Wednesday, March 3, 2010

NHL Goalie Productivity at the Trading Deadline

In a article that is forthcoming in the Journal of Sports Economics, David Berri and I have a paper evaluating NHL goalie talent (and how GM's evaluate NHL goalies), so given today is the NHL trade deadline, my thoughts turn back toward NHL player evaluation. One of the things that GM's must be able to do is to evaluate the value of playing talent in making a trade. In today's salary cap world another issue is to make sure that the team is able to stay within the cap. What I am providing here is a quick measurement of the productivity of NHL goalies, by looking at how productive they are in terms of wins produced above average (WAA).

Here are all the NHL goalies that have played this 2009-2010 NHL season by the NHL trade deadline (through March 2nd) from highest WAA (Wins Above Average) to lowest WAA. I define each of the statistics after the table.

Player Team WAA SOG GA GAA Saves Save Percent
Tomas Vokoun FLA 5.1905 1717 119 2.36 1598 0.9307
Ryan Miller BUF 4.5815 1597 111 2.16 1486 0.9305
Evgeni Nabokov SJS 4.0278 1674 123 2.29 1551 0.9265
Miikka Kiprusoff CGY 3.2828 1565 118 2.18 1447 0.9246
Jimmy Howard DET 3.2483 1350 99 2.28 1251 0.9267
Craig Anderson COL 3.0578 1709 131 2.43 1578 0.9233
Henrik Lundqvist NYR 2.0999 1583 126 2.38 1457 0.9204
Tuukka Rask BOS 1.8668 829 61 2.15 768 0.9264
Jaroslav Halak MTL 1.7284 966 74 2.56 892 0.9234
Jonas Hiller ANA 1.7209 1468 119 2.70 1349 0.9189
Roberto Luongo VAN 1.6987 1449 117 2.35 1332 0.9193
Ilya Bryzgalov PHX 1.5719 1544 127 2.37 1417 0.9177
Ty Conklin STL 0.8810 595 47 2.61 548 0.9210
Johan Hedberg ATL 0.7947 920 76 2.54 844 0.9174
Mike Brodeur OTT 0.7326 87 3 1.00 84 0.9655
Justin Peters CAR 0.6541 93 4 1.34 89 0.9570
Jason LaBarbera PHX 0.6433 331 25 2.23 306 0.9245
Martin Brodeur NJD 0.6194 1544 133 2.33 1411 0.9139
Tim Thomas BOS 0.5548 999 85 2.52 914 0.9149
Peter Budaj COL 0.5286 252 19 2.21 233 0.9246
Anton Khudobin MIN 0.5004 48 1 0.87 47 0.9792
Yann Danis NJD 0.4639 146 10 1.85 136 0.9315
Antero Niittymaki TBL 0.4060 1012 87 2.66 925 0.9140
Semyon Varlamov WSH 0.3385 477 40 2.44 437 0.9161
Marty Turco DAL 0.3282 1329 116 2.68 1213 0.9127
Chris Mason STL 0.3267 1323 115 2.52 1208 0.9131
Thomas Greiss SJS 0.3095 304 25 2.47 279 0.9178
Cam Ward CAR 0.3013 1220 106 2.74 1114 0.9131
Michal Neuvirth WSH 0.1861 464 40 2.75 424 0.9138
Dan Ellis NSH 0.1812 734 64 2.59 670 0.9128
Chad Johnson NYR 0.1583 135 11 2.35 124 0.9185
Brent Johnson PIT 0.1131 458 40 2.69 418 0.9127
Carey Price MTL 0.1042 1124 99 2.76 1025 0.9119
Cory Schneider VAN 0.0328 59 5 3.80 54 0.9153
Alexander Pechurski PIT 0.0233 13 1 1.67 12 0.9231
Andrew Raycroft VAN -0.0179 291 26 2.37 265 0.9107
Matt Zaba NYR -0.0898 16 2 3.53 14 0.8750
Jhonas Enroth BUF -0.1107 37 4 4.14 33 0.8919
Antti Niemi CHI -0.1169 542 49 2.26 493 0.9096
Nikolai Khabibulin EDM -0.2227 602 55 3.03 547 0.9086
Dustin Tokarski TBL -0.2428 16 3 4.09 13 0.8125
Rick Dipietro NYI -0.3534 201 20 2.60 181 0.9005
Michael Leighton PHI, CAR -0.3633 693 64 2.69 629 0.9076
Wade Dubielewicz MIN -0.3703 18 4 4.14 14 0.7778
Alexander Salak FLA -0.3788 40 6 5.37 34 0.8500
Josh Harding MIN -0.4155 421 40 2.79 381 0.9050
Dwayne Roloson NYI -0.4462 1206 110 2.86 1096 0.9088
Joey MacDonald TOR -0.4697 157 17 3.20 140 0.8917
Jonathan Quick LAK -0.5668 1532 140 2.50 1392 0.9086
John Curry PIT -0.5795 14 5 12.50 9 0.6429
Erik Ersberg LAK -0.6855 175 20 2.99 155 0.8857
Manny Legace CAR -0.7105 443 44 2.94 399 0.9007
Marc-Andre Fleury PIT -0.7150 1364 126 2.65 1238 0.9076
Brian Elliott OTT -0.7252 1069 99 2.65 970 0.9074
Patrick Lalime BUF -0.7259 287 30 3.20 257 0.8955
Ray Emery PHI -0.7727 783 74 2.64 709 0.9055
Jose Theodore WSH -0.8268 993 93 2.94 900 0.9063
Jean-Sebastien Giguere TOR, ANA -0.8319 729 70 2.99 659 0.9040
Curtis McElhinney CGY -0.9568 235 27 3.23 208 0.8851
Brian Boucher PHI -0.9832 414 43 2.84 371 0.8961
Martin Biron NYI -1.1034 577 58 3.24 519 0.8995
Steve Valiquette NYR -1.1726 128 19 3.74 109 0.8516
Mathieu Garon CBJ -1.2007 683 68 2.90 615 0.9004
Scott Clemmensen FLA -1.2471 361 40 3.58 321 0.8892
Mike Smith TBL -1.3017 898 88 2.99 810 0.9020
Cristobal Huet CHI -1.4032 968 95 2.31 873 0.9019
Alex Auld DAL -1.4973 558 59 3.00 499 0.8943
Ondrej Pavelec ATL -1.6105 1111 109 3.45 1002 0.9019
Chris Osgood DET -1.6485 524 57 2.94 467 0.8912
Pekka Rinne NSH -1.7225 1074 106 2.81 968 0.9013
Jonas Gustavsson TOR -1.7813 862 88 3.07 774 0.8979
Devan Dubnyk EDM -1.9510 251 35 4.17 216 0.8606
Niklas Backstrom MIN -1.9999 1247 123 2.74 1124 0.9014
Pascal Leclaire OTT -2.1079 670 73 3.02 597 0.8910
Jeff Deslauriers EDM -2.1838 1142 115 3.21 1027 0.8993
Steve Mason CBJ -2.7515 1239 128 3.10 1111 0.8967
Vesa Toskala TOR -3.8991 676 85 3.66 591 0.8743

Where SOG = shots on goal; GA = Goals Against; GAA = Goals Against Average; Saves is well the number of shots on goal that were stopped by the goalie; and Save Percent = Saves/SOG.

WAA is our measure of NHL goalie productivity is based on the absolute value (since a goal against has a negative effect on team wins) of the marginal value of a goal against divided by two (since each win is worth two standings points) times the number of shots on goal that goalie faces times the difference in the save percentage of the goalie and the average save percentage of all goalies for that season (or in this case up to the trading deadline).

5 comments:

Jeff said...

Whoa. Niklas Backstrom is that far down on the list? Does this mean Josh Harding should get more starts for the Wild?

Hawerchuk said...

hi Stacey,

Why not use replacement level for these goaltenders? Mike Brodeur can't be more valuable than Martin Brodeur.

Here's a very simple model I proposed:

http://www.puckprospectus.com/article.php?articleid=298

Stacey said...

Jeff,

One of the main findings of our research is that goalie performance is highly variable, and we wonder why teams are paying so much for such variability. That said, from the model, I would play Harding more than Backstrom, although both are having poor years.

Gabriel,

I like that you have adjusted for 5-on-5 vs. 4-on-5, where our model does not. That said one of the things that I do not like is arbitrary picking a cutoff, as opposed to one that is based on actual performance. Hence the model compares each goalie to the league average save percentage and then calculates their performance based on the coefficient from the regression on standings points and goals for and goals against to get the value of keeping a goal out of the net on the team's standings points to make the playoffs.

Mike Brodeur has only played 3 games, but has played them excellently. His performance in playing those 3 games will likely not continue, and his productivity will change with more playing time. But in terms of a GAA of 1.00 and a save percentage of 0.9655, that is better than Brodeur's numbers through 59 games this year, according to our model.

Hawerchuk said...

Stacey,

Being "league average" is very valuable. Martin Brodeur has played 58 games and has been worth perhaps 4 wins to his team this season. Mike Brodeur had three wonderful games. He contributed 1 win to his team relative to a replacement goalie.

Replacement level is not some arbitrary line. It is the level of performance available for the league-minimum salary. This is not a new concept; using the league average as the baseline caused you to come up with the wrong result.

Jeremiah said...

Stacey,

Is there any way to read your article, without purchasing it for $25?

I am a die-hard hockey fan, and I've recently become addicted to learning about economics. I just finished with some economics classes, post-BA.

Unfortunately, I am not an institution and I can't afford the $25, but I would LOVE to read about your analysis! The goalie position is my favorite to analyze.

If you know of a more affordable way for me to view your study, please contact me!

Cordially,

Jeremiah

You can email me at:
miahs_world@hotmail.com

Thanks again!