2007-08 5v5 Goaltender Performance

by Hawerchuk on Dec 24, 2007 1:41 AM EST Comment 10 comments

I've put together a simple first-order system to analyze goaltender performance. First, I calculated the probability of scoring from 2x2-foot quadrants using data from 2001 through last week's games:

I then took every shot faced by each goaltender and calculated the number of goals we'd expect an average goaltender to allow is he faced the exact same set of shots. [Note that shot distances are relative to the end boards, not to the goal line.] This is obviously a coarse estimate: we have no information on whether shots were defended or screened, and because of the difference in the way shot types are recorded in different arenas, I haven't differentiated between them. [Yes, in hopes of improving the system, I am happy to accept critical comments!]

At any rate, the table below shows the number of goals allowed by each NHL goaltender this season (minimum: 300 mins) and the expected number of goals allowed by an average goaltender.

So as an example, Henrik Lundqvist has allowed 41 goals on 567 shots at 5-on-5, for a 927 save percentage. However, an average goalie who faced the 567 shots from the same locations would expect to allow 58.3 goals, for an 897 save percentage. On a per-60-minute basis, Lundqvist's Goals-Against-Average is 1.90 - 0.80 goals lower than an average goalie. I'll publish some charts for individual goaltenders so we can compare the best goaltenders to the worst in this metric.

PLAYER STATS							EXPECTED STATS
PLAYER	GP	TOI	SOG	GA	SPCT	GAA	EXGA	EXSPCT	EGAA	DELTAGAA
TIMOTHY JR.THOMAS	22	977:05	521	23	955	1.41	44.6	914	2.74	-1.32
HENRIKLUNDQVIST	31	1289:26	567	41	927	1.90	58.3	897	2.71	-0.80
CRISTOBALHUET	18	773:04	411	25	939	1.94	34.9	915	2.70	-0.76
ILJABRYZGALOV	25	924:13	442	25	943	1.62	36.4	917	2.36	-0.74
CHRISOSGOOD	19	797:26	337	22	934	1.65	31.7	905	2.39	-0.73
ROBERTOLUONGO	29	1160:50	536	30	944	1.55	42.7	920	2.20	-0.65
MARTINGERBER	24	964:07	542	38	929	2.36	48.1	911	2.99	-0.62
JASONLABARBERA	22	889:57	477	37	922	2.49	44.3	906	2.99	-0.49
EMMANUELLEGACE	23	934:54	394	28	928	1.79	35.6	909	2.28	-0.48
CAREYPRICE	17	728:33	386	29	924	2.38	34.4	910	2.83	-0.44
MARTINBIRON	28	1088:09	615	41	933	2.26	48.8	920	2.69	-0.43
JEAN-SEBASTIENGIGUERE	27	1013:22	433	31	928	1.83	37.7	912	2.23	-0.39
CAMWARD	27	1086:42	530	41	922	2.26	47.8	909	2.64	-0.37
RYANMILLER	31	1254:29	602	45	925	2.15	51.9	913	2.48	-0.33
RICKDIPIETRO	32	1283:57	601	46	923	2.14	52.4	912	2.44	-0.29
DANELLIS	14	502:01	230	16	930	1.91	18.2	920	2.18	-0.27
TOMASVOKOUN	32	1346:02	726	54	925	2.40	59.7	917	2.66	-0.25
VESATOSKALA	29	1168:59	554	47	915	2.41	52.0	906	2.66	-0.25
PASCALLECLAIRE	23	856:31	374	24	935	1.68	27.0	927	1.89	-0.21
NIKLASBACKSTROM	22	894:17	394	28	928	1.87	30.8	921	2.07	-0.19
MIKAELTELLQVIST	10	393:12	199	15	924	2.28	16.1	918	2.46	-0.17
ALEXAULD	15	620:41	296	22	925	2.12	23.7	919	2.29	-0.17
EVGENINABOKOV	35	1446:58	607	46	924	1.90	49.6	918	2.05	-0.14
FREDRIKNORRENA	15	527:50	237	20	915	2.27	21.0	911	2.39	-0.12
MIIKKAKIPRUSOFF	35	1438:23	660	54	918	2.25	56.8	913	2.37	-0.11
MARTINBRODEUR	30	1286:11	531	44	917	2.05	46.4	912	2.16	-0.11
NIKOLAIKHABIBULIN	25	970:42	421	39	907	2.41	40.4	903	2.49	-0.08
PATRICKLALIME	10	390:04	167	15	910	2.30	15.4	907	2.37	-0.07
MARC-ANDREFLEURY	21	760:41	388	32	917	2.52	32.4	916	2.55	-0.03
KARILEHTONEN	13	472:40	250	22	912	2.79	22.2	911	2.82	-0.03
JOHNGRAHAME	11	435:36	238	22	907	3.03	22.2	906	3.06	-0.03
PETERBUDAJ	22	906:00	419	35	916	2.31	35.0	916	2.32	-0.00
MARTYTURCO	25	1082:27	473	40	915	2.21	39.3	916	2.17	0.037
JOSHHARDING	15	579:25	313	25	920	2.58	23.6	924	2.45	0.138
DOMINIKHASEK	18	656:53	207	19	908	1.73	17.2	916	1.57	0.160
MATHIEUGARON	18	686:55	327	31	905	2.70	28.7	911	2.51	0.193
RAYEMERY	12	432:39	176	16	909	2.21	14.4	918	2.00	0.217
CURTISSANFORD	8	316:27	151	13	913	2.46	11.7	922	2.22	0.235
OLAFKOLZIG	28	1175:33	510	52	898	2.65	47.2	907	2.41	0.242
DANYSABOURIN	18	650:38	316	31	901	2.85	28.3	910	2.61	0.248
DWAYNEROLOSON	22	852:05	463	42	909	2.95	38.4	916	2.70	0.250
CHRISMASON	24	877:12	434	45	896	3.07	39.8	908	2.72	0.350
BRENTJOHNSON	9	344:40	165	17	896	2.95	14.7	910	2.57	0.386
MIKESMITH	12	465:48	206	22	893	2.83	18.9	907	2.44	0.387
HANNUTOIVONEN	10	394:32	195	22	887	3.34	19.1	901	2.90	0.436
JOHANHEDBERG	20	795:09	419	42	899	3.16	35.3	915	2.66	0.499
JOSETHEODORE	15	618:30	280	34	878	3.29	26.4	905	2.56	0.736
JOHANHOLMQVIST	30	1190:12	508	58	885	2.92	43.2	914	2.18	0.741
J-SEBASTIENAUBIN	13	427:15	214	24	887	3.37	17.9	916	2.52	0.848
ANDREWRAYCROFT	10	344:49	166	19	885	3.30	13.5	918	2.36	0.941

0 recs | Comment 10 comments |

Story-email Email Printer Print

Comments

That is absolutely fantastic. Very well done. Looking forward to the further analysis.

by Mojo Tooth on Dec 24, 2007 12:16 PM EST reply actions

Actually, even absent the goalie's personal stats, the analysis is interesting. What I mean is that it's fascinating to see which teams limit opponent's chances to the kind of shots that an average goalie would stop, versus a team that allows shots more apt to beat an average goalie.

The goalie's stats laid over top of those is just icing. (so to speak)

by Mojo Tooth on Dec 24, 2007 12:19 PM EST reply actions

Gabe -

I've played around with this a few years ago. It's cool stuff. I would suggest using road shots only to come up with a factor though - Alan Ryder looked at the problem and concluded that there's some serious bias in the data - the Rangers scorers are insane, for one. Good stuff though - this is absolutely the future.

by mc79hockey on Dec 24, 2007 4:40 PM EST reply actions

MC - thanks, that's a good idea. I'll run the home/road splits later in the season when we have more data.

by Hawerchuk on Dec 24, 2007 11:25 PM EST reply actions

I'm going to make some assumptions about what you did but I personally would recommend using rink as a random variable in the model for expected goals. Then use the broadest inference space to estimate the expected goals (you could get more particular and compare each goalie against the narrow inference space for each game but that sounds a bit to awkward for my liking.)

by Mogen David on Dec 25, 2007 2:14 PM EST reply actions

Simply superb.

by T on Jan 3, 2008 6:14 PM EST reply actions

Simply superb.

by T on Jan 3, 2008 6:14 PM EST reply actions

Very interesting, nice work.

Something that I think would be cool to do with this:
As I see it now, this is a good measure of how a goalie "fits" into his specific situation and is valuable to his team compared to how other goalies would do in his situation (maybe certain goalies are better at taking shots from certain places).
You could also use it though to compare goalies more generally. Say if you find how many shots are taken in each specific grid (in your sample space) and assume that every goalie got exactly that set of shots per game (so that each goalie is facing an equal number of shots from equal distances and so on), then see which goalie would in that instance have the best save percentage.
That would seem more fair in actually ranking the goalies (not that that was necessarily your goal).

by Ariel on Mar 7, 2008 1:22 AM EST reply actions

Awesome, just awesome. From the Rink has just linked this.

by Thrashers Recaps on Apr 27, 2009 9:26 AM EDT reply actions

This is great stuff. I am taking on a project looking at some advanced stats for hockey teams in the upcoming year and this is exactly the type of data we look at. Have you ever done any sort of look at a shooting analysis? I think it would be interesting to see what percentage of shots taken hit the net by player to see who takes "smart shots".

by John on Aug 30, 2009 10:38 AM EDT reply actions