One of the statistics that I use a lot in evaluating player performance is Corsi – a measure which basically works out to a shot plus/minus on one end of the rink or the other (it includes missed shots and blocked shots).  Others prefer Fenwick (the same measure, minus blocked shots) or straight shots (if it doesn’t hit the net…) but all three show essentially the same thing: which end of the rink the puck stays in.


There are a lot of people out there who don’t think this is a fair indication of player performance.  The most frequent comment made is that players of different abilities are more likely to score on different shots, or that defencemen can somehow limit shot quality against.  These are undeniably true, but when you consider players as five-man units I think this gets watered down, and I’d argue it’s really a rather small point in comparison to the most important one – possessing the puck in the offensive zone, or keeping the opposition from possessing the puck in the defensive zone, which are both measured by the shots metrics I’ve listed above.


In an attempt to improve our understanding of the game and give fans access to the kind of information NHL teams track, fans of select teams have been tracking scoring chances all season long.  This started last season with Dennis King tracking the Oilers, but has expanded to include other clubs as well.  On Friday, Vic Ferrari of Time On Ice, who provides a lot of the scripts for and a big chunk of the brains behind advanced statistics, put together a chart comparing scoring chances to the shot ratios to see if the ratios did a good job of showing who was generating scoring chances.


Here’s the chart for the Edmonton Oilers:



To see the full-size chart, click over to Vic’s site.


The pattern there is undeniable, and shows up in the correlations.  According to Vic, the Fenwick number above has a 92.0% correlation with scoring chances, Corsi has a 90.0% correlation, and shots alone has an 89.0% correlation.


That may not be conclusive, but it’s compelling evidence.


Also of interest, to the shot quality club: the numbers that include missed shots had the highest correlation.  One would assume that the team which did a better job of getting their shots on net would have a better matchup with scoring chances (given that the team missing more shots would, logically, have less accurate shooters.  That turns out to not be the case.


This is a very strong indicator that Corsi/Fenwick and the shots ratio are important player metrics on an individual basis.

Comments (15)

  1. Also makes sense that the measure that doesn’t include blocked shots has the highest correlation with outchancing (as well as winning, as previously seen).

  2. Doogie2K: Yeah, I’m definitely starting to lean towards Fenwick as the best measure of what’s happening on the ice.

  3. Well, knowing that we automagically consider a blocked shot as “not a scoring chance” (even if it’s Ovechkin, 8 ft in front of halak, seeing his shot carom off Gorges’s ass) while missed shots and SOG can be SC, Fenwick logically should correlate more.

    But at the scale of a game, I still prefer Corsi as a zone time / zone split proxy. Even at the level of a player it’s pretty good, I think. Fenwick, for those duties, could sometimes brush over some pretty significant amounts of data.

    But over the course of a season? I’m agnostic, really.

  4. Olivier: Good points, and I think Corsi’s the best indicator of zone time. I suspect that as a metric of five man unit talent, however, the fact that Fenwick gives credit for blocking shots probably makes it a little better.

    I’m not sold on that, but that’s how I’m leaning.

  5. I think Corsi should correlate better with the abilities of a forward, while Fenwick should correlate better for defenseman, as for the most part shots from farther away have a better chance of being blocked so Fenwick does somewhat compensate for where the shot is taken from and can give Defensemen credit for keeping teams to the outside

  6. Also, Jonathon, do you have a full list Fenwick ratings for this year by any chance?

  7. @JW: I think that’s exactly what Vic found by correlation. I know he found Corsi was the closest indicator of zone time, and I’m pretty sure (but can’t find) that he also found Fenwick had a stronger association with winning percentage than Corsi.

  8. Anybody else notice that 5-6 months ago Phoenix played, Well like Phoenix then the NHL takes over the team and Bam,they start winning left right and center. Now im not taking anything away from the Coyotes they have great goaltending and afew good players,What i’am saying is the officiating this year is brutal,Now i wonder if the ref’s think “Hey the Redwings are playing againts the leagues team{Phoenix}Ohh oh i better make them happy after all- this is my job”…We all see bad calls, goals taken away, penalties given, or not given, look at game 6 the second fight with Abdelkader, Det gets the extra 2min “why” or right at the end of the game- a slash to the back of the legs on a detroit player No call,sure there was only seconds to go but a call is a call in any game not just this one.Bad calls make for bad hockey ,veiwers dont want to see their team get eliminated from the playoffs especially by a ref making the wrong call,I guess we’ll see im right with my thoery come game 7–Good luck Detroit….

  9. Johnathan: Fenwick is a better consideration of play.

    But the block is an important change of the stats. It is a Save! people do not get that. to me it is still a shot. Would it have gone in if no one saves it. I fthe answer is yes then it is a save! Forward Save (block), Dman Save (block) Goalie save (save)

    When rating a goalies Save percentage we should be using the shot position chart to get a cumulative rank for the goalie Save % using context rated scoring values. Basketball does this! Why not us?

    Ex. A wrist shot from the slot on edge of the left circle yeilds a goal xx% of the time. A slap from the point on a screen yeidsXX%.

    Then the block becomes more Valuable. A Dman who blocks 5 xx% opportunities has made a diffrence! The dman has made the save.

    Yes shot quality is important. Making a save on it counts whether they are dmen, forwards or goalies.

  10. Interesting. From what I’ve read,the correlation to team points for Corsi (.626) and Fenwick (.602). That’s from TPSH.
    In other words,slightly better than flipping a coin (ok,it’s much better than flipping a coin but still an incomplete metric). My contention with Corsi and Fenwick is that they conceptualize ‘scoring chances’, I prefer sticking to definitive events like scoring an actual goal.
    A basketball style shot chart with numerical values assigned to the various zones* is the answer-ie shots from the slot are more valuable than those from the point 5v5 (I would assume but maybe not,after all; how many point shots result in goals due to juicy rebounds?)- although no single metric will give you the complete picture. Get on it.

    *split the offensive zone into 3 vertical zones (point-blueline to top of circle,slot-circle,crease-bottom of circle to icing line) and 5 lateral zones (wide left,left,middle,right,wide right) and determine the scoring% from each zone over the last 5 years to get the default values for each zone.
    Any shot coming from any zone will then have a value that can be used to measure an individual’s scoring abilities. Taken in conjunction with Coris or Fenwick, I’d think you’d get a decent picture of any individual players contributions. Even if that doesn’t hold, I’d still like to see shot charts in the NHL.

  11. @rickithebear You sir,are a genius. I didn’t even read your post before I posted mine, I took about 20 minutes trying to come up with a general ‘formula’. Kudos,great minds think alike.

  12. Good Lord. The Oilers finally send someone out to read this stuff, and he may very well not be clever enough to understand the complete foolishness of the two posts above. Normally I would stay silent and enjoy watching the silliness play out. But here we have an opportunity.

    S: These folks above may well be nice people, but they are childlike and simple (and don’t want to argue with either of you).

    Your competitors are a light year ahead of you (and me for that matter), I can tell by their online actions at IOF and TOI. In fact i’ve learned a lot just by watching their focus shift.

    Who have you guys contracted? Blueline (Melloy is sharp), Coleman?, Bain? Be careful, you’ll likely make yourself worse.

    The business of the NHL is trivial compared to the business of NHL wagering. Still, I’d like to see the Oilers do well. My hope is that you hire in someone like Keenan, MacTavish or Murray. Then bring in a competent (but not brilliant) Bayesian mathemetician to turn their thinking into math.

    Best of luck

  13. Yeah,you’re right igor. Who needs a offensive zone probability grid when you have bayesian math…

  14. ricki/gill:

    What I’m getting at is that ‘scoring chances’ is a measure of shot quality – it only tracks shots that have a reasonable chance of scoring. And if it’s saying almost the same thing as a straight shots count, that means that over time the difference in quality of chances evens out (for the most part), and that the remaining difference is pretty small.

    Now, I imagine this varies from team to team (the Lemaire system, for instance, almost undoubtedly has an impact on shot quality) but for players on the same team we should be able to use Fenwick – or the other shot metrics – as a proxy for scoring chances.

  15. JW- I wasn’t really arguing against Corsi or Fenwick, I understand the measurements but I question,what is the ultimate goal of these two metrics? To measure individual players performance in game situations shift by shift and segmented by the different manpower situations ie-5v5. That’s how I understand the two variables. Feel free to correct me if I’m wrong or missing something there.
    For measuring an individual player, I think that goals are important and I really don’t believe in luck. For breaking down goal scoring, I like shot charts. Basketball uses them to evaluate players abilities,track for performance and team tactics. Hockey and basketball share a lot of similarities, the shot chart should translate well.
    I’d like to see if goals tend to be scored from certain areas of the ice (in general and for specific teams/goalies) and if teams generate more shots from certain location as opposed to others and,if teams allow more shots from certain locations as opposed to others etc…
    While I think that recording the data would be a ball breaker; I think that kind of information could be useful for measuring players, teams, styles, possible outcomes…
    Anyhow,I don’t intend to try and write a novel here; I was just stoked that someone else thought shot charts would be cool.

Leave a Reply

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