A Corsi-centric Discussion of Advanced Stats



Justin Bourne and I got together to discuss a few things in the realm of hockey analysis recently. The full thing can be found here at Backhand Shelf, but here’s a taste of the back and forth:

Kent Wilson:

Perhaps we could discuss this from your latest “Thoughts on Thoughts” piece: “Guys who lose a lot of puck battles end up with poor Corsis, and with the (slight) rise in the prevalence of that stat, I think those guys are going to be more exposed than ever.” Is this something you intuited or is it something that’s been studied? I, personally, assume there could be a relationship there and in fact have thought it would be great if we could track puck battles to see if they correlate with possession (like the way Eric did with zone entries).

That said, I’m not totally convinced this is necessarily true, particularly if the player has other well developed skills that may compensate for poor puck battling abilities (on the flip side, I think it’s possible to have a lousy corsi even if you’re strong in the trenches due to poor skating, puck play etc.). Related topic – the rise of corsi. Is that your assumption based on it’s prevalence in independent circles or have you heard whispers that some in the league are paying attention to it?

Justin Bourne:

I made the comment that guys who lose a lot of puck battles will have poor Corsis from my own experience as a player. When you have a linemate battling for a puck down low, and you’re in your position, you’re praying he comes out with it. If he doesn’t, you’re headed back on defense. So many times I remember personally getting bumped off pucks, and being partly pissed because I wasn’t going to get an offensive opportunity, but mostly because I knew I now had to skate the length of the ice (hard), and would likely be spending the shift in the d-zone.

It’s pretty clear to me that losing puck battles will negatively affect your Corsi in a significant way. I do agree that it’s possible to compensate for lost puck battles with other skills, specifically by making good decisions with the puck. (Though, you’d have to be making *great* decisions with them in the neutral zone to make up for something as damaging as losing battles.) One thing I hope people who follow Corsi know, is that it’s mostly a decision-making metric, at least in my mind. Guys who turn the puck over a lot have their Corsi hurt. Guys who can’t figure out a way to get the puck out of their D-zone have their Corsi hurt.

It’s largely mental. And I don’t follow the stat crazy-close, I just think I have a good grasp of the things that dictate the direction of the puck. Something like good skating can help you win a puck race or two and improve your Corsi, but that happens a time or two a game. You probably touch the puck and need to make a decision 20 times a night…

We go on to talk about the rise of possession and advanced metrics in mainstream coverage and why people might find some of them intimidating. Again, the rest can be found at Justin’s place.


  • MC Hockey

    Nice article…is it supposed to be that short? Props to the Bourne Identity…drum roll and then sad trombone for the patheticness of the joke. Sorry bad humor is part of the lockout.

  • RexLibris


    Perhaps its just me, but I seemed to infer from the conversation that there is developing an analytical separation between winning the puck battle and size of the player. The emphasis on intelligence and decision-making skills as opposed to the traditional big-body-hard-to-knock-off-the-puck line of thought.

    I remember watching J.F. Jacques, a big body if there ever was one, losing a lot of puck battles because he simply wasn’t at the same level of game comprehension as his opponent. Linus Omark, on the other hand, was often referred to as being “surprisingly strong on the puck”. Surprisingly because he was small, but he made the right decisions and positioned his body such that, in some cases, when the battle was lost a penalty was drawn.

    Corsi might be a sort of entry-level drug for further statistical analysis, opening the gateway into other ways of examining a player’s strengths and weaknesses and what they can provide to a team by further breaking down how that player operates in certain situations.

    A stat within a stat – music to Brian Burke’s ears.

    MacTavish used to say “don’t let the play die on your stick” and this seems to be echoed by Justin Bourne.

    A very interesting conversation, thanks for posting it.

  • This post is noteworthy because Bourne’s explanation of why he makes the association between puck battles lost and corsi goes some way to defeating the claim itself. He explains that he hates it when other players on his line lose a puck battle in the offensive zone. But wait: it’s the other guy that is losing the puck battle, yet its Bourne’s corsi that suffers just as much as his presumably suffers. This is evidence against the claim that a guy’s own puck battles are associated with his corsi.

    To repair the damage one might weaken the claim by subbing “Guys who lose a lot of puck battles” to “Guys whose lines lose a lot of puck battles”. Then you’ve got yourself a case. But you’ve also taken all the focus off the individual you were supposed to be evaluating in the first place.

    • That’s more a criticism of corsi itself – indeed, of any stat in hockey, since every result you generate is necessarily going to be somewhat dependent on the guys you are playing with.

      That said, that’s the reason we try to collect a lot of data, we do things like WOWY (with or without you) analysis and why I argued we’d need to correlate puck battles with other measures to ensure they are meaningful.

      • Quite right, but if this is the best that can be said in favour of the rise in the significance of corsi, the excitement is probably unwarranted.

        My own crazy view, FWIW, is that there is no such thing as an individual’s ability/skill (full stop) to be measured by advanced or counting stats. Ability/skill is always relative to an environment. It’s not something that inheres in an individual. This necessarily limits the reliability of statistical measures we are all familiar with because and insofar as they presuppose otherwise. WOWY is a notable exception, but only goes part of the way to remedying the problem.

  • I find it interesting that teams would go to the trouble to collect extra data on things like puck battles and turnovers and not have the extra intrigue to test and make sure it means something. I mean, I get it…they carry into it an assumption, that it’s meaningful on its own, but I don’t get the scenario where you don’t, at some point, become curious to investigate just how strong the effect might be, or whether it can be predictive.