• Dealing With People Who Cite Corsi% And Single Games

    by  • December 7, 2013 • Hockey • 3 Comments

    If you’re a guy who talks about numbers, you get a lot of this:

    I assume that it was something similar that prompted Eric Tulsky to write “Advanced Stats Proponents Are Not Idiots”.” Let’s take the criticism that my friend @kissel87 came up with.

    To start with, he’s misunderstanding the thrust of the whole Corsi% thing. Score effects are a well known phenomenon – we’ve been discussing it here since November of 2008. Teams that are leading in the third, or leading by a couple of goals earlier, tend to get outshot attempted. There is buckets of evidence of this.

    Over the course of a season, this washes out to an extent. In a single game, a team can easily play 70% or more of the 5v5 TOI in a game in a situation that isn’t close. Over the course of the season, not true.

    A way to deal with this is to look at the Corsi% when the game is close – within a goal in the first two periods or tied in the third. What if we ask ourselves about records for teams when winning the shot attempt fight with the score close and then when losing the shot attempt fight with the score close? We get a table like this.

    The bottom line says it all to me – teams that win the Corsi% when the score is close are on a 97 regulation point per 82 game pace. Teams that lose it are on a 67 point pace. If your standard is that something needs to tell you with 100% certainty who’s going to win the game, well, this isn’t it. If you’re willing to consider information that merely provides considerable insight, you might like this.

    Why do some teams seem to be doing better in games in which they lose the Corsi% close fight? In addition to special teams, which aren’t considered in Corsi% close, there’s the issue of percentages. I went through and gathered the goal data for teams when the game was close and sorted it based on whether or not the team won the Corsi% when the score was close. It’s revealing.

    First of all, if you look at the line at the bottom, you’ll see that there’s effectively no PDO edge when the score’s close. There’s a difference of two hundredths of a point. Collectively, all of the difference in goals between teams that win the Corsi% and lose the Corsi% when the score is close is the edge in shots.

    If you look at the first table, you’ll note that eleven teams – from the Panthers down – haven’t really enjoyed a nice winning percentage cushion in games in which they’ve won the Corsi%-close fight. You’ll note that all of those teams have enjoyed better PDO in games in which they lost the Corsi%-close fight. There’s your explanation: they’ve had better percentages in games in which they’ve been outshot so far. You’d be stupid to bet on that continuing, given what we know about the regression of shooting and save percentage and the fact that, on a league level, there’s essentially no difference.

    Understanding all of this prevents you from saying stuff like this:

    I assume he’s done his arithmetic correctly and that that’s true but it misses the point: many games in which a team outshoots the other team, it does so because it is desperately trying to get back in the game and the other team is playing it safe, chipping pucks in and not chasing goals.

    The sooner a wide swath of people grasp this, the less irritating my mentions on Twitter will be. In the mean time, feel free to cite Corsi%-close and score effects to these people and direct them to this post.

    Email Tyler Dellow at tyler@mc79hockey.com


    3 Responses to Dealing With People Who Cite Corsi% And Single Games

    1. OIlanderp
      December 7, 2013 at

      The coming war with the Cylons will make people understand the logic of Corsi%-close. But then, of course, it might be too late.

    2. Pingback: [Gameday Bits] Real Hexed Guys of Orange County | All New York Islanders

    3. Front7
      December 11, 2013 at

      That’s the country singer Brett’s older brother.
      Remember him from school. Not surprised.

    Leave a Reply

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