• On Corsi% Rel, TOI and Benoit Pouliot

    by  • April 3, 2014 • Hockey • 22 Comments

    For the uninitiated, Corsi% Rel is a stat that takes your Corsi% and compares it to what your team does when you aren’t on the ice. So if the Oilers have a Corsi% of 48.7% when Martin Marincin is on the ice and 40.2% when he isn’t, we say that Marincin has a Corsi% Rel of 8.5%. (It’s pretty depressing if you’re a fan of a team with a guy who has as Corsi% of 48.7% and a Corsi% Rel of 8.5%.)

    One of the objections that you hear to numbers is that teams don’t buy them. In a general sense, I think that that’s true. I doubt that coaches are keeping a close eye on the Corsi% Rel on their roster. However, they seem to come to similar conclusions.

    If Corsi% as a metric of how a player is doing makes sense, and if coaches want to play their best players the most, you’d expect to see that players who get more ice time have a better Corsi% Rel. I was thinking about Benoit Pouliot, who has frequently had an excellent Corsi% Rel, today and thought I’d do a quick graph of Corsi% Rel vs. 5v5 TOI for forwards.

    You can see from the graph that as 5v5 TOI increases, so does the Corsi% Rel, generally speaking. Although coaches might not think the game in terms of Corsi%, they are, in a general sense, behaving as one would if they did. Numbers people think guys who are better at generating Corsi% are better players, all other things being equal; coaches tend to play forwards who are better at generating Corsi% more. So if you don’t buy into Corsi% you can, to a certain extent, take it up with the coaches. They’re the ones who behave as if it’s a good metric.

    I bring this up because I was thinking about Benoit Pouliot today. Cam Charron wrote about him last summer in the context of him being someone that the Leafs should acquire. The Leafs did not and the Rangers picked him up. He’s been a third liner on their team but he has a killer Corsi% Rel, particularly when you take into account what guys playing 11.15 5v5 minutes generally do.

    As Charron laid out, this isn’t unusual for Pouliot. It’s bizarre to me that he keeps bouncing around and making teams better and then not getting much of a contract. He’s currently on a one year deal at $1.3MM in New York and available again this summer. The Rangers are probably the favourites to come out of the Metro Division, in large part because the bottom end of Pittsburgh is a wasteland. I assume that there’s something irritating about Pouliot – sideburns or something – but whatever it is, it’s hard to imagine it being more important than the fact that he seems like an awfully useful bottom of the roster type who hasn’t seemed to require a lot in terms of money or term.

    Email Tyler Dellow at tyler@mc79hockey.com

    About

    22 Responses to On Corsi% Rel, TOI and Benoit Pouliot

    1. Triumph
      April 3, 2014 at

      My guess as to part of why Pouliot struggles to find work every summer is his propensity for penalties – he takes a fair number for a guy who isn’t tough, w/ a P taken/60 that looks to be around 1 . I think a lot of them are in the offensive zone too, and this is something I’ve been wondering about watching Jaromir Jagr and his half-hooks all year – I think offensive zone penalties are underrated by coaches, in that guys who take a fair number of stick fouls in the O-zone tend to be big play drivers because of all the turnovers they force. Coaches tolerate it when it’s Jagr and Kovalev doing it, but if it’s Pouliot or Alexei Ponikarovsky they tend to have less of a tolerance.

      More to the point, players who are offensively gifted but don’t kill penalties and don’t play well enough to stick on higher lines tend to be undervalued – there’s always a few of them available in late July.

      • Triumph
        April 3, 2014 at

        Also, having a French last name doesn’t help.

      • PopsTwitTar
        April 3, 2014 at

        would be interesting to see if your idea about Ozone penalties is correct. I dont know if there are enough of them to draw any solid conclusions about a type of player who tends to take them though.

    2. PopsTwitTar
      April 3, 2014 at

      Until these types of stats are more widely used in front offices, he’s the kind of guy who cant overcome his potential. Decent size guy who doesnt play particularly “big”. Great hands who doesnt particular make a lot of plays. And he was a top 5 pick who never had that outstanding offensive year. But hes exactly the kinda guy Id want on my bottom 2 lines.

      • Roke
        April 3, 2014 at

        Another one of those cases where people evaluate players based on what they think they should be rather than what they are.

        Pouliot was a part of the best Habs’ 4th line I’ve seen in my time following the club playing with Desharnais and Darche. Granted, two of those guys were playing in roles below their abilities but they provided some offense with a slight Dzone tilt against the opposition bottom-6 and pushed the puck toward the offensive zone.

        Pouilot’s better than heralded bottom-6 forwards like McClement, Moen, and Prust yet because a lot of teams are risk-averse and try to minimize mistakes from their bottom-6 Pouliot ends up a bargain every year.

        • PopsTwitTar
          April 4, 2014 at

          Well this is another thing against Pouliot – he’s not a “banger” or “energy guy” or a “defensive specialist”. Coaches/GMs love to fit bottom 6 guys into one of those roles.

    3. DD
      April 3, 2014 at

      I think your sideburns comment is probably apt.

      The memories of Pouliot I have from junior are that he would these poor stretches during games where he would be turning the puck over, not backchecking and generally playing terribly. He would get a earful from the coaches on the bench and his play would dramatically improve the next shift.

      I have a feeling that might have carried over to the NHL where he’s had to get the occasional kick in the pants from coaches. And no matter how well he plays the rest of the time, that’s what sticks in their mind. I’d imagine players like that irritate coaches more than any others, even more than players who aren’t NHL-caliber. Should we call it the Dustin Penner effect?

    4. daryl
      April 4, 2014 at

      Classic line: “It’s pretty depressing if you’re a fan of a team with a guy who has as Corsi% of 48.7% and a Corsi% Rel of 8.5%.”

    5. Pat Mc
      April 4, 2014 at

      Sideburns is right. I have a friend who is very involved in hockey in Sudbury and his description of Pouliot was that he was (as a junior) a guy who was very good at hockey but who did not particularly love it. It happens to be his job and hockey tends to be pretty conservative, they love their rink rats. I would suggest that is probably part of the reason he bounces around right there. Maybe he’s like the Ball Four guy, Bouton? Sports’ lifers don’t trust guys like that.

    6. sacamano
      April 4, 2014 at

      “Although coaches might not think the game in terms of Corsi%, they are, in a general sense, behaving as one would if they did. Numbers people think guys who are better at generating Corsi% are better players, all other things being equal; coaches tend to play forwards who are better at generating Corsi% more. So if you don’t buy into Corsi% you can, to a certain extent, take it up with the coaches. They’re the ones who behave as if it’s a good metric.”

      What a perverse, ass-backwards way of thinking about the usefulness of statistics.

      Coaches don’t “behave as if they believe Corsi% is a good metric”. Corsi% is a good metric because it helps the rest of us understand why coaches play the players they do. As Fenwick has often said, useful statistics are much more likely to confirm what we already know than to reveal something we don’t.

      A player’s Corsi% is one measure (a pretty good measure it turns out) of a player’s usefulness. But the fact the some players get more or less ice-time than their Corsi% suggests they “ought to” demonstrates that other factors are also important that are not included in Corsi%.

      It is amazing to me that you consistently seem to imply that coaches just fluke or back their way into doing what your stats tell them they ought to do (you often phase this as ‘intuition’), rather than starting from the viewpoint that the coaches in the room are actually pretty good judges (i.e., they assemble a whole lot of sensory data including but also above and beyond the information that contributes to Corsi% and use their judgement) of the talent on their own team, and that stats can help the rest of us learn why they use players the way they do.

      If I were a coach, I would look at your chart and say to myself: “This is exactly why I don’t spend a lot of time considering Corsi% — it doesn’t give me any more information than I already know; I’m already playing the best players the most. And furthermore, my judgement includes lots of information that Corsi% misses.”

      It is similar to what Hall said. Tracking his Corsi% doesn’t do much to tell him how he should improve, what he is doing wrong.

      It is a fantastic tool for fans, general managers and sports betters. But on a day-to-day basis, I can absolutely see why coaches and players don’t spend a whole lot of time on it. For the most part, it is simply telling them what they already know, without all of the extra information which is what they actually need to know.

      See — those Corsi% guys are just using mathemagic to tell me something I already know (I

    7. Tyler Dellow
      April 4, 2014 at

      It is amazing to me that you consistently seem to imply that coaches just fluke or back their way into doing what your stats tell them they ought to do (you often phase this as ‘intuition’), rather than starting from the viewpoint that the coaches in the room are actually pretty good judges (i.e., they assemble a whole lot of sensory data including but also above and beyond the information that contributes to Corsi% and use their judgement) of the talent on their own team, and that stats can help the rest of us learn why they use players the way they do.

      Jesus. I praise coaches all the time. It’s been a theme here for a long, long time – I have a substantial respect for them. I literally have a post that opens with this sitting in my drafts right now:

      A while back, I was listening to some very successful hockey coaches talk about some fine point of the game. I was incredibly impressed at how they had thought through the problem from every angle, considering every single possibility and the risks and rewards associated with each.

      As I’ve said for a while here, I’m leerier than most of criticizing coaches. I think that they know their players incredibly well and have put a ton of time into thinking through how things should be done. However, I do think that there’s a bit of a potential weakness in terms of knowing what a player does well and does poorly and assembling it into a proper value judgment of a player.

      My complaints with coaches, to the extent that I have any, are that a) they’re too conservative and b) they sometimes fail to see the forest for the trees. I think they’re making complex judgments, most of them are getting most of them right and that it’s amazing that they’re able to process the volume of information that they’re processing without any useful data.

      Do I think that incorporating data into what they’re doing would help most of them? Sure. Do I think a lot of GMs are outright awful? Sure. But I’ve spent a lot of ink defending coaches and this is a bullshit observation. I was pretty much the last guy in Edmonton defending MacTavish when he was coach. A lot of Oiler fans wanted Krueger gone after last year – I spilled a lot of ink defending him and arguing that there should be deference to the coach. The team took an obvious step back and I said he should get a summer to figure out what he’d done that worked and what he’d done that hadn’t worked. I think the same about Eakins this year.

      FYI: a search for “intuition” reveals no results. A search for “intuitive” brings up 31. I looked through the ones this year and last and they’re all things like me saying “This isn’t intuitive” or “this is intuitive.”

    8. sacamano
      April 4, 2014 at

      Well, that’s a new way to cite authority — use your FUTURE writings to support your past statements. Beautiful.

      Look, I didn’t say you hack on coaches’ abilities — it is the way you support them which reveals a rather peculiar way of thinking about what constitutes “data”.

      It is again illustrated in your response where you note that “it’s amazing that [coaches] are able to process the volume of data that they’re processing without any useful data.”

      The idea that “useful data” = ‘numerical summaries’ is an awfully unimaginative way to think about information and decision making.

      Can numerical summaries be helpful? Of course! Are they inherently more “useful” than other kinds of data? Not necessarily. Do coaches who do not use numerical summaries not have any “useful data”. Puh-leeze.

      In the way you discuss stats, you tend to put the ass before the head, so to speak. In this instance, you suggest that coaches “behave as if they believe it is a good metric”. In fact, it is the metric which ‘behaves’ as if it is a good coach (i.e., the metric does a good job of modeling what actually happens in the world – coaches’ decisions regarding ice-time).

      Statistics like Corsi% are approximations of events going on in the world — they are not the events themselves. Coaching decisions (player performance, etc.) are the events that your metrics are trying to model and understand.

    9. Tyler Dellow
      April 4, 2014 at

      Well, that’s a new way to cite authority — use your FUTURE writings to support your past statements. Beautiful.

      Quickest thing I had at hand. I guess I could have just invented something about always having praised the intuition of coaches. There are a ton of examples in the archives here of me defending coaches that people wanted gone.

      It is again illustrated in your response where you note that “it’s amazing that [coaches] are able to process the volume of data that they’re processing without any useful data.”

      The idea that “useful data” = ‘numerical summaries’ is an awfully unimaginative way to think about information and decision making.

      I actually went back and edited the comment, presumably while you were commenting and changed the first “data” to “information” so as to distinguish between numbers and information, which was my intent when I wrote the comment. So I don’t really disagree with this but then I don’t think that I think this way.

      In the way you discuss stats, you tend to put the ass before the head, so to speak. In this instance, you suggest that coaches “behave as if they believe it is a good metric”. In fact, it is the metric which ‘behaves’ as if it is a good coach (i.e., the metric does a good job of modeling what actually happens in the world – coaches’ decisions regarding ice-time).

      I think you’re reading an awful lot into one line. The point I’m making is that, whatever coaches may say about this stat and its value, they behave in the way one would if one had faith in it. I really think that this is pretty clear from this sentence: “Although coaches might not think the game in terms of Corsi%, they are, in a general sense, behaving as one would if they did.”

    10. sacamano
      April 4, 2014 at

      Your edit on the information-data line makes it even more obvious that you only consider numerical summaries to be “useful data”. All of the other information coaches rely is . . . what? Not data? Not useful? This is exactly the problem with the way you discuss it. It gives a pretty clear impression that numerical summaries are useful and everything else is not. Can you really not see how a coach would read that?

      “The point I’m making is that, whatever coaches may say about this stat and its value, they behave in the way one would if one had faith in it. I really think that this is pretty clear from this sentence:

      And the point I’m making is that by phrasing it that way you are patronizing coaches. You are essentially saying to them: “You may not think that Corsi% is useful, but look at how you coach, you do it the same way as people who do think Corsi% is useful; therefore, what is really going on is you are unknowingly using Corsi%. Har har, we know why you are giving players ice time better than you do because you don’t use any “useful data”.

      When in fact, coaches have a very good idea of talent that does not depend on their use of Corsi%.

      I think you need to be careful in distinguishing between “Corsi% is not useful” and “Corsi% is not useful for ME”. Again — if I were a coach, I’d look at your chart and say “Why would I use Corsi%, when it is simply telling me what I already know, but without lots of other information that I also know”.

      coaches are NOT behaving in the way they would if they had faith in it. They are behaving using the information they believe is important, and can easily point to your graph and say

      • Tyler Dellow
        April 5, 2014 at

        Your edit on the information-data line makes it even more obvious that you only consider numerical summaries to be “useful data”. All of the other information coaches rely is . . . what? Not data? Not useful? This is exactly the problem with the way you discuss it. It gives a pretty clear impression that numerical summaries are useful and everything else is not. Can you really not see how a coach would read that?

        If you’re determined to read it a certain way and unwilling to accept my assurances that I don’t mean it the way that you’re taking it, there’s really nothing that I can do for you. I’m trying to distinguish between numbers and other forms of information.

        Again — if I were a coach, I’d look at your chart and say “Why would I use Corsi%, when it is simply telling me what I already know, but without lots of other information that I also know”.

        My response to that would be that there are probably some cases where the coach is playing a guy who Corsi doesn’t like where he’s right, and probably some cases where he’s not playing a guy who Corsi likes and he’s wrong. I would bet that coaches have a tendency to overvalue guys who are on a PDO run, for example. We’ll see if Tyler Bozak’s still playing number one minutes two years from now. We saw with Whitney that when the PDO dried up the party was over.

        It’s another piece of information to be reconciled with the information that the coach already uses. In most cases, it’s going to tell him that he’s right. I don’t view that as being “Haha, we know why you’re doing what you’re doing” – I view the cases where there’s disagreement as being more interesting.

        • sacamano
          April 5, 2014 at

          If you’re determined to read it a certain way and unwilling to accept my assurances that I don’t mean it the way that you’re taking it, there’s really nothing that I can do for you. I’m trying to distinguish between numbers and other forms of information.

          And herein lies the problem. You fail to think about the way in which you phrase and frame the issues will be read by others, and then you wonder why coaches, players, etc., don’t jump into the your arms.

          Read these line through the eyes of a coach:
          “Although coaches might not think the game in terms of Corsi%, they are, in a general sense, behaving as one would if they did . . . So if you don’t buy into Corsi% you can, to a certain extent, take it up with the coaches. They’re the ones who behave as if it’s a good metric.

          And think about the cause and effect relationship your words imply by phrasing it that way. Again — coaches do not “behave” as if Corsi is a good metric. That’s a bizarre way to phrase it and to think about what metrics are and do. Corsi is a good metric not because coaches “behave according to it”, but because IT does a good job of modeling player quality as reflected (in this instance) through coaching decisions regarding playing time. The model comes after the actual events, not before.

          I’m not arguing at all that the metrics aren’t useful. I love them. I’m arguing that the way you discuss them tends to reify them as a “thing” that coaches unconsciously model, rather than phrasing things whereby the metrics are the model of the things people do (coaching decisions, player quality, etc.).

          I don’t disagree with your general point at all that your chart demonstrates that Corsi Rel% is useful. But why not phrase it as:

          “If Corsi% as a metric of how a player is doing makes sense, and if coaches want to play their best players the most, you’d expect to see that players who get more ice time have a better Corsi% Rel” . . . . Sure enough, when we examine the relationship we see that Corsi% Rel has a strong positive linear relationship with 5v5 ice-time/game. As such, it is pretty clear that Corsi Rel% does a good job of describing player quality as judged by the coaches, which demonstrates that it is a useful measure.

          And, yeah, this is all very interesting. But, to me, the really interesting implications of this post are:

          1) This strong positive correlation probably explains WHY coaches are not so keen on Corsi%. Again, there is a difference between “This is not useful” and “This is not all that useful for ME”.

          2) The really interesting part of this is, as you note, will be “cases where there’s disagreement”. In order to explain the disagreements, either the coach is making a “mistake” (e.g., your example of playing a guy too much based on the lag from a recent hot streak) OR the metric models are missing information that coaches find useful (e.g., maybe coaches have found it useful to play a guy who has just entered a cold-streak MORE to help him out of a bad luck slump).

          How to sort out what is missing on each end is the trick.

          • sacamano
            April 5, 2014 at

            One last one just for the hell of it.

            You note “I would bet that coaches have a tendency to overvalue guys who are on a PDO run, for example.”

            But is this really a mistake? Is it really ‘overvaluing’? It would certainly be a mistake for a GM to sign someone to a big contract on the basis of a PDO run, or for a sports better to mortgage their house on the basis of past results of a player on a PDO run. But is it a mistake for a coach to give a guy on a PDO run more ice time? PDO runs are real. They happen. They exist. This is not the same as saying that they are predictable, or repeatable or the same as talent. But, from a coaching perspective, in the here-and-now, does it matter? Players on PDO-runs are producing results, so doesn’t it make sense to keep throwing them out there knowing until they cool off? Add to this the fact that there is sometimes a sort of positive feed-back with players on hot-streaks — they are hot, so gain confidence, so play more assertively, which can drive results independent of but adding to their luck, etc. The same thing with players on cold-streaks, they get down on themselves, play more tentatively, which further drives their results down, etc.

            I can see these stats having a ‘scale of usefulness’, depending where on the hockey chain you reside

            GMs & Gamblers — very, very useful for identifying over-performing/under-performing players over the long-term, and $ and contract negotiations

            Coaches – Useful, but more as a second-glance for what my eyes are telling me.

            Players – Even less useful. I already know if I’m playing well or poorly or if I’m on a streak or a slump. HOW do I change this is what I need to know.

      • May 6, 2014 at

        Cormac, I’m just pushing your botutns. The more mature fans here can appreciate the fact that there is more to the rivalry than that 7-0 game. It dates back for decades and has been back and forth for a long time. It’s a rivalry that deserves some respect and I found your choice of video to be way too self aggrandizing. It’s even worse given the outcome of the game don’t you think? It would be like editing a video of Ali Frazier and only show the punches landed by one of the boxers.

    11. sacamano
      April 4, 2014 at

      errrr. That last line starting with “coaches” should have been deleted. Damn cut and paste.

    12. April 11, 2014 at

      This is precisely how I’ve felt about Kyle Wellwood. What made it worse about Welly was that he’s a decent faceoff guy, and the Jets not only passed him up as a depth forward but never played him at center (choosing Jim Slater or Alex Burmistrov instead).

      Wellwood would have netted league minimum, and not a single team in the league could apparently spare 3rd line and 2nd PP unit minutes. And now he’s friggin retired.

    13. Robb
      May 1, 2014 at

      Maybe GMs are like me, and keep getting him confused with Marc-Antoine Pouliot. I know I would avoid signing him if I made that mistake.

    Leave a Reply

    Your email address will not be published. Required fields are marked *