Roles and Stats Part V: Roles and Defensive Stats – Defensemen

This is part five of my series on roles and stats. In part four of the series I looked at how roles relate to the offensive statistics of defensemen and found that there was a relationship however it was not nearly as strong as for forwards (as one would expect). Today I will look at roles of defensemen and defensive statistics which one would think ought to show a stronger relationship. Let’s have a look.

(If you have not already, it is recommended to at least read part I of the series where I discuss methodology before proceeding).

Leading and Trailing TOI%


Ok, this is a bit unexpected. There doesn’t seem to be much if any relationship between a defenseman’s role and limiting goals against. It is surprising because we saw a relatively strong relationship for forwards.


There is some indication here that more defensive roles are somewhat correlated with more shots against which seems odd for defensive defensemen to do. However we know that many defensemen used in shutdown roles (Phaneuf, Weber, etc.) actually have poor corsi ratings so maybe this isn’t all that surprising. That said, there isn’t much of a compelling story to tell here.


Not much here either. This is somewhat surprising.

Defensive and Offensive Face Offs


I am going to go through these quickly because there really isn’t much to say. There isn’t much interesting to read from these charts.


Again, there is some indication that defensive roles lead to more shots against.


However defensemen don’t seem to be able to influence save percentage relative to their roles.

Penalty Kill and Power Play


Same as above.




By now you know the story.


This is a bit surprising. There is very little relationship between defensive ability of defensemen and their roles. It is surprising because we see some relationship between roles and offensive statistics for defensemen and between roles and defensive statistics for forwards. However for some reason roles have no relationship with defensive statistics for defensemen. One theory I have on this is that offensive defensemen are often paired with defensive defensemen more so than for forwards. Think Erik Karlsson and Marc Methot or Morgan Rielly and Matt Hunwick. Brent Burns played mostly with Paul Martin last year. A few years ago P.K. Subban played a lot with Josh Gorges. It’s possible defensive pairs are put together and used in more balanced ways than forward lines and that might be what we are seeing in these charts for defensemen.

I want to take a moment to reflect on Matt Cane’s Hockey Graphs article on the subject. In it he calculated correlations between year one and two statistics with year three and four statistics. He looked at nine separate correlations with future save percentage statistics for forwards and defensemen separately. Matt Cane shows p-values for his statistics and he pointed out that only four of the 18 regressions had a p-value less than 0.05. All of these were correlations for forwards. What is even more interesting is that of the 9 forward regressions all but one had a p-value equal to or better than 0.10. Conversely, for defensemen 5 of the 9 regressions had p-values above 0.50 and only one had a p-value below 0.30 (a p-value of 0.16). It is interesting to see that even Cane’s work finds stronger relationships for forwards than for defensemen.


One comment

  • Derrick

    In addition to your supposition that D pairs are made to be more balanced (whereas forward lines tend to have well defined offensive/defensive roles) I wonder if it may be due to just the raw number of available bodies.

    With 12 forwards on the bench most times it gives them the ability to slot them into specialized roles. Perhaps one line is a pure offensive scoring line that gets to play more in the O-zone with heavy PP time, another is a traditional shutdown line who takes on a more defensive role with heavy D-zone starts and work on the PK. Then they have some guys who are more two-way forwards that sometimes slot in with one of those units depending on where they are needed. And with all that they still have enough bodies available that the 4th liners can be sheltered and only used in situations in which they will not be a liability (many are PK specialists who get lots of D-zone time, but some get heavy O-zone time because the coaches don’t trust them as much), which also means they are more likely to sit when in a leading/trailing situation (although sometimes with a huge lead it is the superstars who get a break while the scrubs get more ice time).

    Defense on the other hand only has 6 bodies to work with, which means it is much more difficult to shorten the bench and use them is specific roles. When guys are playing 25-30 minutes a night they are going to have to play both ends of the ice, help out on PK even if they are more suited towards the PP, and be used both to defend a lead and to mount a comeback. Even the 3rd pairing guys can’t be sheltered the same way 4th line forwards can be. They may not get PP time and might spend less time on the PK than others, but the top pairing can’t play the entire game so they still wind up having to take shifts when leading and trailing and in both zones. So that may have something to do with why there is seemingly no rhyme or reason to these charts, because the D really don’t have the leisure of playing in predetermined roles.