Black Box: Week 1



Every Monday we’ll be posting a weekly statistical update on how the Flames have fared with each of their individual players on the ice. This first week will be rather tame not only because they’ve played just a single game, but because two of the major statistical sites on which we rely aren’t set up with 2011-12 data yet. Consider this a trial run, and a good way to see what it will look when it’s all put together.

Advertisement - Continue Reading Below

OZQoC Charts (explanation)

For the past four years one of famed statistician Gabriel Desjardins’ many contributions has been the variety of statistics he either introduced or popularized over at Behind the Net, including the Offensive Zone Starts and Quality of Competition data required for our OZQoC Charts. Unfortunately the site isn’t set up for the 2011-12 season yet, so we’ll have to wait a little while before we have an objective presentation of how each of the Calgary Flames are being used this season.

Even-Strength Scoring (explanation)

If there’s another statistician whose contributions rival Desjardins, it’s Vic Ferrari over at Since extracting Corsi (shot-based) data is truly painful without the assistance of his handy utilities, we’ll wait a short while until he’s got the site up and running for the 2011-12 season. Until then we’ve got Kent’s scoring chance data, and the rest from NHL’s official stats feed. All chances for and against below are per 60 minutes of ice.

Advertisement - Continue Reading Below
Player          ESP/60 CEF CEA CE% SCF SCA  SC%  GF  GA   G%
Scott Hannan     5.1   N/A N/A N/A  25  10 71.4% 5.1 0.0 100%
Mark Giordano    4.2   N/A N/A N/A  21  13 62.5% 4.2 4.2  50%
Lee Stempniak    0.0   N/A N/A N/A  33  20 62.5% 6.5 0.0 100%
Niklas Hagman    4.7   N/A N/A N/A  14   9 60.0% 4.7 0.0 100%
Olli Jokinen     4.0   N/A N/A N/A  24  16 60.0% 4.0 0.0 100%
Roman Horak      5.2   N/A N/A N/A  10  10 50.0% 5.2 0.0 100%
Rene Bourque     5.0   N/A N/A N/A  10  10 50.0% 5.0 0.0 100%
Alex Tanguay     4.0   N/A N/A N/A  20  24 45.5% 4.0 8.1  33%
Chris Butler     0.0   N/A N/A N/A  18  22 45.5% 0.0 0.0   0%
Jay Bouwmeester  0.0   N/A N/A N/A  16  19 45.5% 0.0 0.0   0%
Jarome Iginla    0.0   N/A N/A N/A  17  21 45.5% 0.0 7.0   0%
Curtis Glencross 4.9   N/A N/A N/A  19  29 40.0% 4.9 9.7  33%
Tim Jackman      0.0   N/A N/A N/A  18  36 33.3% 0.0 9.1   0%
David Moss       0.0   N/A N/A N/A   6  29 16.7% 0.0 0.0   0%
Anton Babchuk    0.0   N/A N/A N/A   4  23 14.3% 3.8 0.0 100%
Cory Sarich      0.0   N/A N/A N/A   0  21  0.0% 0.0 0.0   0%
Matt Stajan      0.0   N/A N/A N/A   0  17  0.0% 0.0 8.5   0%
Tom Kostopoulos  0.0   N/A N/A N/A   0   8  0.0% 0.0 8.4   0%

Special teams (explanation)

Once again we’ll postpone the inclusion of Corsi-based data until Ferrari has his site set up for this season, and just look at time on ice this season. First the power play.

Player              TOI/GP PTS/60 CE/60
Alex Tanguay         4.7    0.0    N/A
Mark Giordano        3.9    0.0    N/A
Olli Jokinen         3.1    0.0    N/A
Rene Bourque         3.1    0.0    N/A
Jarome Iginla        2.9    0.0    N/A
Niklas Hagman        2.7    0.0    N/A
Lee Stempniak        2.5    0.0    N/A
Anton Babchuk        2.1    0.0    N/A
David Moss           2.0    0.0    N/A
Jay Bouwmeester      1.7    0.0    N/A
Curtis Glencross     1.4    0.0    N/A

Alex Tanguay was the key man, working the point on the power play along with Mark Giordano. Up front the Flames iced Olli Jokinen between Rene Bourque and Jaroma Iginla on the primary unit. Now the penalty killing.

Player           TOI/GP CE/60
Mark Giordano     4.6    N/A
Scott Hannan      4.6    N/A
Jay Bouwmeester   4.5    N/A
Matt Stajan       3.6    N/A
Chris Butler      3.5    N/A
Tom Kostopoulos   3.1    N/A
Curtis Glencross  3.0    N/A
Rene Bourque      2.3    N/A
David Moss        2.1    N/A
Lee Stempniak     1.8    N/A

Sutter trusted everyone except Cory Sarich and Anton Bachuk on defense, and mostly used Matt Stajan and Tom Kostopoulos as forwards. Using fourth-liners on your penalty kill has the advantage of freeing up your other players to play more minutes at even-strength and with the man advantage.

Goaltending (explanation)

Despite stopping 92.3% of shots at even strength, Miikka Kiprusoff was still dinged with a non-Quality Start since he let in two power play goals.

Advertisement - Continue Reading Below
Goalie           GS QS  QS%  ESSV%
Miikka Kiprusoff  1  0  0.0%  .923

That’s it for this week. With the exception of the missing data, that’s what the weekly updates will look like. What else would you like to see?


  • Just to supplement things here, Sutter has buried Bouwmeester, Butler, Iginla, Moss and Tanguay through the first few games. He even had Babchuk and Sarich starting more often in the defensive end.

    Hannan, Giordano, Glencross and Stempniak, on the other hand, have seen most of the o-zone face-offs.

    Don’t know what’s going on here, but most of that doesn’t makes sense to me (even if it’s a small sample to work from).

    And one of the reasons Moss is getting crushed? Lots of d-zone draws but he probably doesn’t get a lot of the “play catch-up” ice time in the last 5-10 minutes of the third like Tanguay and Jarome.

    • ChinookArchYYC

      I wonder if Sutter is rewarding players with better or worse zone starts. It seems odd, if he is, but Hannan, Giordano, Glencross and Stempniak all played well against the Pens, and they top the list. On the other hand, I really didn’t like what I saw from Bouwmeester, Butler, Iginla, or Moss, and the are were given slot heavy lifting work. That said, it’s hard to judge performance when the zone starts are so different.

      I also expect Stajan will be relied on to take on PK duties this year. If he is to elevate his status, in the eyes of the coach he will need to do so on the PK. I’m okay with this, since I thought he did a pretty good job.

      • It’s possible. A sort of “go with the guys who look most dangerous” thing. Last year when Moss and Jackman were united and kicking the cap out of their counterparts Sutter started putting them in the offensive zone a bit more. In part because they looked dangerous and in part because the rest of the team didn’t.

  • Those Iginla-Moss-Tanguay numbers are baffling. It does, however, highlight a long standing gripe I have with the Ozone % numbers, being that it excludes far too much information and makes the starting positions look way more unbalanced than they are, especially when you are talking about usage as opposed to adjusting corsi rates.

    So Iginla’s Ozone start percentage is seen as 27.3% or it appears he started 72.7% of his shifts in the defensive zone. That’s crazy! And it is also totally misleading.

    That is based on being on the ice for 16 defensive zone faceoffs and 6 offensive zone faceoffs. But Iginla also started another 12 shifts in the neutral zone at a faceoff, so he really only started 16 of his 34 shifts, that he started at faceoffs, in the defensive end. That is only 47% of his shifts.

    And that totally ignores when he starts a shift without a faceoff. has iginal with 50 shifts. He has 4:30 PP time – call that 4 shifts. That means the faceoff data fails to incorporate another 12 shifts or something over 25% of Iginla’s EV shifts per game. As most on the fly line changes happen when the puck is either in the neutral zone or in the offensive zone, probably all 12 of those shifts started with the puck somewhere other than the defensive zone. Now Iginla’s true zone start in the defensive end is something more like 16 out of 46 shifts or 34% of his shifts – pretty close to an even third. Given that the Flames only appeared to have had 25 out of 101 EV faceoffs in the other teams end, the fact that Iginla didn’t have a bunch of shifts start there probably is not that surprising.

    While I can appreciate that you need to adjust corsi figures depending on zone start, using that same percentage as a true measurement of usage can be pretty misleading.

    • Yeah, the nomenclature is deceiving, although it’s been standardized like this because neutral zone starts just don’t have much or any effect on possession. In reality, zone start is “offensive-defensive zone faceoff ratio”, which is a mouthful.

      I hear you on the “on-the-fly” change thing. If we could reasonably capture that data it would help. Probably will one day.

      That said, in the context of ZS% and how it helps or hinders player, I’d prefer to see Iginla et al. with the higher percentage on this team. Certainly not the lowest. Maybe Sutter was focusing more on individual player match-ups, but man…

      • and @Robert Vollman

        I pretty much agree with your points in terms of use of zone start (that is why I said I recognize the Corsi adjustment), but what I was specifically responding to was this line from Kent.

        “Sutter has buried Bouwmeester, Butler, Iginla, Moss and Tanguay through the first few games”

        I don’t think zone start tells us that is what happened through two games. I think the real explanation is that the Flames were getting out played, resulting in lots of faceoffs in their own end. These 5 guys were playing the most minutes and therefore Sutter had no choice but to start them in their own end if they were going to be on the ice. Here is how I justify that view on the numbers.

        The Flames took 101 EV faceoffs by my count (I am adding up all the faceoff wins and losses on and dividing by 5 as they were all EV I think this is an accurate total EV count). 25 in the OFF end 34 in the DEF end and 42 in NEU. Clearly, there was an edge here for the Flames’ opponent.

        The Flames played 101:47 at EV. The five players Kent listed played the following portion of the Flames’ EV TOI during those two games ( TOI :

        Bouwmeester – 40:36(39.9%)
        Butler – 35:54(35.3%)
        Iginla – 37:18 (36.6%)
        Moss – 25:06 (24.7%)
        Tanguay – 30:08 (29.6%)

        If we extrapolated those to the proportion of the Flames face offs in each zone we would have expected the following faceoffs for each:

        Bouwmeester – OFF – 9.98, DEF – 13.57, NEU – 16.8
        Butler – OFF – 8.83, DEF – 12.00, NEU – 14.83
        Iginla – OFF – 9.15, DEF – 12.44. NEU – 15.372
        Moss – OFF – 6.175, DEF – 8.398, NEU – 10.374
        Tanguay – OFF – 7.40, DEF – 10.06, NEU – 12.43

        So the difference from the expected starts less the actual starts is:

        Bouwmeester – OFF – +4.98, DEF – +0.57, NEU – +2.76
        Butler – OFF – +3.83, DEF – -2.00, NEU – +1.83
        Iginla – OFF – +3.15, DEF – -3.56, NEU – +3.37
        Moss – OFF – +1.18, DEF – -1.60, NEU – +0.37
        Tanguay – OFF – +2.4, DEF – -1.94, NEU – +2.43

        So Iginla was “buried” because he started 1 or 2 less of his 23 EV shifts per game in the offensive zone than we would have expected? Bouwmeester was buried because he actually started about as many shifts with a faceoff in the defensive end as we would expect? And why did Bouwmeester start a disproportionate number of his shifts on the fly?

        I think we need to count the zone starts, but until we look at the absolute measures, saying things about usage of players is very suspect. I think when you look hard at these numbers, and not the ratios, the story is Calgary getting outplayed, but Sutter getting his “top line” out for their ice time regardless of where the faceoffs were taking place. The ratios are misleading.

        • Fair enough. I extrapolated based on context of a larger sample of games when, in fact, we’re talking numbers too small and particular to properly project.

          I’ll put it this way: I hope Iginla et al. don’t have similar ZS% 40 games into the season.

          • That we can agree on. Frankly, I think it is fair to say Sutter could have worked harder to give them higher ground as opposed to basically ignoring the starting position which is what I see.

            I tried to comb through the play by play to see if maybe Sutter was trading zone start for matching the Iginla line agasint lesser competition – but I couldn’t identify, never mind quantify, a pattern.

          • That’s helpful.

            I still can’t find a reliable pattern. Against Pittsburgh at home it looks like Sutter went the first half matching Iginla’s line against the Malkin line as there only appears to be one EV shift that they weren’t matched up. Then in the second half tried to get away from that matchup a bit and you see more of the Dupuis-Cooke line showing up with Iginla on the ice. Only one shift all game with Iginla’s line against the Staal line so I guess he was keeping him away from the second best line anyways.

            In St. Louis it looks to me like St. Louis tried to match McDonald-Langenbrunner-Backes (which would be their first line with Stewart-Oshie-Berglund as the second?) but Sutter was trying to work away from the match up with Iginla coming on half way into other lines shifts, right after power plays etc.

    • Valid points, but your beef seems more to be about the intrepration of the statistic rather than the statistic itself.

      Of the face-offs Iginla has taken so far at even-strength, most are in his own end. No one is disputing that.

      You are disputing how some people are interpreting that to mean Iginla is being used in his own end to the same extent.

      Nothing is wrong with the statistic. Rather than change the statistic to match people’s intrepretations, people should be changing their interpretations to match the statistic.

      That being said, over time they tend to align – we won’t be having this argument if we’re at the 40 game mark and Iginla’s offensive zone start is still just as low.

  • BobB

    Despite stopping 92.3% of shots at even strength, Miikka Kiprusoff was still dinged with a non-Quality Start since he let in(??) two power play goals.

    This is exactly what I commented on (comment #7) in the original article. The influence of PK on Quality Starts.

    It’s almost like the measurement needs ev focus or at least another column (like QualESSt), because busting on a goalie for the undisciplined play of his team in front of him (in that game) and the team effects of the PK… misses the mark to me.

    There is a distinction between Kipper’s first game at a .923evsv% and Karls at a .844evsv%. 2 even strength goals vs 5?

    If there isn’t a distinction, there should be one. Why not just use only evsv% numbers?

    If there is an argument why a .923ev performance gives you the same chance of winning as a .844 even performance, I’d be interested in hearing it. Especially if those two numbers continued over a very large sample size.

    Imagine if an opposition team went 5 for 15 on a fictitious one game PP bonanza… we would pin it on goaltending?

    • Why not just use even-strength save percentage? Good question.

      First of all, goalies shouldn’t get a free pass when killing penalties – it’s part of their job to stop those, too.

      Secondly, the same sort of question was asked in baseball when Quality Starts were introduced there. Why not include hits, walks, wild pitches, and so on?

      There’s a strong temptation to add things to every stat, but the idea of Quality Starts (in both sports) is to take a quick look at the scoreboard and/or the box score and get a sense of whether the goalie did or didn’t do his job. It’s meant to replace wins after all, not something more complex, like GVT.

      In the short run there may be some aberrations (like Karlsson’s 8-for-11 last year), but it will work out in the long run. And was Kiprusoff’s game one performance really a Quality Start, despite his .923 even-strength save percentage? Did he really gie the team a 75% chance to win? I think QS got it right: no.

      I like was Section205 was doing – instead of changing QS let’s introduce something new.

      Let’s figure out an expected save percentage, and then make the comparison.
      Method 1: Use scoring chances, as Section205 described
      Method 2: Use shots, but break it down by ES, PP and SH situations.

      Average ES SV% is .920
      Average SH SV% is .871
      Average PP SV% is .913

      Given that, Kipper should have stopped (.920 * 26 + 10 * .871) = 32.63 shots (he stopped 32), and Karlsson should have stopped (.920 * 32 + 2 * .871 + 1 * .913) = 32.095 shots (he stopped 30).

      So neither start was above-average, although Kipper could have been with one more save – which actually would have pushed him into Quality Start territory. The added complexity might not be helping us.

      There are other methods we could explore, like using all Corsi events (broken down by PP, SH and PK) – that reduces scorer bias but is far more complex.

      I don’t know, what does everyone else think?

      • BobB

        I agree, and we are looking for more information than the scoreboard: “We lost 5-3 and 5-2, goaltending wasn’t very good” We want more fineness, not more or equal coarseness. (As an aside… I think game 1 the goalie DID give the team a 75% chance of winning…. they were down by one goal, pulled the goalie, and were scored on in an empty net! Did the skaters give the team a better chance of winning being outshot 37-17?) Baseball is not hockey. A pitcher is nothing like a goalie, he’s more like a QB. They are the inverse in that pitchers are the leading edge of the sequence.

        I see that we’re trying to track consistency with this, but with goaltending it has to be:

        “Did the goalie play well/poorly?, did the team play well/poorly? How do those influence one another? How did both of those influence the final score?” The metric should function and reflect “storm-of-the-century” limits: “Best goalie in history playing on the worst team in history vs worst goalie in history playing on the best team in history and not coming back 50/50 or 55/45”

        Otherwise it’s just another vague measure for goalies like wins/losses. Hoping for that 60% mark. There are plenty of games where you can play well and get blitzed and (by this measure) won’t get a QS and vice-versa. Then the same criticisms are true as they are of “Wins for a goalie”. 13SA, 1 brutal GA, Team dominates – QS! (Team win, not a goalie win, and the goalie is thinking “That was brutal”). Even then tracking standard deviation from their own avg sv% is a better indicator of consistency, and just as flawed!

        “What rate (expressed in a % of 100) was the goalie the weak link in the defensive game? What rate was the defence (Team) the weak link in the defensive game?” – That tells me something other goalie stats don’t. And that tells me more of the difference between the goalie performances in Game 1 and Game 2.

        We should either track it simply by attributing special teams to the team, and pulling ES out as more info, (as PK frequency is much less influenced by the goalie, being put in high quality against situations for 2min at a time) or complexly by developing a QS metric for the team, to compare against the goalieQS metric.

        I don’t want a stat that makes me dig up Team effects to cross reference and validate… like PP/PK, SF/SA, chances for/against… I want the stat to tell me by looking at it (especially on a team as inconsistent as Calgary, game to game, period to period, shift to shift). I want it to say “Goalie X was good this often/bad this often… but also: and here is WHY” (or at least more reasons why, even if the answer is: because he sucks)

  • Section205


    So far we have seen 4 goaltender performances up close:

    In order of SV%
    1. Kipper .889SV% 36 shots against, 22 scoring chances against
    2. Halak .882%, 17 SA, 8 SCA
    3. Karlsson .857%, 35 SA, 16 SCA
    4. Fleury .850%, 17 SA, 14 SCA

    No quality starts yet. I think that is accurate. But that is interesting for analysis, because there are no clear cut leaders. Was Fleury the worst? No he was the best.

    I look at ratio of SCA/SA as a proxy for “degree of difficulty” – knowing that not every SCA is a shot on goal.

    Degree of Difficulty: Fleury .823, Kipper .611, Halak .471, Karlsson .457

    My Rating = Multiply SV% x Degree of Difficulty:
    1. Fleury .700
    2. Kipper .543
    3. Halak. .415
    4. Karlsson .392

    I would say Fleury was closest to a quality start.

    This is not perfect, but I think best reflects the performances. Halak had it much much easier than Fleury, and Halak’s bad first goal hurts his stats. STL won in spite of Halak, they prevented a lot of SA and SCA. Both Karlsson and Kipper got shell-shocked, but a lot of Karlsson’s shots were not scoring chances. Karlsson had at least one bad goal.

    What about (SCA-GA)/SCA?
    1. Kipper .818%
    2. Fleury .786%
    3. Halak .750%
    4. Karlsson .688%

    • BobB

      It’s interesting what you’re trying to do, but I get nervous as soon as we start adding in “Scoring Chances” because it relies so heavily on that metric being concrete or it will multiply further out of whack.

      I also think of the four goalies we’ve seen, Kiprusoff has been the best (and I sincerely hope that’s not just homerism.)

      Fleury faced half the shots, and 8 less scoring chances but he let in only one less goal?

      Whether both Kipper and Fleury were great, or were she–it. Kipper to me played the better game.

      Another thing I was tracking last year is Quality Starts for the TEAM!

      It’s all fine to say how consistent our goalies are, but why not do the same for the Team with Shots/Shooting %/ Shots against/ PPfor/PPagainst.

      What if our goalie is 60% QS and the team is 90%QS….. or goalie 60%QS and the team 25%QS at the end of the season?

      Cause it’s fine to say “Neither goalie gave us a Quality Start in the first two games”, but did any of the other 20 players give a quality start…. getting so few shots and allowing so many?

  • Section205

    Rain Dogs, I like the Team QS idea and would like to see your findings.

    You make a good argument for Kipper being the best (relatively) of the Non Quality Starts. I think My Rating puts too much emphasis on Degree of Difficulty.
    My (SCA-GA)/SCA calc seems to support your argument for Kiprusoff.

    I’m leaning towards a Rating (similar to QB rating) for goaltenders. Knowing that it will be far from perfect.