Even Strength Player Ratings for this season

Yesterday I presented my even strength player ranking system and rankings for last season which clearly created some controversy. Hopefully when I present this years rankings I can clear up a bit more of the concerns that people have with the system. I know some people found some of the results difficult to believe but if I can show some consistency from year to year I think that should help show the value that this rating system can have.

So first let me start off by showing the top 40 rated players from this season.

Player Team Offense Rating Offense Rank Defense Rating Defense Rank Overall Rating Overall Rank
PIERRE-MARC BOUCHARD Minnesota 1.89 23 5.13 2 3.51 1
KEITH YANDLE Phoenix 0.48 417 6.18 1 3.33 2
MATTHEW LOMBARDI Calgary 1.75 34 4.68 5 3.22 3
JAROSLAV MODRY Dallas 1.63 52 4.72 4 3.17 4
DANIEL SEDIN Vancouver 2.31 6 3.90 6 3.10 5
STEPHANE ROBIDAS Dallas 0.79 326 4.75 3 2.77 6
NIKLAS HAGMAN Dallas 1.36 101 3.90 7 2.63 7
MIIKKA KIPRUSOFF Calgary 1.66 47 3.36 10 2.51 8
NIKITA ALEXEEV Tampa Bay 2.14 12 2.62 16 2.38 9
BRIAN ROLSTON Minnesota 1.72 36 2.92 12 2.32 10
ALEXANDER PEREZHOGIN Montreal 1.71 38 2.75 15 2.23 11
BRETT MCLEAN Colorado 0.90 276 3.49 9 2.20 12
SCOTT HARTNELL Nashville 1.05 216 3.04 11 2.04 13
ILJA BRYZGALOV Anaheim 0.33 444 3.75 8 2.04 14
THOMAS VANEK Buffalo 2.60 2 1.35 93 1.98 15
ROBERT LANG Detroit 1.72 37 2.19 22 1.96 16
OWEN NOLAN Phoenix 2.51 3 1.26 115 1.88 17
KEITH TKACHUK St. Louis 1.22 146 2.53 18 1.87 18
JASON WILLIAMS Detroit 1.35 106 2.35 19 1.85 19
SIDNEY CROSBY Pittsburgh 1.84 27 1.83 34 1.84 20
ROBERTO LUONGO Vancouver 2.32 5 1.36 89 1.84 21
RADEK MARTINEK NY Islanders 0.89 280 2.79 14 1.84 22
MARK RECCHI Pittsburgh 1.02 226 2.61 17 1.82 23
PAVEL DATSYUK Detroit 2.04 15 1.58 55 1.81 24
STEVE SULLIVAN Nashville 2.79 1 0.83 303 1.81 25
MIKE JOHNSON Montreal 1.27 131 2.35 20 1.81 26
SCOTT GOMEZ New Jersey 2.21 7 1.32 101 1.77 27
TOM PREISSING Ottawa 1.89 24 1.57 58 1.73 28
DANNY MARKOV Detroit 1.57 64 1.84 33 1.71 29
STEVE MCCARTHY Atlanta 0.50 409 2.85 13 1.68 30
MARC-ANDRE FLEURY Pittsburgh 2.15 11 1.15 150 1.65 31
MIKE DUNHAM NY Islanders 1.11 193 2.17 23 1.64 32
SAMI SALO Vancouver 1.19 164 2.06 26 1.62 33
NICLAS WALLIN Carolina 2.51 4 0.71 364 1.61 34
WOJTEK WOLSKI Colorado 2.01 18 1.22 123 1.61 35
MICHAEL NYLANDER NY Rangers 2.19 8 1.04 191 1.61 36
MILAN MICHALEK San Jose 1.69 41 1.49 71 1.59 37
SAMUEL PAHLSSON Anaheim 1.27 132 1.92 29 1.59 38
NICKLAS LIDSTROM Detroit 1.80 30 1.35 94 1.58 39
MICHAEL CAMMALLERI Los Angeles 1.29 127 1.86 31 1.57 40

The first thing one will notice is that the ratings are much higher than those for last season. This is because we are working with much less data this season than last so one bad or good game will make a huge difference. This is not ideal because essentially it means we may not have enough data to reliably evaluate players. For example, I don’t expect Keith Yandle to remain at the top of the list as the season goes on but I must also say that to play over 100 minutes of even strength ice time with the Coyotes and only have one goal scored against you is pretty impressive as Phoenix is a horrible defensive team.

So, it is still early in the season to rely too much on these rankings but I think there is some value in comparing players with their values of last season to see if there is much consistency showing up. Here is a list of each players overall ratings comparing last years with this years.

Gelinas 1.70 last year vs 1.43 this year
Donovan: 1.67 vs 0.85
Zetterberg: 1.66 vs 1.05
Thornton 1.64 vs 0.81
Cheechoo 1.64 vs 0.51
Nylander 1.64 vs 1.61
Schneider 1.63 vs 1.53
Armstrong 1.61 vs 1.43
Jagr 1.57 vs 1.15
Selanne 1.56 vs 1.18
A. Markov 1.55 vs 1.27
Weber 1.55 vs 1.05
Legwand 1.54 vs 1.29
Shanahan 1.53 vs 1.1
Sakic 1.51 vs 1.21
Komisarek: 1.45 1.17
P. Bergeron: 1.45 vs 1.17
Kalinin 1.44 vs 0.92
Crosby: 1.44 vs 1.84
P. Boucher: 1.40 vs 0.85
Morrow: 1.38 vs 0.87
Rozsival: 1.38 vs 0.96
Ponikarovsky: 1.36 vs 1.31
Gionta: 1.36 vs 1.51
Zhitnik: 1.35 vs 0.98
Pothier: 1.35 vs 1.36
Malik: 1.35 vs 1.31

Clearly there are some significant differences (Donovan, Thorton, Cheechoo, Boucher, Kalinin, Morrow, Zhitnik) but clearly many of them haven’t gotten off to good seasons. Thornton and Cheechoo haven’t gotten anything going yet. There are many similarities too. Most other top rated players from last season are still rated well above the 1.00 mark. Last years highest rated player in Gelinas is also fairly highly rated this year. Brian Pothier who changed teams has an almost identical rating. It might also be worth pointing out that Chara also has an almost identical rating this year compared to last year (0.72 vs 0.82).

While we are on the Senators lets take a look at their defensemen from last season more closely. Here are last seasons and this seasons overall ratings.

Chara 0.72 0.82
Redden 1.22 1.05
Meszaros 0.95 0.73
Phillips 0.85 1.16
Pothier 1.35 1.36
Volchenkov 0.98 1.07

Pretty good consistency really as Phillips is the only one who really changed their ratings significantly. But what I really wanted to point out that in the playoffs last year most people said that Redden and Pothier were Ottawa’s best defensemen and Chara and Meszaros were big disappointments. In fact, Chris McMurtry gave Redden and Pothier a B grade, Chara an F and Meszaros a D when he handed out grades on each players playoff performance. Phillips got a C+ and Volchenkov a C. I find it quite interesting that Chris’s analysis agrees almost perfectly with my ranking system. Maybe 5-on-5 (which is generally more important in playoff hockey) Chara isn’t the stud defensemen everyone thinks he is and a good skating, puck moving defenseman like Pothier is what matters most. Certainly Chara joining the Bruins hasn’t done a lot for them in the standings.

Once this years ratings stabilize more (say in a month or so) I’ll do a more formal analysis but what I see so far is that there is some consistency in ratings from year to year.

This article has 7 Comments

  1. While you still have not explained the guts of your player rating system, its evident that you do not have enough games to produce anything meaningful (assuming this algorithm produces something meaningful with enough games played – which is at this point unproven), I think its clear that the highest rated player is a fluke. Its some guy who has played a role on a team but hasn’t been on the ice when they allowed goals and wsas when they scored them. Given 30 teams leaguewide somebody will fit that description whether it is PM Bouchard, Keitzh Yandle, Martin Gelinas or Shean Donovan.

  2. Yes, it is still early and if Yandle gives up goal on his next shift his defensive rating will drop about in half. So yes, there really isn’t enough data yet to draw any real conclusions about a guy like Yandle.

    But Bouchard’s ratings from last year were 1.09, 1.51, 1.30 which put him in the top 10% of qualified players last season so while Bouchards rating is currently over infated the fact that he is at or near the top of the list isn’t a shocker.

    I think its clear that the highest rated player is a fluke. Its some guy who has played a role on a team but hasn’t been on the ice when they allowed goals and wsas when they scored them.

    At what point does it not become a fluke? At what point do we have to conclude that Gelinas being on for 40 goals his team scored and just 13 goals opponents scored is not a fluke. At what point do we decide that maybe Gelinas’ ability to play hockey is a factor? Over the course of the whole year the Panthers scored 168 even strength goals. Gelinas was on the ice for 40 of them (~24%) over less than 50% of their games. Lets say they scored 60% of their goals from January 18th on. That would be 101 goals. Gelinas would have been on the ice for nearly 40% of Florida’s goals scored. Jokinen was only on the ice for 38 goals from Jan 18 on and Jokinen played 12% more even strength ice time. At what point does this not become a fluke?

  3. When the laws of sample size catch up with the data. Obviously Gelinas had a terrific season last year, but I suspect that both the goals he was on the ice for and the lack of goals against were both well above expectation and not necessarily due to his actions. I think Greg is correct in saying that there is bound to be such a player somewhere in the league for a given season, and that 82 games are not enough to draw such conclusions.

    There’s not sufficient evidence to prove that Gelinas as a LW is preventing a significant number of goals simply by his presence – and given his career +/- I suspect this is true. Correlation does not imply causation – at least not with the sample sizes we’ve been dealing with here.

  4. Are expected goals = sum(ice time with player *average scoring rate * players score) over all players in the NHL
    where player score is the above score… (so you can score at a rate 6x below average against Yandel and be average) So if you play against a lot of good defensive players or with better players expected to get more plusses.

    You can then iterate this until it comes to an equilibrium?

  5. JavaGeek, yes, you could do that though I am not sure you would come to equilibrium. I’d have to think about that more. I do something similar but not exactly the same formula you use. I don’t currently iterate but I am trying to come up with methods that would further isolate a players value and one of them would involve iteration. The one thing I worry about with iteration is the potential to increase, not decrease, the error if not done correctly.

Comments are closed.