Tuesday, February 21, 2012

Scott Gomez Modified Corsi

I recently watched the film Moneyball - in it, the essence of sabermetrics and finding efficiencies in the 2002 baseball market was reduced to three words: Get on base. Baseball has come a long way since then - on-base percentage has come to be correctly valued by the market. I also re-read the book Moneyball has a chapter devoted to the concept of DIPS (Defense Independent Pitching Statistics). Voros McCracken found that aside from home runs, walks, and strikeouts, pitchers have no control over how many hits they give up. However, that's not quite true - some pitchers have a little better control and it manifests itself over a career. It's not worth all that much - maybe a few hits over a season, maybe a run or two a year - but some pitchers give up a few more hits and some a few less hits than we would expect by just assuming that everyone gives up hits at the same rate. I got to thinking about applying this sort of winnowing to hockey - as the years go by, we're accumulating more and more data, now it's time to really look at some of it.

There's been some amazing work done on shaping Corsi/Fenwick into something more intelligible than it is in its raw form; I don't need to list off every blogger in this field. It's not like these concepts are being ignored. So someone may well have done this before, but I'm not consciously stealing their material.

Now, to hockey - I'm a Devils fan, and so I watched Scott Gomez play for 8 years. I didn't really understand shooting percentage when I watched him, but now I do, and I can say unequivocally that Scott Gomez is a horrible shooter. Hockey-reference.com finds only one forward with more than 1000 shots on goal and a career shooting percentage below his - Henri Richard, for whom we only have shot data for the final 8 seasons of his career. Bear in mind that New Jersey, the place where he played the majority of his career, tends to undercount shots - there could be between 25 and 100 saved shots that went uncounted over the course of his career there, driving his shooting percentage even further down.

Yet I see bloggers all over praise Gomez's Corsi and his potential value even as his shooting percentage slides further into the toilet. For me, the reason why Gomez's Corsi has always been high is pretty simple - he loves to shoot the puck from all angles, and many of his shots are extremely low-percentage shots that are merely an attempt to generate a rebound that a teammate can score on. Keep in mind that before I conduct this experiment, I think that Gomez's Corsi will not be affected very much at all - it's still going to be quite good. Still, we have to account for the fact that he is a terrible shooter, and that we finally have enough data to be pretty darn sure that his presence on the ice lowers shooting percentage.

Now unfortunately for ease of data collection (and the fact that behindthenet has been spotty the last couple of days), I am only closely examining data from between 2007-08 to 2009-10. timeonice.com presents the data I wanted in a much easier format for me to compile. I'll spare you the raw data and just present a simple chart: shooting percentage with Gomez on the ice at even strength versus shooting percentage with Gomez on ice with Gomez's shots removed. In all 3 years, his teammates actually shot better than he did while on the ice with him. This doesn't seem that remarkable, but we have to remember that 2 of his teammates are always going to be defensemen, who typically shoot much at a much lower percentage than forwards.

YearS% With GomezS% W/O Gomez Shots

Now unfortunately to complete the post, I had planned on getting more shooting data so I could compare Gomez's shooting rate to an average forward both in terms of shots on goal and in shooting percentage at ES, but I can't find that data anywhere. It would be a haphazard look anyway - what I will do instead is look at Gomez's Shots % while on ice with Shots For normalized such that 8% of shots were goals, as that is generally considered to be the NHL average at even strength.

YearShots % ActualShots % Modified

Is this 'correct'? No - obviously in 2008-09, Gomez's teammates were just as much of a letdown as Gomez himself, as the shooting percentage doesn't move much with Gomez's shots taken out. It looks like that year he was certainly unlucky, in addition to being a poor shooter.

We don't have quite enough data included to make any sort of sweeping statement here. We know that the Gomez trend continues - according to behindthenet, Gomez himself shot 3.5% at 5 on 5 in 2010-11, and while he was on the ice, the Canadiens shot 4.7% as a whole. This year, he has 0 goals in 38 5 on 5 shots.

And yes, before I get a deluge of complaints, I am ignoring the effect that additional shots have - an increased chance of drawing a penalty, more offensive zone starts, not being in the defensive zone, and so forth. I'll leave that to someone with more comprehensive data. All I'm saying is that with nearly 5 seasons worth of data, and a close examination of 3 seasons worth of it, we have to account for the fact that Gomez is only a +2 player at even strength despite having 406 more shots than his opponents while on ice, and that much of it has to do with Gomez's horrific shooting percentage. Corsi and its brothers are something, but not everything - we mustn't be completely wedded to such a proxy for what's actually going on.


  1. Here's a chart of all of Scott Gomez's even strength shots last season:


    You can reproduce that chart for other seasons and players as you'd like. It looks like most of his shots are from reasonably good areas, so I don't think bad shot selection is a good explanation for his SH%.

  2. That's a cool chart.

    I'd get upset if I weren't the one saying this, but while Gomez's shots are from good areas, I'm not convinced they're good shots. Either A: he's just a bad shooter B: he's intentionally not trying to score at times or C: while his shots appear to be from good areas, the lack of context isn't giving us enough information about how good these shots are. I suspect it's some combination of A, B, and C.