mason Raymond (52.8% chances, 49.4% Fenwick). The two players who performed the worse by this measure among regulars were Samme Pahlsson (45.6% chances, 50.3% Fenwick) and Kevin Bieksa (49.7% chances, 53.4% Fenwick).


  • MC Hockey

    One of my comments made it into a post? Sweet. Thank you for the acknowledgement.

    And for the record, I agree with your supposition that Henrik and Daniel are poor defensive players. Simply looking at it from a traditionalist’s perspective would show you that they are not very aggressive playing the body, not fast skaters, and aren’t overly willing to give up the body to block shots. I would argue that most “good defenders” (ie: those that suppress opponent shots, force turnovers, keep play to the outside, etc.) are at the very least one or two of these things. But at the same time, it doesn’t really matter how good they are at defending if they’re deployed exclusively for offense and proceed to lead the entire league in that regard.

    On to the relevant topic, I really wish that you labelled your axis on your graphs. I’m lost as to what data the graphs are showing. I’m guessing the y-axis on all graphs is a player’s Chance Differential expressed as a percentage of total chances when that player was on the ice, whereas the x-axis are Fenwick%, Corsi% and Goal% respectively. Is this correct?

    Also, if I understand it correctly, goal% is just on-ice goals for expressed as a percentage of total on-ice goals for and against. This means that the third graph you have there is essentially using chance differential to predict +/-.

    My question is this: when we know that we need to score more goals than your opponent to win a hockey game, why are we looking at trying to predict Chance Differential when you’ve demonstrated that is has a weak correlation with goal differential? I would think that it would be far more relevant to look at predicting goal differential, as that is the strongest predictor of winning hockey teams that I have been able to find.

    The damning thing about +/- (or goal differential, or goal rate, or goal%, etc.) is that while it’s near impossible for a single player to control his +/- entirely due to it being so heavily influenced by so many forces outside of his control, it’s still the single most important statistic in all of hockey because at the end of the day, which of two teams playing each other wins the hockey game? The one with the better +/- at the end of 60 minutes. This being said, I still don’t believe that it’s an accurate reflection of individual player ability.

    It’s great that you and Thom have demonstrated a definitive strong correlation between chance differential and Corsi/Fenwick. If nothing else, this allows us to make the assumption that in most cases, players with better Corsi or Fenwick ratings are getting more chances as well. However, I would also be interested in seeing Chances For expressed as a percentage of Corsi/Fenwick For to see if certain players consistently generate a higher ratio of their shots from a scoring chance area than others.

    Keep up the good work, guys.

    • I think the problem with using goals as talent indicators is that they’re entirely random.

      If I can sit through 82 games, count up all the quality chances and who was on the ice, and find that they’re closer to shot attempts than goals, it shows just how limited one’s understanding of the game will be when they restrict themselves to purely looking at goals.

      A player can continually blow coverage and get bailed out by a good goaltender. A same player can blow coverage once or twice but a goal will be scored on. On the replays, we pin fault on the player, but we forget that it takes four, sometimes six, good opportunities for a goal to go in because the quality of NHL goaltending is outstanding.

      • orcasfan

        I’m not sure I can buy your supposition that goals are random events in a hockey game. If they were truly random, you would expect a couple of major prevalent outcomes:

        a) Distribution of offence league wide (on a team-by-team basis) would be more or less equal, influenced predominantly by the quality of goaltending against and possession metrics for

        b) Year to year, we would expect to see massive turnover in the individual goal scoring leaders

        Neither of these things listed above is true. By making the statement that “goals are random events” and proceeding to discount them entirely as a talent indicator, the argument being made is that possession metrics are what individual players should be evaluated on.

        In that case, Alex Steen is the best forward in the league (#1 in Corsi On, #3 in Corsi Rel). And judging by Steve Stamkos’ numbers (236th in Corsi On, 96th in Corsi Rel), he’s a little above average at best. Why then does Stamkos have 60, 45 and 51 goals in the last three seasons, compared to Steen’s 28, 23 and 29 goals/82 games over that same period? This difference cannot be accounted for by variance. This is not random. Steven Stamkos will score more goals than Alex Steen this upcoming year, and barring injury, will continue to do so every future year they are both in the league.

        Using the same databases and information sources that you have used, I’ve done my own linear regression and correlation analysis, and the resulting R-squared values do not seem to support the premise that puck possession is a good indicator of either goal differential or winning hockey. In fact, the relationship between 5-on-5 team Corsi and wins is quite weak (R-squared of about 0.33 over the last five seasons). These results would seem to indicate that the basic assumptions on which the chance data argument (that more chances = better players) is based are somewhat flawed. If goals are what you need to win a hockey game, isn’t it logical to look for a strong indicator of goal scoring, rather than a strong indicator of possession?

        Going back to my earlier post, perhaps expressing chances as a percentage of either Corsi or Fenwick events for will yield stronger correlations to offense generated. Maybe the difference in offence between the Sedins and David Booth can be attributed to a larger proportion of chances relative to shots attempted. Maybe this will reconcile some of the difference between success and possession, maybe it won’t. I can’t tell from a cursory glance at the numbers.

  • NuckfiSh

    “Note Cody Hodgson. Much better than Zack Kassian. That is how Zack Kassian will forever be judged, unfortunately.”

    That’s a bit unfair… Cody had spent the last 2+ years learning the Canucks system… Zack had a 17 game crash course.

    Next summer, compare Kassian’s numbers from next season to Hodgson’s numbers from this season, and then we’re free to judge.

    Good stuff tho, cheers.

  • orcasfan

    I’m not sure I understand what you mean when you say that goals are entirely random. I assume you mean that they cannot be predicted. Anyway, not the point I wanted to make.

    Actually, I wanted to ask if any of you have done (or are in process of doing) any study on the zone entry stats for the Canucks. Over at Broad Street Hockey, they have done some fine work on that element, and the results are very useful when we look at one of the aspects offensive play. I would love to see something along those lines for the Canucks!

  • orcasfan

    Henrik Sedin’s 5v5 ZS adjusted (by ignoring first 10 seconds after a zone face off) PDO over the past 4 seasons are 1.039, 1.036, 1.058, 1.038 for a 4 year average of 1.043. Kesler’s over the past 4 seasons are 1.008, 0.977, 1.007, 1.019 for a 4 year average of 1.000. Henrik Sedin is consistently well above Kesler so maybe Henrik really is a better player than Kesler.