clock menu more-arrow no yes mobile

Filed under:

NHL Adjusted Plus-Minus Part 4, Even Strength Overall

By Brian Macdonald

This is part 4 of a series of posts about Adjusted Plus-Minus (APM). Since I've been a slacker, its been a while since the last one. So as a reminder, in Part 1, Part 2, and Part 3, we gave an intro, discussed best offensive players, and discussed best defensive players (both at even strength). This post we'll talk about the best overall players in even strength situations according to Adjusted Plus-Minus. We've been using Goals (G), Shots (S), Fenwick (F), and Corsi (C), and I'm also going to start including Expected Goals (E) now, from the stuff I posted about here. I'll also briefly discuss ridge regression, which is the regression technique I'm now using. Now for the top 5 players at even strength according to APM....

The top 5 players at even strength according to APM using Goals (G).
Player Pos Team G E S F C G60 E60 S60 F60 C60
1 Pavel Datsyuk C DET 22 13 13 12 12 0.98 0.59 0.57 0.54 0.55
2 Daniel Sedin LW VAN 15 9 11 11 11 0.72 0.43 0.52 0.52 0.54
3 Alex Ovechkin LW WSH 15 9 11 12 13 0.64 0.38 0.46 0.49 0.56
4 Zdeno Chara D BOS 15 7 6 7 7 0.52 0.24 0.22 0.22 0.25
5 Sidney Crosby C PIT 14 7 7 7 6 0.73 0.38 0.34 0.37 0.30

Reminder: S, F, and C were rescaled using league average shooting percentages at even strength and are roughly the same range as G. Typically these estimates are smaller than G though.

No big suprise that Datysuk is first, given that he was the top defensive forward, and in the top 5 in offense. Crosby was first in offense, but because of his defense and injuries, he's the fifth in this list. Notice that he is second in G60, but still behind Datsyuk.

I have started using ridge regression instead of ordinary least squares regression. Ridge regression can help when you have variables that are very correlated, which in this case means when one player spends most of his time on the ice with another player. Think the Sedin twins, or your favorite defensive pairing that has been consistently playing together for a few years.

There are two nice things about the ridge results compared to the ordinary least squares results:

  1. Lower standard errors
  2. Higher year-to-year correlations.

What are the standard errors in these estimates? For G60, they are roughly 0.25, for E60, they are roughly 0.14, and for S60, they are roughly 0.10.

Now let's sort by expected goals:


The top 5 players at even strength according to APM using Expected Goals (E).
Player Pos Team G E S F C G60 E60 S60 F60 C60
1 Pavel Datsyuk C DET 22 13 13 12 12 0.98 0.59 0.57 0.54 0.55
2 Alex Ovechkin LW WSH 15 9 11 12 13 0.64 0.38 0.46 0.49 0.56
3 Daniel Sedin LW VAN 15 9 11 11 11 0.72 0.43 0.52 0.52 0.54
4 Jonathan Toews C CHI 13 8 7 7 6 0.62 0.39 0.34 0.31 0.30
5 Zach Parise LW N.J 14 8 9 8 6 0.83 0.47 0.52 0.49 0.39

And now by shots:


The top 5 players at even strength according to APM using Shots (S).
Player Pos Team G E S F C G60 E60 S60 F60 C60
1 Pavel Datsyuk C DET 22 13 13 12 12 0.98 0.59 0.57 0.54 0.55
2 Henrik Zetterberg LW DET 6 6 11 12 12 0.25 0.26 0.49 0.53 0.53
3 Alex Ovechkin LW WSH 15 9 11 12 13 0.64 0.38 0.46 0.49 0.56
4 Daniel Sedin LW VAN 15 9 11 11 11 0.72 0.43 0.52 0.52 0.54
5 Joe Pavelski C S.J 5 5 11 10 8 0.25 0.27 0.56 0.54 0.41


Datsyuk is number one in both of those tables, and is number one in everything except Corsi. Ovechkin wins Corsi, mostly because of offense.

What about defensemen? Chara is the only one that showed up on the previous lists. So let's restrict to defensemen only, and let's sort by goals:

The top 5 players at even strength according to APM using Goals (G).
Player Pos Team G E S F C G60 E60 S60 F60 C60
1 Zdeno Chara D BOS 15 7 6 7 7 0.52 0.24 0.22 0.22 0.25
2 Brian Campbell D CHI 10 5 5 4 5 0.36 0.18 0.18 0.16 0.19
3 Nicklas Lidstrom D DET 10 3 -1 -1 0 0.35 0.10 -0.05 -0.03 0.00
4 Niklas Kronwall D DET 8 4 7 5 5 0.33 0.16 0.30 0.23 0.21
5 Ryan Suter D NSH 8 4 3 4 3 0.32 0.18 0.13 0.16 0.12

and by Expected Goals:

The top 5 players at even strength according to APM using Expected Goals (G).
Player Pos Team G E S F C G60 E60 S60 F60 C60
1 Zdeno Chara D BOS 15 7 6 7 7 0.52 0.24 0.22 0.22 0.25
2 Brian Campbell D CHI 10 5 5 4 5 0.36 0.18 0.18 0.16 0.19
3 Stephane Robidas D DAL 5 5 4 3 3 0.19 0.19 0.18 0.13 0.12
4 Ryan Suter D NSH 8 4 3 4 3 0.32 0.18 0.13 0.16 0.12
5 Duncan Keith D CHI 7 4 8 7 8 0.22 0.14 0.25 0.22 0.25


and by Shots

The top 5 players at even strength according to APM using Shots (S).
Player Pos Team G E S F C G60 E60 S60 F60 C60
1 Duncan Keith D CHI 7 4 8 7 8 0.22 0.14 0.25 0.22 0.25
2 Chris Pronger D PHI 4 3 7 7 6 0.16 0.13 0.30 0.29 0.25
3 Niklas Kronwall D DET 8 4 7 5 5 0.33 0.16 0.30 0.23 0.21
4 Zdeno Chara D BOS 15 7 6 7 7 0.52 0.24 0.22 0.22 0.25
5 Kris Letang D PIT 2 3 6 7 7 0.07 0.11 0.26 0.29 0.29


I was a bit surprised by Campbell being on these lists, given the criticism of his play relative to his contract over the recent years. But he comes out 2nd according to both G and E, and he is 7th in shots. Chara is 1st in Goals and Expected Goals, and is 4th in Shots, which suggests to me that he is the best defensemen at even strength.

Of course, that doesn't necessarily mean he's the best overall defensemen, since estimates of power play offense and short handed defense contributions can impact the rankings. We'll talk about special teams next time.

As always, comments/suggestions/clarifications/questions/answers are welcome.
-bmac