Scoring chances, Corsi, and why Mikhail Grabovski is great

When I first joined The LeafsNation, I had a few people asking questions as to whether I’d be counting scoring chances or not for the team next season. The gut reaction was “yes” because it’s always good to have a bit more data to analyze, but there’s really no point.

Eric T. of Broad Street Hockey, and recently NHLNumbers, came out with a piece today about scoring chances and Corsi that can be found here. For those unfamiliar with the concept of Corsi, please read on.

Corsi was a metric developed by Jim Corsi, a goaltending coach for the Buffalo Sabres. Corsi doesn’t stand for anything and it isn’t a real complex concept. Corsi could more appropriately be written out as “shot attempt differential”. 

It measures the amount of shots directed at the opponent’s net while a player was on the ice, including blocked shots and missed shots, and subtracts those directed at the player’s defensive net.

You can express the stat as a rate per sixty minutes, as in, “Mikhail Grabovski had a Corsi per 60 of 8.9 at 5-on-5 last season” which would mean that every 60 minutes of ice-time, the Leafs would fire between 8 and 9 more shots at the opponent’s net than his own, meaning Grabovski is a “plus-possession player”. 

If you subtract Grabovski’s teammates’ combined Corsi when he’s off the ice from that 8.9 number, you’re left with his “relative Corsi” which is a better way to compare players between different teams.

Corsi can also be expressed as a ratio as well. “When Grabovski was on the ice, 53.5% of all shot attempts were directed at the Leafs’ opponents’ net“. This also tells us he was a “plus-possession player”.

We say “plus-possession” because a former Edmonton Oilers blogger named Vic Ferrari found that teams that get a lot of shot attempts also spend a lot of time in the opponent’s end. Teams that spent a lot of time in the opponent’s end were also found more likely to make the playoffs. If Grabovski has a 53.5% Corsi ratio, there’s a strong chance that the Leafs were on offence for 53.5% of the time Grabovski was on the ice, and if every Leaf could “drive play” as well as Grabovski, there’s a strong chance the Leafs would make the playoffs.

The basic criticism of the Corsi metric is “well, it doesn’t account for the quality of shot”. As it happens, a few team bloggers did undertake the effort to count each and every quality shot for an against their team over a full season. Eric today showed that the correlation between players with good Corsi rates and quality shot rates was absurdly high:

Whatever tendency certain players might have for driving their team to get more scoring chances than a simple shot differential predicts is small and swamped by random noise. This suggests tracking scoring chances isn’t adding much information to the readily available shot differential numbers.

So, no, I won’t count scoring chances for the Toronto Maple Leafs the traditional way we’ve done it. Enough teams have done been recently and the results that leave no indication that any one player or team can consistently and continually create more quality shots without creating a higher quantity of shots.

  • MediumWell

    Thanks for doing this — haven’t been able to find as comprehensive a breakdown of how Corsi works anywhere else. It was driving me insane.

    And you have to wonder if Lupul isn’t in the trade discussion because the Braintrust is afraid of shaking the chemistry he has with Kessel. Rightly or wrongly, I think it’s something to consider when both have career years.

  • “The basic criticism of the Corsi metric is “well, it doesn’t account for the quality of shot”. As it happens, a few team bloggers did undertake the effort to count each and every quality shot for an against their team over a full season. Eric today showed that the correlation between players with good Corsi rates and quality shot rates was absurdly high”

    I don’t think the work that Eric did proves or disproves whether shot quality exists. To me, the way to do that is to look at whether shooting percentage are repeatable and when we do that with sufficient data we find that it is. I found that r^2 of 2007-10 on-ice Sh% to 2010-12 on-ice Sh% is 0.2854 and I also found that 2007-10 SH% predicts 2010-12 goal rates (r^2 of 0.2531) as 2007-10 fenwick rates (r^2=0.2441) are. To me the fact that Eric showed a high correlation between scoring chance and and corsi data tells me that the scoring chance counting process isn’t refined enough to identify the existence of shot quality.

      • Danny Gray

        Because his value will never be higher. He’s a great story and everything but chances are he doesn’t produce like that again and he will be a UFA next summer and likely too expensive for the Leafs given his age.

        He could be moved to help fill the other holes in our roster.

        • Danny Gray

          I agree to a certain extent, but MacArthur is also UFA at the end of this year and will be due a raise as well. They’ll both be looking for a raise in the $1-2 million dollar range from their current salary. I’m guessing close to $1.25.

          This is a decision that should be made later in the year, closer to the deadline. Hopefully after initial talks have started about an extension, so we have an idea what player’s are looking for in terms of salary and term. Personally I think Kadri could more easily step in and take MacArthur’s spot and fill it well. In time I think he’d actually be better.

          With Armstrong, Connolly and Lombardi (combined $11.25 million) all UFA next season and likely not back we have more than enough to give Lupul a raise of $1-1.25 million. Keep the term short because of injury history if possible (3-4 years? sounds reasonable). We’d still have more than enough to go after a coveted UFA next season. Getzlaf?

          Lupul-Getzlaf-Kessel
          JVR-Grabovski-Kadri
          Kulemin-Bozak-Frattin

          Easy.