© 2013 Michael Parkatti Baby_Sheep_by_MickeyIW

Season Player-with-Player Matrix

I’m working on an exercise to optimize the Oilers’ line and pair matchings to maximize Corsi for the team.  The first step was to compile player-with-player Corsi percentages for the season to date — to be included, two players had to play at least 9 minutes of ice-time and play more than 5 games (except Brown).  I’d thought I’d post this while I’m working to give you a sense of what goes into this process.  This table will drive my mathematical programming to figure out who to play with who.  Click on it to enlarge.



I’m also listening to my iTunes library while I work.  This song just came on, and is helping to deaden the pain:



  1. Michel
    Posted March 9, 2013 at 3:33 pm | #

    Very interesting chart. A few things I noticed:

    – MPS appears to be a fit on almost every line
    – The opposite seems to be true of Hemsky this season
    – A lot has been made about the team missing Horcoff, and justifiably so, but it appears as though Haartikainen has also been a bigger loss than advertized
    – Every player on the team saw their corsi lowered when sharing a line with Lennart Petrell…

  2. Michel
    Posted March 9, 2013 at 3:42 pm | #

    What do you think of the following lines based on the data?
    L1: RNH-Hall-Eberle (all three appear to make each other better)

    L2: Gagner-Hemsky-MPS (MPS pairs well with Gagner, and he pairs well with Hemsky. Given that Gagner and Hemsky paired well with each other in the past, hopefully this line would perform well if given more opportunity).

    Bottom 6 pairings:

    Those two pairings appear to work well together. Until Horcoff and Hartikainen come back from injury I’m not sure any third man on either line will be of much use, so just try to keep those pairs together and rotate the extra help based on matchup.

    Also, it seems as though the Petry-Smid defensive pairing would be better served in the #1 role again so they can get more mins with the #1 fwd line. Takes some pressure off of J.Schultz too while he gains experience.

    Anyhow, looking forward to seeing what you come up with Michael. Thanks for the post! Always enjoy reading your blog.

  3. PutzStew
    Posted March 9, 2013 at 7:36 pm | #

    I’m new to this, hate advance stats but this make sense.
    Do you have anything for the totals from past years.

    Michel. Why would you not trade Hemsky and put Yak in that position?


    • Michel
      Posted March 10, 2013 at 4:32 am | #

      PutzStew: I’m just assuming that the goal of this exercise is to try to optimize the assets currently on the roster. If we could suggest trades, I would want to massively revamp the entire bottom 6.

      Long term, I would think that Yakupov’s development plan with the Oilers is to eventually have him succeed Hemsky on the second line once he improves on his positioning.

      It’s important to consider that this exercise does not consider offensive zone starts. Yakupov has been much more sheltered this season with his zone starts slanted towards the offensive zone as compared to Hemsky who is deployed in a much more balanced manner. I would assume that means guys will generally show better shot diff numbers with Yakupov as it means they were more likely in the off zone when paired with him than when they were paired with Hemsky.

      I would think that variable also partially explains Petrell’s atrocious numbers. He has taken draws on a 70/30 basis in the defensive zone – more than anyone else on the team. SO I’m sure that skews his results to the negative. With that said, he split draws 50/50 last year against weak competition and was still a – player so I would think his numbers would still look poor, just not to the same extent.

  4. Woodguy
    Posted March 10, 2013 at 7:19 am | #

    You should make any square that is under 40% black to represent two things:

    1) The gravitational force of sucking
    2) The despair we feel when those players are on the ice


    I’ve been looking at all the Oilers’s WOWY’s.

    MPS is only behind Hall in terms of making team mates better.

    I know there are other factors at work in those numbers, but its something to see.

  5. Sean
    Posted March 10, 2013 at 1:12 pm | #

    Thanks Michael. It boggles the mind that 91 was scratched twice… EVERY winger on the team plays better with him around. I think the season was over as soon as Horc went down which is pretty damning. Is there another team in the league that would be as cripple by the loss of their #3 center?

    The bright side: I don’t think it would take more than 2 weeks for a good GM & coach to make this a >50% corsi team.
    The dark side: I strongly suspect we won’t have a good GM & coach any year soon.

  6. Posted March 15, 2013 at 10:18 pm | #

    I actually just started working on the same thing for the Sharks, but for Defensive pairings. One thing I noticed quickly was the influence of sample size. I think it would be extremely prudent to regress Corsi% based on TOI to the mean. For 5v5 data, my estimated r = .50 was at 180min. Thus to regress any TOI, %regression = 1-(TOI / (TOI + 180))

Post a Comment

Your email is never published nor shared. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>