• Boudreau and Turning Corsi% Into Goal For%

    by  • April 20, 2014 • Hockey • 7 Comments

    I just want to wrap up my fixation on Boudreau and offensive zone faceoff losses, which I’ve discussed (in order) here, here and here. It occurred to me that I should probably discuss the goals for/against.

    Nobody who pays attention to Corsi cares about it for abstract reasons. They care about it because Corsi is tightly correlated with goals, there are way more Corsi events in a game than there are goals and goals are subject to considerable randomness. So when I’m talking about a player who I think helps Corsi, what I’m really saying is “I think this guy does things that will help his team outscore the opposition.”

    As I’ve mentioned repeatedly, I’m a believer in breaking the game down into digestible pieces based on sensible criteria. One of the ways that I do this is by identifying the period of time following a faceoff in which there is an effect based on the location/outcome of the draw. I have a database that has all of the 5v5 faceoffs from 2007-13 and has classified all of the Corsi events based on whether there’s a faceoff influence. Here’s what the data shows us:

    The relationship between Corsi% and GF% shows up again. The astute observer will note that teams winning offensive zone and neutral zone faceoffs don’t get quite the GF% that their Corsi% would suggest. I suspect that that’s tied to the defensive team being more likely to have five guys behind the puck when the shot is taken. It’s not a massive effect though – if someone offered you the chance to only take offensive zone faceoffs, all of which you’d win, you’d live with the fact that the GF% isn’t quite up to the Corsi%.

    Anyway, Boudreau. Piling up Corsi% without getting goal difference is pointless. This is why stat guys just sort of stare blankly at people who say “What, you think guys should just shoot when they cross centre ice?” Stats guys don’t think you should fire from anywhere; they do think that you should do the things that lead to you having more opportunities to shoot the puck from the places people shoot the puck.

    If the Corsi% that Boudreau-coached teams have piled up after losing offensive zone faceoffs has any value, it will show up in their GF%. Higher Corsi% should, over time, equal higher GF%. Does it check out?

    Yep. What I find striking about this is how different the Ducks and Capitals are. Nobody would refer to the Capitals as a big, heavy team but you hear ad nauseum about what a big, heavy team the Ducks are. It doesn’t seem to matter – either team, Boudreau gets them to murder the opposition after they lose an OZ faceoff at 5v5 and turn it into goals.

    At the risk of raising the ire of my good pal Sacamano, I don’t know how you could reasonably expect a coach who coaches 82 games a year against 29 different teams and has his own team to be responsible for as well to suss out who’s got the most effective tactics for him to steal without a back room running this sort of analysis. The fact that Boudreau’s been getting these kinds of results for years and it hasn’t been sussed out and either nullified or lifted by every other team in the NHL is amazing, right up until you consider that, without numbers, breaking the game into tiny pieces like this and comparing thirty teams is pretty much impossible.

    Email Tyler Dellow at tyler@mc79hockey.com


    7 Responses to Boudreau and Turning Corsi% Into Goal For%

    1. jvuc
      April 20, 2014 at

      Ummm are you suggesting coaching can influence SH% at a team level. Blashphemy.
      Good stuff.

      • Tyler Dellow
        April 20, 2014 at

        Good news – I’m not saying that.

    2. Steve
      April 20, 2014 at

      Some fascinating work there Tyler.

      Most good coaches try to convince their players to keep the puck, and force the issue when they don’t have it. I.E. “It’s our puck and we are coming after it.” always trumps “Get ready, here they come!”

      This is some solid analysis to back that notion up.

      (Someone please explain to Mr. Simmons that this is one of the great examples of how we can turn to advanced data to help support positions or investigate anomalies.)

    3. da
      April 21, 2014 at

      Great article! It would be great to see the same breakdown of the Kings? A Corsi% machine with a consistently low GF? Would they be a good comparable? Thanks

    4. sacamano
      April 21, 2014 at

      No ire from me on this series of posts! I’ve really enjoyed them, for two main reasons:

      1) They frame the relationship between the statistics and the events in a helpful way (e.g., when you said, “It’s pretty cool that we’re getting to the part where we can use the data that the league gathers to illustrate a coach’s impact.”). Indeed! The stats help the rest of us (including other coaches) understand what it is a coach/player/team is doing and why it is successful.

      2) And more than simply framing the relationship in a helpful way, you’ve also done a nice job of pointing out WHY the Corsi% has been higher — (i.e., “The Ducks seem to have the inside winger pressuring the puck, whether by going in behind the net or by going hard to take away the far boards a lot more. The Leafs seem to have one guy chase the puck carrier behind the net while the other two forwards are setting up to back up and defend.”)

      This is the step that is often missing in these discussions: what is it that is driving the Corsi results?
      We all know this, but it seems to be forgotten on occasion: Corsi is not a “thing” that drives results — Corsi% is a useful measure of those events; it is really just a big flashing arrow that tells us where to look.

      Unfortunately, most analyses fail to follow the arrow and stop at simply calculating a player or team’s Corsi. This is fine for some purposes (e.g., sports bettors and general managers who tend to benefit from long-term data over which short-term strategy decisions, etc., get evened out in the wash) but not so useful for other purposes (e.g., players and coaches who tend to benefit from knowing the short-term strategy decisions in order to know how to face a particular opponent or what decision to make in a particular on-ice scenario). Taylor Hall was right when he noted that simply telling him his Corsi is low doesn’t tell him much beyond what he already knows (i.e., that he is playing poorly), and it doesn’t give him any guidance on what he needs to do to fix it.

      So, yeah, I’ve liked this series of posts a lot. You not only identify the atypically high Corsi after a faceoff-loss, but you do a nice job of following that arrow to identify the events that producing the results (in this case, a strategy of post-faceoff pressure vs dropping back). A textbook case of how the stats would be useful to a player or coach — by pointing to a particular event and how it is played on the ice.

    5. SeanL
      April 22, 2014 at

      Fun fact, the Capitals were actually the heaviest and the 2nd tallest team in the league this year. While it’s certainly not the same team that was in place while Boudreau was coaching, many of the same key players are around. Some of the forwards who are gone were pretty big (Knuble, Steckel, Bradley, Brashear, Arnott) or at least bigger than they’re given credit for (Semin is 6’2″, Kozlov is 6’4″).

    6. chelch
      April 30, 2014 at

      Maybe advanced stats are most useful when looking for a good coach. I’ve never been able to buy into acquiring players with good Corsi% because it’s not an individual stat. But coaches… especially the way you break down Boudreau’s OZ- Corsi% compared to the league average… I love that! Can we get more? All coaches, all situations?

    Leave a Reply

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