Team-Adjusted GAETAN

by red army line on Jun 11, 2011 11:37 AM EDT

Goal Asymmetry Experienced by Team Above Normal. Here's GAETAN in two sentences:

GAETAN tells us how many more or fewer goals a player's team scored and allowed while he was on the ice than we would have expected from an average player. To calculate GAETAN, we first determine how many goals for and goals against would have occurred while an average player played the same number of minutes as a particular player, and then we compare those numbers to the actual goals for and goals against that occurred while the particular player was on the ice.

Using Gould Old Days' method, instead of rating players against league averages, I wanted to rate players against their team averages, to correct (at least partly) for the quality of the team for which a player players on.

This method isn't a perfect adjustment for the team a player plays on--a great player on a great team may look worse than a good player on a terrible team--nor does it account for quality of competition and some other situation effects, but I feel that it's a step in the right direction, like relative plus minus as opposed to raw plus minus.

In order to get team averages, I simply used team power play, penalty kill, and even strength goal differentials per minute (most of which Gould Old Days was kind enough to provide in his spreadsheet in the above link). These values only differ because of penalties drawn to taken values (for obvious reasons). I used those team-adjusted values to calculate expected goals for and goals against values while the player was on ice. I did not account for players having changed teams, however. Each traded/waived-and-claimed player was judged against his initial team. I did not make the assumption that players do not affect shooting percentages to any significant degree and base this off of shots as opposed to goals, moreover, so shooting/goaltending luck (along with any shooting percentage skill) is included. I did not adjust team situational values for goaltending skill, moreoever, since it is so difficult to estimate.

Here is a link to the spreadsheet. You'll notice that I listed GAETAN versus league-wide rates, GAETAN versus team rates, and GAETAN (team) per game. For the rest of this post, "tGAETAN" refers to team-adjusted GAETAN.

Top ten forwards in the league, sorted first by descending tGAETAN, then by tGAETAN per game, then alphabetically:

									League GAETAN	Team GAETAN
Player	Team	Pos	GP	ES TOI	PP TOI	SH TOI	Team GF	Team GA	+/-^X	'+/-^X	'+/-^x per game

Mikhail Grabovski	TOR	C	81	1232.6	254.6	81.1	93	57	19.48	31.15	0.38
David Backes	STL	R	82	1250.9	221	144	106	66	32.67	30.31	0.37

Daniel Sedin	VAN	L	82	1216.8	296	8.6	130	44	58.64	28.01	0.34
Jonathan Toews	CHI	C	80	1257.4	247.1	156.7	116	67	40.39	27.97	0.35
Ryan Getzlaf	ANA	C	67	1192.7	240.1	31.3	111	58	33.12	26.29	0.39
Michael Grabner	NYI	R	76	960.4	69.6	116.3	64	49	19.45	26.19	0.34
Loui Eriksson	DAL	L	79	1249.9	267.7	107.4	95	57	22.74	26.12	0.33
Sidney Crosby	PIT	C	41	656.8	222.5	19.4	78	33	25.67	26.03	0.63

Anze Kopitar	LAK	C	75	1192	272.3	154.8	94	58	24.82	23.10	0.31
Henrik Sedin	VAN	C	82	1271.9	294.8	12.6	133	52	54.14	22.77	0.28

(Sorry the last three columns do not line up perfectly)

The "Care Bears" finished just outside the top ten. Going a bit further down the list, Anaheim's super group five of Perry-Getzlaf-Ryan and Lydman-Visnovsky shows up early and the Sedins are both near the top. Meanwhile, "Toews for MVP" was probably a more legitimate argument than I'd given it credit for.

Crosby's tGAETAN rate was in a league of its own. He ranked first among players with at least 40 games played, with the next highest as Getzlaf's 0.39. Among players with at least 30 GP, Ryan Whitney was at 0.55 and Grant Clitsome as 0.41.

And top ten defensemen:

									League GAETAN	Team GAETAN
Player	Team	Pos	GP	ES TOI	PP TOI	SH TOI	Team GF	Team GA	+/-^X	'+/-^X	'+/-^x per game
Lubomir Visnovsky	ANA	D	81	1596.4	336	35.3	146	76	41.38	31.62	0.39

Toni Lydman

ANA

1391.6

290.8

25.21

28.03

0.36

Tomas Kaberle

TOR, BOS

1456.1

333.1

23.9

100

11.57

23.37

0.29

Scott Hannan

COL, WSH

1325.8

9.3

208.3

-0.06

21.29

0.27

Andrew MacDonald	NYI	D	60	1078.2	183.1	143.6	80	65	11.24	20.92	0.35
Ryan Suter	NSH	D	70	1382.2	240.3	141.6	92	55	27.60	20.74	0.30

John-Michael Liles	COL	D	76	1346.9	245.1	80.7	94	75	3.35	20.29	0.27
Dustin Byfuglien	ATL	D	81	1553	326.9	7.8	121	81	9.63	20.06	0.25

Anton Babchuk

CAR, CGY

1088.1

210.3

42.1

15.99

19.86

0.24

Ryan Whitney

EDM

641.1

151.2

94.6

6.61

19.34

0.55

We see three defensemen that saw mid-season trades land them on better squads. I think that is most of the reason why Kaberle, Hannan, and Babchuk finish so high. Meanwhile, keep an eye on Andrew Macdonald among other young Islanders blueliners (Ty Wishart has a strong showing despite a downgrade in team in moving from Tampa to Long Island, and I bet Travis Hamonic would have a strong showing as well, but for some reason he is absent from the spreadsheet).

Shea Weber is noticeably absent, even though his partner shows up...

Caps leaders:

League GAETAN

Team GAETAN

Player

Team

Pos

ES TOI

PP TOI

SH TOI

Team GF

Team GA

+/-^X

'+/-^X

'+/-^x per game

Alexander Semin	WSH	L	65	904.6	208.8	61.4	74	33	26.97	21.04	0.32
Alex Ovechkin	WSH	L	79	1330.3	354	3.6	110	49	27.65	20.99	0.27
Nicklas Backstrom	WSH	C	77	1218.9	256	111.2	93	50	29.22	20.49	0.27
John Carlson	WSH	D	82	1484.8	181.2	190.8	93	66	27.91	15.33	0.19
Brooks Laich	WSH	C	82	1076.2	245.7	188.7	74	47	21.57	12.16	0.15
Mike Knuble	WSH	R	79	1119.4	183.8	109.4	66	45	13.92	5.43	0.07
Mike Green	WSH	D	49	900.9	217.9	115.6	58	40	8.26	1.32	0.03
Sean Collins	WSH	D	4	56.6	0	2.4	3	2	1.23	0.80	0.20
Brian Willsie	WSH	R	1	6.3	0	0	1	1	0.00	-0.04	-0.04
Marcus Johansson	WSH	C	69	868.5	65.9	81.5	45	40	6.48	-0.52	-0.01
Keith Aucoin	WSH	R	1	5.9	5.9	0	0	0	-0.56	-0.56	-0.56
Eric Fehr	WSH	R	52	553.1	101.6	0	30	18	2.33	-0.70	-0.01
Karl Alzner	WSH	D	82	1436.5	30.6	174.2	60	62	11.67	-1.22	-0.01
Brian Fahey	WSH	D	7	77.4	4.4	1	3	4	-1.32	-1.83	-0.26
Andrew Gordon	WSH	R	9	71.5	6.6	0	4	5	-1.63	-2.06	-0.23
Jeff Schultz	WSH	D	72	1243	4.8	176.2	53	60	9.32	-2.50	-0.03
Jay Beagle	WSH	R	31	286.7	3.2	35.7	7	10	0.09	-2.52	-0.08
DJ King	WSH	L	16	90.1	0.9	0	3	6	-3.09	-3.68	-0.23
Mathieu Perreault	WSH	C	35	362.8	52.7	0.3	19	16	-1.99	-4.07	-0.12
Tom Poti	WSH	D	21	303.5	28.7	53.4	16	20	-1.65	-4.57	-0.22
John Erskine	WSH	D	73	982.9	3.9	95.7	36	41	3.74	-4.71	-0.06
Matt Bradley	WSH	R	61	631.1	5.2	3.4	20	23	-3.17	-7.40	-0.12
Tyler Sloan	WSH	D	33	401.2	0.2	11.1	11	19	-6.96	-9.85	-0.30
Matt Hendricks	WSH	C	77	763	23.6	95.8	33	43	-3.13	-9.99	-0.13
Boyd Gordon	WSH	C	60	644.2	4.4	134.3	19	38	-6.63	-13.62	-0.23
Jason Chimera	WSH	L	81	955.3	48.8	69.6	38	49	-9.02	-16.47	-0.20

Thanks to the Caps' below-average power play and above-average penalty kill, players who played a little less on the power play jumped ahead of players who played more, like Alexander Semin ahead of Nicklas Backstrom. Towards the bottom of the list we continue to see tough minutes and terrible players. Note that Schultz, Erskine, Bradley, and Hendricks are, in part, so low because they played so many minutes.

I think you could make a case that Semin was the best forward on the Caps this past season. I doubt any GM will be buying that story, though, unfortunately.

Here is how tGAETAN compares with GAETAN for players with at least 20 games played:

via i987.photobucket.com

The x-axis is GAETAN, the y-axis is this tGAETAN. I like that the range of y values (~70) is less than the range of x values (~100). Even though this concept is not the same as that underlying goals-versus threshold (GVT), the highest GVTs on record belong to Dominik Hasek, in the low-to-mid 50 goals-above-replacement, and the top value in GAETAN this past season was Daniel Sedin at nearly 60, which, considering Hasek's GVT, I find little high for my liking, personally. I think Visnovsky's 31 in tGAETAN sounds a little more realistic in terms of painting a picture of player value, although I'm not about to start championing Visnovsky-for-Hart anytime soon (or ever, probably).

That's all for the time being. Any suggestions or ideas?

If this FanPost is written by someone other than one of the blog's editors, the opinions expressed in it do not necessarily reflect those of this blog or SB Nation.

Tweet Comment 6 comments | 10 recs |

Do you like this story?

Comments

Two comments:

1) The baselines are set using the whole team, correct? So the baseline everyone’s compared against for tGAETAN includes the player himself? Seems that would suppress tGAETAN for good players (since they’re being compared against themselves rather than just their worse teammates) and elevate tGAETAN for bad players (again, since they’re being compared against themselves rather than just their better teammates). This would be an effect that would show up much more in tGAETAN than regular GAETAN.

2) Could you explain why the GAETAN and tGAETAN numbers come apart dramatically for some players (Schultz and Alzner, for instance) but not others? It seems like that’s telling us that in these cases, given the ES/PP/PK TOI, the league average player might not do nearly as well, but the team average player would do better. Now compare Alzner to Mike Green; Alzner has higher GAETAN but lower tGAETAN. Is that telling us that Alzner is playing in roles that the team can better replace (PK probably) than the roles Green is playing? And vice versa in the case of Green?

Awesome work, thanks.

"...what're you gonna do?"

by Rainbow, Kitty, Beer on Jun 11, 2011 3:46 PM EDT reply actions

1) Yes, I would need to subtract out. I didn’t have the GF/GA breakdowns by manpower, though (Gouldie listed totals only), so I didn’t subtract out this time. That is the logical next step, though.

2) Alzner compared to the league was above average on the PK, but on the Caps he was slightly under average. Green was above average both ways. Green was average as a Caps PP D but Alzner, in his small sample, wasn’t too good.

Here’s the breakdown, rounding. I couldn’t find 4v4, 4v3, 3v3, and 5v3 data, so I assumed neither Alzner nor Green had been on ice for a goals against 4v3 or 5v3 or a goal for 3v4 or 3v5.

Alzner
GAETAN XGD: -13
actual: -2
tGAETAN XESGD: +9 (act: +13)
XPPGD: +3 (act: 0)
XSHGD: -13 (act: -15)
XGD: -1

Green:
GAETAN XGD: +10
actual: +18
tGAETAN XESGD: +6 (act: +8)
XPPGD: +19 (act: +16)
XSHGD: -9 (act: -6)
XGD: +16.5

Alzner is two better than Green after ES play, but Green wins special teams +6 to -5..

Red Line Station and @RedArmyLine, featuring coverage of the most frustrating team in the NHL
To help with basic Timeonice functions.
If I reference a lot of stats, just assume I haven't seen anything to contradict or invalidate them.

by red army line on Jun 11, 2011 5:09 PM EDT up reply actions

Thanks for doing this, it’s great stuff, and a great advancement of the original concept. I haven’t dug in yet (been away), but it looks really good.

I don’t think I have any more raw data than I posted — I would love access to the underlying data used by NHL.com and behindthenet, but barring that, I don’t know of any way to get the data other than copying it from the web pages one screen at a time.

Atta dinnin stick a who!

by Gould Old Days on Jun 12, 2011 2:58 PM EDT reply actions

Re: Kovalchuk versus Semin, check out where Kovalchuk’s numbers are.

"Essentially, all models are wrong, but some are useful" George E.P. Box

by Knee high to a duck on Jun 12, 2011 6:53 PM EDT up reply actions

Yup. He had a bad year for a bad team.

Mind you, when it comes to offense, Kovalchuk’s bad year for a bad team still gave him 3 goals and 3 assists more than Semin. And the rink wrap voters saw this as a down, but not really all that bad year for Semin.

This was Kovalchuk’s first year below 40 goals since the lockout. Semin has never broken 40 goals. Kovalchuk played 81 games — his low since the lockout is 76. Semin played 65 games, and we were all pleased because twice in the previous three years he played even fewer games.

And from watching Kovalchuk play with the Thrashers, I remember that he could actually play decent defense when motivated. It’s just that he’s been on such a string of terrible teams, he’s hardly ever been motivated. Put him on the Caps to the right of Backstrom and Ovechkin and yes, I think he’d have tremendous success. Much more than Semin.

I’m not advocating picking up his contract, mind you. I think he’s so damaged the Devils with that thing that he now may never play with a winner and show that he’s capable of being a well-rounded player. But do I think Kovalchuk is inherently a better hockey player than Alex Semin. You bet your ass I do, no matter how much he struggled last year.

Atta dinnin stick a who!

by Gould Old Days on Jun 13, 2011 9:07 AM EDT up reply actions

Kovalchuk has a couple of offensive abilities: First things first when discussing 17 for the Devils is his shot. He’s the best shooter in the NHL, period, hands down. This isn’t in dispute – Kovalchuk is more likely to score from a given spot on the ice than any other player, according to Desjardins.

What is Ilya Kovalchuk Worth
Observed Distribution of Shooting Talent
Estimated NHL Shooting Talent
Who’s a Better Shooter, Ilya Kovalchuk or Alex Tanguay?

After you get past that, though, it’s ugly. Despite getting pretty choice starts in ATL and NJ and never playing tough competition, at least relative to his teammates, Kovalchuk has never been above 1.0 in Corsi Rel as far back at the BTN data goes. Semin’s worst season is better than Kovalchuk’s best Corsi Rel season, on better teams; Semin has a higher bar to clear to get the same result.

Semin vs. Kovalchuk Player Card

Couple of things jump out at me – Semin is a negative in the penalties taken/drawn department when you adjust for his being a forward. He ought to be drawing more than he takes, but it isn’t so. Kovalchuk is more in line with the league average there, but no special skill. Another thing that really jumps out at me is that in spite of easier minutes than a lot of his teammates got in ATL, the Thrashers gave up more shots when Kovalchuk was on the ice than off of it.

I don’t buy that he’s not motivated and therefore can’t be bothered to play good defense. Those teams would have been indisputably better if their best player could have been bothered to take some responsibility in the defensive zone. That’s a failing of the player whatever reason it came about, whether it’s innate inability to perform the task or innate inability to give enough of a fuck to compete. Alexander Semin does a better job of shot suppression and generation against better competition. I’ve never observed Ilya Kovalchuk play great defense and I’ve certainly observed many of the lazy, disinterested plays that Semin routinely takes flak for (that have become less common since the ’08 season when Fedorov arrived) coming from Kovalchuk.

Now, fun with rate stats: of the two players, the two seasons with the highest Goals/60 both belong to Semin. Both of the top two Points/60 belong to Semin. Semin owns the top spot in Primary Assists per 60, as well as 3rd place. That doesn’t paint the picture of Kovalchuk as the overwhelmingly superior offensive player – he’s getting his stats though huge minutes. It’s great that he can play those minutes, but Kovalchuk takes very long shifts; if I recall, the only forward with longer shift lengths was Alexander Ovechkin.

The durability concern is a legitimate one, but I am absolutely not buying that Kovalchuk is an inherently better hockey player than Alex Semin. You’ve often pointed out that a player’s most explosive offensive years aren’t always the years that they’re most effective at helping the hockey team win, because they’re playing defense and preventing goals from going in behind them at a better rate than they were earlier. That same analysis applies cross-sectionally across the population: offense is only one facet of a player’s value.

"Essentially, all models are wrong, but some are useful" George E.P. Box

by Knee high to a duck on Jun 13, 2011 6:50 PM EDT up reply actions 5 recs