• Seriously, Points Aren’t Offence

    by  • July 8, 2014 • Hockey • 19 Comments

    I had an exchange yesterday with everyone’s favourite favouriter on Twitter, @tsetse_fly.

    A little context. IPP (individual point percentage) is a stat that measures how many points you get per goal scored when you are on the ice. Forwards tend to be clustered around 72% in the long run, with stars a little higher and scrubs (and the occasional non-scrub, like RNH) somewhat lower. It’s generally believed that wild deviations from that 72% figure tend to be chance and if a guy has a year with an unusually high or low point number, this can be a reason why.

    So, for example, I wouldn’t bet on Taylor Hall’s points remaining as high as they’ve been over the past two years if the Oilers don’t start putting up drastically better possession numbers. Hall’s IPP over that time is 96.6%.

    Defencemen are a little more muddled. The guys you think of offensive defencemen (which we’ve been taught means “defencemen who get points”) tend to be pretty high up in terms of IPP, with numbers in the high 30s and 40s. Off the glass and out types tend to be lower down.

    I’ve been kind of beating the “points aren’t offence” drum with defencemen for a while. Essentially, my argument is that judging a defenceman’s offence by the points he produces is flawed, something that I don’t necessarily think is true for forwards. If a forward is putting up points, I’m inclined to think that there are a lot of goals being scored when he’s on the ice. With a defenceman, I don’t think that that’s necessarily the case.

    I figured I’d test this by pulling together two quick graphs that show the relationship between points/60 and goals for on/60 at 5v5 for forwards and defencemen. For those of you who aren’t familiar with looking at relationships by way of a graph, the closer the graph is to a line (whether up, down or straight), the tighter the relationship is between the two variables.

    Screen Shot 2014-07-08 at 1.32.08 PM

    Screen Shot 2014-07-08 at 1.32.19 PM

    It is, I think, pretty obvious that this relationship is a lot tighter for forwards than it is for defencemen. A cautionary note: this doesn’t mean that points are necessarily a good measure of a forward’s offensive contribution – play with Crosby (the dot that’s almost off the forward graph) get more points, probably through nothing you’ve done. It does mean that we can draw an inference from seeing a high or low points total that the guy was probably on the ice for many or few goals for.

    With defencemen, that’s not so much the case. There are defencemen who get points despite not seeing their team score a particularly noteworthy amount of goals and defencemen who don’t get points despite seeing their team score lots when they’re on the ice. This might be easier to see if I re-cast this in terms of a percentage – D scoring as a percentage of the average D in my group and the goal for on rate as an average of the defencemen in my group.

    Screen Shot 2014-07-08 at 2.06.53 PM

    The red dot is Erik Karlsson. His point rate is 81% above the average for defencemen from 2007-14. The Sens scored just 7.2% more goals than the average defenceman saw his team score when he was on the ice though. Forwards like that don’t really exist. If I sort the forwards by their PTS/60 as a percentage of the average, Nikolai Zherdev has the highest rate of points relative to average with a percentage increase in his team’s GF rate that’s lower than Karlsson’s – he scored 23.8% more points than the average player and his team put up 6% more goals than the average team.

    I should note that the Senators shot just 7.3% at 5v5 with Karlsson on the ice over the seven years that I’m looking at. The average is about 8.0%. Still, you’d assume that if more pucks went in, Karlsson would get more points and the issue would remain. That may not be a safe assumption but that’s a question that teams (and analysts) will be getting into once we have SportVu data available.

    I want to be clear that I’m not saying that Karlsson is a bad offensive player and that I think he does other things that make him an exceedingly valuable player – he’s rarely in his own end. At the same time, I don’t think that the relationship between points and offence works for defencemen in the same way that it does for forwards. If you run an NHL team, this is a kind of troubling thing to have to wrap your head around because players – even defencemen – get paid for points. This rests on the assumption that points equal offence. It’s not a safe assumption to make.

    Email Tyler Dellow at tyler@mc79hockey.com


    19 Responses to Seriously, Points Aren’t Offence

    1. Triumph
      July 8, 2014 at

      Solid piece. I think something about Karlsson – he gets 7 shots/60 on 5 on 5, which is a lot. Over the last three years, Byfuglien (who played some at forward this year) is the only guy over 7, then it’s Souray, Krug, and Yandle above 6, and everyone else is below 6 shots/60. The thing about a D taking so many shots is that they have to be lower percentage than forward shots, and indeed you have to get pretty low on the list to start finding guys with forward-high shooting percentages (the highest, incidentally, is Justin Schultz, who is shooting nearly 12% 5 on 5).

      I imagine the culprit, points-wise, is the dreaded secondary assist. I wonder if there’s some causal relation here – teams who score more points by D may tend to score them in transition, and teams who tend to score a lot in transition aren’t good Corsi teams, and therefore don’t score as much in general? Something to consider, I guess.

      • Stephan Cooper
        July 8, 2014 at

        Stupid autocomplete, can the above be deleted?

        Similarly, you can notice that Karlson historically tends to do much better by Corsi than other metrics (Fenwick, straight shots, goals). I would make sense that running an offense through a defenseman, even one that plays as deep as Karlsson does is going to involve some trade-offs, one of which is a large proportion of your attempts at offense end as shots from a lower percentage area.

      • buan
        July 10, 2014 at

        Would it be interesting to remove secondary assists from the analysis? I think I’ve read a piece form Eric T. about secondary assists having much less correlation from season to season. Getting rid of these fluctuations might help in clearing the picture. Just a quick thought.


    2. Triumph
      July 8, 2014 at

      Oh, duh, that’s part of the solution – the critcal thing with defensemen is that they end up playing with everyone. So Karlsson had 1620 minutes at 5 on 5, but 200+ of those minutes were with Chris Neil and the rest of the 4th line. Whereas someone like Crosby does appear to have some double shifts here – 81 minutes with Tanner Glass, for instance – he still doesn’t have to play with the entire forward roster. Crosby spent over 80% of his minutes playing with Chris Kunitz, who may not be Jari Kurri but he’s certainly better than Erik Karlsson’s most common linemate (Marc Methot). So yeah, I’d be curious about how top D do with top line forwards and how they affect one another’s scoring rates. I imagine you will get different results once you eliminate D minutes played with 3rd and 4th lines.

    3. John
      July 8, 2014 at

      Equal points definitely doesn’t mean equal offense for defensemen, but he was still first in the league in CF/20 among defensemen with at least 300 5v5 TOI as I’m sure you know so he’s definitely creating a lot of offense. Also just wondering, do you think Vlasic is a top 5/10 defenseman in the NHL?

    4. July 9, 2014 at

      One thing that’s interested is that lone dot at 1.1, 2.5 on the forward graph. I looked through the numbers and found that that’s Brent Burns. That’s very interesting because over the 7 season period you looked at, he was a defenseman in Minnesota for 4 years, a Dman in San Jose for 1 year, and then a forward on Thornton’s line for 2 years.

      So his Pts/60 are more like that of a Dman (skewed by all that time on defense) but his G/60 are more like that of a forward on a high scoring line, (Last 2 years, his G/60 was 3.32 [8th best in the NHL] because he played on San Jose’s top line.), but not quite there as they’re being pushed down by his time spent on defense.

      Basically Burns himself serves as a nice litmus test for your hypothesis and I think only strengthens it.

      • Triumph
        July 9, 2014 at

        That’s fine work in finding out who that is, but that serves to bolster my hypothesis – Brent Burns the defenseman has to play with all kinds of dross – here’s his most common linemates in those 5 years – Nick Schultz, Mikko Koivu, Andrew Brunette, Antti Miettinen, PM Bouchard, etc. – but still, he spent only 1/5th of his ice time with Koivu, 3/8th with Schultz. Brent Burns the forward spent 80% of his time with Joe Thornton. So while it is true that forwards’ P/60 correlates more with GF/60, there’s a recursiveness built in with forwards – they are usually playing with other good forwards – that is simply not true of defensemen.

    5. Grant Bronk
      July 9, 2014 at

      Not a stats wonk – lets start there. But, I stopped immediately at your comment regarding Hall and his 96.6 IPP. You stated that it’s likely it won’t stay there unless the Oilers improve their possession numbers. It seems more logical to me that it works in reverse, and that it is precisely because the Oilers have such bad possession numbers (have trouble cycling and their D is annually graded as a D) that drives that number. This is because all their offense goes through Hall, and until there is a supporting cast, that is likely to continue. This tells me that Hall is unbelievably valuable to a bad team that has not surrounded him with much support. Once he gets that support, and others contribute to the creation of offense through possession those numbers will then start to trend toward the mean.

      • MaxPower417
        July 9, 2014 at

        His comment was that Halls POINTS won’t stay at their current level unless the Oilers improve their possession numbers. The IPP is certainly going down no matter what happens.

        If you’re going to stop reading an article because of a sentence, make sure you reread that sentence to make sure it’s the writers fault and not the readers.

        • Grant Bronk
          July 9, 2014 at

          I’m afraid I’m not so into ‘blame’ as you are. Just visiting, making a comment and looking for hockey discussion. Obviously we’re a little defensive ’round these parts. No need for it, but some folks are like that.


          • MaxPower417
            July 9, 2014 at

            Okay bud

            ” But, I stopped immediately after X” is pretty clear way of saying “I thought I’d check out your site, but it’s clear you don’t know what the fuck you’re talking about”.

            If you were looking for a reasonable discussion, that’s really not the way to kick things off.

            • jqstave
              July 10, 2014 at

              Nah man, you’re off base. I didn’t read it like that at all. Chill out and don’t be so defensive. You didn’t even write the article. He was just asking a question.

    6. Pingback: Salary arbitration dates; Hawerchuk on Kane; hockey on a navy ship (Puck Headlines) - FourLeagues

    7. EthrDemon
      July 10, 2014 at

      I have a feeling that the looser correlation for point v goals-on for defensemen might be partially explained by the higher likelihood of a defeseman getting a second assist for a goal which he was no longer on the ice for.

      (Or to rephrase, the lower average elapsed time for a forward between touching the puck and a goal being scored)

      • swelrod
        July 16, 2014 at

        Does anyone have any data on the frequency of this occurring? I can’t imagine that second assists are given out long enough prior to see this too often. It absolutely happens, but I wouldn’t think enough to create much noise.

    8. zach frey
      July 10, 2014 at

      obviously the correlation with a defenseman’s IPP is going to get weaker than a forward’s–probably about half as strong, given the 30 vs. 70 ratio of forwards factoring in goals to defensemen. there still, undoubtedly, has to be a correlation, though. a great offensive D is going to factor into a higher percentage of the goals than his weaker counterpart. it’s just not as strong a correlation–muddled, as the author noted. lots more error and room for deviation will cause the defenseman’s IPP to fluctuate more statistically just for the simple reason of smaller sample size. different breakout and offensive strategies on different teams will also swing a D-man’s likelihood to factor into the points sheet wildly.

    9. Biff Morenz
      July 11, 2014 at

      Yes, points aren’t offence and those goals pre-Corsi and Fenwick we saw in the Air Hockey ’80s wasn’t offence. Oh brother. Dial it back a bit. I’ll take the guy who does zip all game and scores on the two shots he takes over the David Booth has the puck in the offensive end and skates his way into a corner.

    10. Pingback: Salary arbitration dates; Hawerchuk on Kane; hockey on a navy ship (Puck Headlines) | NHLRealTalk

    11. jpm79
      July 11, 2014 at

      Are there two types of players we think of when we say offensive defencemen? One with strong corsi that succeed in moving the puck and possession up the ice to the forwards (who factor in most of the recorded scoring) and another that do more on their own (like Karlsson ) and have similar out of whack numbers.

    Leave a Reply

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