SBNation.com User Blog: JpR-dc

Need a distraction? (wtf Game 2) Come follow.

JpR-dc — Fri, 17 May 2013 15:33:09 -0000

It is probably true that anything to do with numbers is the last thing most people will want to read after the first two games of this series. It might go something like this:

A: Man, we lost!!
B: Still out-chanced ‘em. By my count--[A punches B]
A: OH GOD NUMMMMMMBERS [staggers, pets a wounded, yet still-breathing shark, sheds a tear, dries the tear, whispers something to wherever a shark's ear is located, exeunt]

For some, numbers provide no quarter. For others, their ease of manipulation allows us to "keep believing" in spite of harsh realities. Maybe you’re in the middle. I’m still in the latter group. Come on over. Try to the Kool-Aid.

[A re-enters, gives audience skeptical look.]

In a previous Fanpost, I looked at the notion "score effects". I looked at just the recently completed regular season and created expected outcomes for a team Down 2 at even-strength (as this happened to the Sharks in Game 1—sorry to bring that up). There were positives to be drawn (Sharks outperformed the league in Shots For) and negatives (Quick SV%=100).

There was an interesting finding I didn’t mention. In Down 2 situations, at even-strength:

Goals Against (GA) rate > Goals For (GF) rate

I didn’t mention this because I felt that I needed to look at more data to verify it. In looking at the past 6 regular seasons, it is always the case. The rate of a Down 2 team allowing another goal was greater than the rate of getting one goal back.

Now, it’s a little tricky to understand this because I believe the rates are determined by totaling GF and GA in Down 2 situations and then dividing by time on ice (TOI). However, we can conclude that a GA occurs more frequently (and, if I may, is more likely) than a GF. This is the case in each individual season (and obviously in the overall mean).

Then I looked at the last 5 playoffs. The data itself here is a little spotty, plus it is a smaller N, and I think cross-checking with NHL RTSS would help, but I calculated some means anyway. Three quick points: 1) the GF and GA rates fluctuate A LOT more; 2) in two playoffs (2007-2008 and 2010-2011) we saw GF rates larger than GA rates; and 3) in the overall mean, however GA rate was still greater than GF rate in the playoffs and the difference was actually much greater in the 5 playoffs than the 6 regular seasons.

[A reaches for bludgeoning weapon, any bludgeoning weapon.]

	TOI	GF/1	GA/1	SF/1	SA/1	MF/1	MA/1	FEN+	FEN-	opp SV%
REG MEANS (N=6)	3.780055556	0.04697037	0.053122222	0.553132407	0.429240741	0.234322222	0.183561111	0.834425	0.665924074	0.921729457
POFF MEANS (n=5)	3.335	0.04089359	0.055173718	0.455249145	0.365640598	0.210475855	0.159383333	0.70661859	0.58019765	0.917576966
DOWN 2, even-strength; from BTN

What’s this mean? First, regarding Game 2 (sorry to bring it up) Sharks outperformed the league by generating positive Fenwick events and getting a goal in 2.12 minutes of Down 2 play. Second, regarding Games 1 and 2, Sharks outperformed the league in terms of a) positive Fenwick events; and b) not allowing a GA. Three, when we say "score effects" we should know that really we are simply talking about a positive Fenwick differential that is greater than what we typically see in Close situations (demonstrating this with data would be a next step). In the aggregate, the explanation seems to be: in Down 2 situations, teams take greater risks to get shots off, which leads to more shots, but can also lead to odd-man rushes and GA the other way. In fact, this seems to happen more often than actually generating a GF from a Fenwick event. This has not happened to the Sharks, which means that when they are Down 2, they are outperforming most other teams. This is a good thing.

[A yells some stuff like WHERE GRAPH YOU NUMMMMBERS, exeunt.]

Now, if they could just not be Down 2 again on Saturday. Until then...

[Shared spreadsheet with data (individual sheets for season/playoff years; sheet with means by season and in the aggregate, expected outcomes from season/playoff, and Sharks actual outcomes in Games 1 and 2) here.]

It is probably true that anything to do with numbers is the last thing most people will want to read after the first two games of this series. It might go something like this:

[A re-enters, gives audience skeptical look.]

There was an interesting finding I didn’t mention. In Down 2 situations, at even-strength:

Goals Against (GA) rate > Goals For (GF) rate

[A reaches for bludgeoning weapon, any bludgeoning weapon.]

	TOI	GF/1	GA/1	SF/1	SA/1	MF/1	MA/1	FEN+	FEN-	opp SV%
REG MEANS (N=6)	3.780055556	0.04697037	0.053122222	0.553132407	0.429240741	0.234322222	0.183561111	0.834425	0.665924074	0.921729457
POFF MEANS (n=5)	3.335	0.04089359	0.055173718	0.455249145	0.365640598	0.210475855	0.159383333	0.70661859	0.58019765	0.917576966
DOWN 2, even-strength; from BTN

[A yells some stuff like WHERE GRAPH YOU NUMMMMBERS, exeunt.]

Now, if they could just not be Down 2 again on Saturday. Until then...

Unpacking score effects (i.e. dangitquick)

JpR-dc — Wed, 15 May 2013 16:13:38 -0000

Preface: The notion of "score effects" comes into the conversation when we see one team outshooting another by a significant margin, yet losing. Say a 2-0 loss where the losing team outshoots the other by 50% at even-strength (e.g. Game 1; apologies for bringing it up). This notion is important if we believe shots drive goals for (positively) and goals against (negatively). And by the way, I'm a believer, but this doesn't mean we should stop trying to falsify this (at least, at this stage). But this post isn't about that. Let's move on.

Man, that 3rd period. Thought something was going to happen. Really did. But it didn't. Burns. Man, thought he had one. But he didn't. So, question: should I be upset about this? And if so, at whom?

When the game ended, yeah, I was deflated, but hey, Game 1, right? I figured the shot statistics would favor the Sharks because of the 3rd period. Right away, I could hear someone's internet voice sternly saying: guys, guys... GUYS!... score effects. [Silence.]

I agree. But I wanted to measure up how well the Sharks performed in this score effects situation, i.e. Down 2. To do this, I did a simple test.

1) I looked at the Down 2 Fenwick data from BTN (and I believe this is even-strength).

2) I calculated several per minute average rates across the entire league for the 2012-2013 NHL regular season: Goals For, Shots For, Shots Against, and Save % (of the team Up 2). I did this by simply finding the means for these columns and dividing by 60. The reason I did this is to create a set of expected outcomes for the Sharks time on ice Down 2. Note: I'm leaving Missed shots, er, Fenwick, aside for the moment. Here are the rates:

GF/1 minute: 0.0498
SF/1: 0.579
SA/1: 0.455
SV%: 91.4%

3) The Sharks played 21.63 minutes of Down 2 time at even-strength. (Note: After Richards goal, there were two penalties (Torres, Richards) and Niemi was on the bench for 1.87 minutes. 27.50 minutes from goal 2 to the end of the game. Subtract 2, 2, and 1.87 and you get 21.63.)

4) Given 21.63 minutes of Down 2 time, and based on league-wide rates for Down 2 during the 48-game regular season, what are the expected outcomes? What are the actual outcomes. ... Feels like we're about to see a tabl--

	Expected	Actual	Difference
GF	1.077	0.000	-1.077
SF	12.520	16.000	3.480
SA	9.850	5.000	-4.850
Quick SV%	91.4%	100.0%	8.6%

Table 1. Expected and actual outcomes in the Down 2 time-on-ice. Expected derived from 2012-2013 league-wide Down 2 statistics at BTN. Actual derived from WCSF Game 1 SJS v LA. I cross-checked with NHL's play-by-play to be sure, correct me if I'm wrong.

Let's look at these and answer the original questions. First, if the Sharks had performed in step with the league during the regular season, they should have gotten one back. So, that sucks. Second, the Sharks were putting 50% more shots on goal and (concurrently, if not consequently) also reducing shots against on goal by almost 50%. So, that's good. Third, and I think this is the strongest statement I can make given the data pres--goddamnitjonathanquick.

Bottom line: First, while Sharks outperformed the league in making the opposition goalie make save after save, they were outperformed by dammitthatguy. This is another piece of evidence that Quick contributed significantly to the Game 1 win (since his SV% was so much greater than the league average Down 2). Second, this also suggests, that while we can say that score effects has some impact on the end-game shot differential, the Sharks team talent at possessing the puck and getting shots on goal had a significant impact on Game 1 as well (again, by virtue of outperforming the league average rates by about 50%). Therefore, I ain't so upset. ... Um, drop the puck already.

Man, that 3rd period. Thought something was going to happen. Really did. But it didn't. Burns. Man, thought he had one. But he didn't. So, question: should I be upset about this? And if so, at whom?

I agree. But I wanted to measure up how well the Sharks performed in this score effects situation, i.e. Down 2. To do this, I did a simple test.

1) I looked at the Down 2 Fenwick data from BTN (and I believe this is even-strength).

GF/1 minute: 0.0498
SF/1: 0.579
SA/1: 0.455
SV%: 91.4%

	Expected	Actual	Difference
GF	1.077	0.000	-1.077
SF	12.520	16.000	3.480
SA	9.850	5.000	-4.850
Quick SV%	91.4%	100.0%	8.6%

An exercise to flatten fluctuations in optimism: historical cases of below-average goalie performance, yet winning the Cup

JpR-dc — Wed, 01 May 2013 06:12:39 -0000

Just finished watching ANA-DET and got caught up in thinking about two things that have been topics of discussion here and elsewhere recently. The first is goalie performance in the playoffs and the second (related) are factors that contribute to playoff success. (Also, I need a bump in optimism. If you are already optimistic, check out these sweet line graphs that illustrate "hotness" entering the playoffs. How can you not like them?)

I had said recently that I liked how this year's team was generating more pucks to the net (Fenwick For/60) than any other Sharks team with the exception of the team from 2010-2011. (Note: I think the behindthenet.ca statistics have been updated to reflect the final game since I wrote that fanpost). Plus, Niemi is playing ballsawesome.

As Snark points out (in the comments), goalie performance in the playoffs is volatile in small sample sizes and shouldn't be trusted on its own. Well, fair enough. A fanpost by holidaypark suggested the combination of Fenwick Close%, even strength SV%, and PP/PK% to think about predicting playoff performance. Though preliminary (and pretty high up the "grandness" scale of theorizing), it was helpful in thinking about the following scenario:

Suppose Niemi is below-average during these here playoffs in 2013. What would the Sharks need to look like as a team to compensate? If this first week consists (as it probably will) of me squirming in the fetal position on the couch mumbling cmoncmoncmoncmon because Niemi's SV% has fallen off the cliff, what possession and special team numbers can I find to ~~self-medicate~~ make myself feel better?

Here's an outline of the steps I took to answer this question:

(1) Determine what historical cases exist of goalies playing below average in the playoffs, but their teams still finding high levels of success. (2) Determine what these teams' possession and power play numbers were and eyeball compare them to the Sharks regular season sample of 2012-2013. (3) If differences seem great, freak out. If differences seem reachable, freak out less. (I thought of making a decision tree for this, but it seemed superfluous.)

So, I dragged out data from the past 5 years from NHL.com on goalie playoff performance. There were 122 player-years total, covering 872 total game starts (thus 436 playoff games in the 5 year sample). I summed the total shots and saves and divided to get an average SV%. This--it is true--covers all time on ice, regardless of score or man-advantage. Goalies in this 5 year playoff sample stopped 91.5% of all shots on goal in all situations.

Now, of the Cup-Winning teams, did any of the primary goalies have a SV% of less than 91.5%? For the last two years, we have Jonathan Quick at 94.6% and Tim Thomas at 94.0%. Those two won 16 games on their own. The next two were exactly what I was looking for. Antti Niemi at 91.0% and Marc-Andre Fleury at 90.8%. Both also won 16 games on their own. Lastly, Chris Osgood was at 93.0% and won 14 games en route to the 2008 Cup.

Goalie	SV%
Quick	94.6
Thomas	94.0
Niemi	91.0
Fleury	90.8
Osgood	93.0

Table 1. SV% for last five primary goalies for Cup-winning teams.

Step one done. Step two. I looked at the playoff possession and power play statistics (since penalty kill statistics rely on the goalie to some degree, I set them aside for the moment). I plucked behindthenet.ca stats on Chicago 2009-2010 (22 games) and Pittsburgh 2008-2009 (24 games).

For simplicity, I'm presenting data for these teams (1) in Close 5v5 situations, in terms of Fenwick%, goals for and against, and shots for and against, to get a sense of how they are performing during tight moments in the playoffs where there's no room and you still need to possess the puck, and then (2) in 5v4 situations, looking at goal and shot generation.

Close 5v5	FEN%	GF/60	GA/60	SF/60	SA/60	5v4	GF/60	SF/60
CHI 09-10 PLOF	55.44	3.13	2.11	32.00	25.19		7.40	48.30
PIT 08-09 PLOF	52.40	3.06	2.55	30.26	25.42		7.70	53.30
SJS 12-13 REG	52.41	2.48	2.02	28.29	26.47		6.60	58.10

Table 2. Comparing puck numbers between two Cup-winning teams with below-average performing goalies and one current playoff team that might have a below-average performing goalie.

Eyeball test discloses the following points:

1) These two Cup-winning teams are scoring at lot of goals (>3 GF/60) in 5v5 Close play. Like more goals/60 than the best goal-generating Sharks team in comparable situations in the past 5 years (Sharks 2009-2010, 2.91 GF/60).

2) Related to 1), these two teams also generated more shots on goal/60, though the Sharks team Fenwick is almost identical to the Penguins 2009 Cup team. (On a total nerd note: the difference between 55.4% and 52.4% in Fenwick% looks like about three more Fenwick For events and two less Fenwick against. There is some preliminary stuff here, but this needs more attention to understand more precisely what's going on.)

3) The power play was running at a higher success rate for the Cup teams than the Sharks, though the Sharks were getting more pucks at the net.

This means that if Niemi has a below-average run (that is, a drop from his season SV% of 92.4% in all situations to a SV% < 91.5%) the team may not be out of luck. If the Sharks can continue to possess the puck, they just need to convert a little more often in Close 5v5 and on the power play. Consider this statistic: the Sharks SH% in 5v5 play this season was 6.68%. If we get a SV% of 91.5 and a PDO of 1000, this would actually entail a SH% improvement. I point this out to argue that this is not beyond the variation we might see for teams that can possess the puck in the 48-game stretch we have just completed. In short, to compensate for a lower SV%, we need a bump in SH%. Without it, a deep playoff run is unsustainable. Fortunately, such a bump is definitely reachable, and should not be entirely unexpected.

Ahh. I feel better now.

[Final FYI: feel free to interpret this entire post as a Niemi reverse-jinx.]

Suppose Niemi is below-average during these here playoffs in 2013. What would the Sharks need to look like as a team to compensate? If this first week consists (as it probably will) of me squirming in the fetal position on the couch mumbling cmoncmoncmoncmon because Niemi's SV% has fallen off the cliff, what possession and special team numbers can I find to ~~self-medicate~~ make myself feel better?

Here's an outline of the steps I took to answer this question:

Goalie	SV%
Quick	94.6
Thomas	94.0
Niemi	91.0
Fleury	90.8
Osgood	93.0

Table 1. SV% for last five primary goalies for Cup-winning teams.

Close 5v5	FEN%	GF/60	GA/60	SF/60	SA/60	5v4	GF/60	SF/60
CHI 09-10 PLOF	55.44	3.13	2.11	32.00	25.19		7.40	48.30
PIT 08-09 PLOF	52.40	3.06	2.55	30.26	25.42		7.70	53.30
SJS 12-13 REG	52.41	2.48	2.02	28.29	26.47		6.60	58.10

3) The power play was running at a higher success rate for the Cup teams than the Sharks, though the Sharks were getting more pucks at the net.

Ahh. I feel better now.

[Final FYI: feel free to interpret this entire post as a Niemi reverse-jinx.]

Sharks vs. Sharks: A glance at the numbers (and a case for Niemi-inspired optimism)

JpR-dc — Sun, 28 Apr 2013 19:43:28 -0000

Something I did during the lockout was start to read (and catch up on) all the work statistical work that was being done on hockey. There's way more out there than I had initially thought. Like a(n) (Imperial) shit-ton more. I humbly began to look at historical Sharks team stats on a team-season basis, and I wasn't sure how to interpret what was going on. With this season over, I thought I'd revisit this by asking: in terms of the statistical aggregates, what Sharks team is the ~~2012-~~2013 Sharks team the most like?

Yeah, here's a table:

close 5v5	fenwick	toi	gf/60	ga/60	sf/60	sa/60	mf/60	ma/60	fenFOR/60	fenAG/60	totFEN/60	pts	ppg	playoff result
2007-2008	56.33	29.78	2.51	2.36	28.08	20.96	11.28	9.14	41.87	32.46	74.33	108	1.32	2nd RD
2008-2009	55.86	28.12	2.50	2.16	29.64	22.74	11.92	9.91	44.06	34.81	78.87	117	1.43	1st RD
2009-2010	51.10	27.67	2.91	2.35	29.19	28.69	12.64	11.77	44.74	42.81	87.55	113	1.38	WCF
2010-2011	54.44	30.37	2.48	2.10	30.33	26.19	13.01	10.07	45.82	38.36	84.18	105	1.28	WCF
2011-2012	52.19	32.14	2.39	2.25	29.51	26.50	12.00	11.45	43.90	40.20	84.10	96	1.17	1st RD
2012-2013	53.03	30.04	2.51	2.00	28.51	26.39	13.98	11.47	45.00	39.86	84.86	57	1.19	??

Table. TABLE. Aggregate numbers in Close 5v5 situations, Sharks, 07-08 through 12-13.

First, I looked at Fenwick% as this encapsulates goals, shots, and missed shots. In this perspective, this year's team is most like last year's team, but a little better. In what ways better? Well, they have given up the least of amount of goals in Close 5v5 play than any Shark team in the time period. This is probably due to Antti Niemi's sick season. Thanks, Antti. Seriously.

Then I looked at other "defensive" aspects, like shots against/missed shots against, to get some context, and what struck me when I first looked at these numbers came roaring back. The teams from 07-08 and 08-09 were ridiculous in terms of Fenwick Against. It's a real break in the data, if you consider their shots against and missed shots rates (summed in the fenAG column). What's also striking is that, with the exception of 11-12 Sharks, every other team had a better Fenwick For figure than 07-08 and 08-09. (And yeah, 09-10 had the worst Fenwick% and went to the WCF.)

So, what this all mean? I'd actually like to think that the Sharks of this year are more like Sharks 10-11 than Sharks 11-12. They are closer in terms of goals for and goals against, and total fenFOR. I want to think that with Niemi playing extremely well, as long as this team gets pucks toward the net (this year's iteration the 2nd best at doing so), they have a good a chance as any of the previous teams to go deep.

Yeah, here's a table:

close 5v5	fenwick	toi	gf/60	ga/60	sf/60	sa/60	mf/60	ma/60	fenFOR/60	fenAG/60	totFEN/60	pts	ppg	playoff result
2007-2008	56.33	29.78	2.51	2.36	28.08	20.96	11.28	9.14	41.87	32.46	74.33	108	1.32	2nd RD
2008-2009	55.86	28.12	2.50	2.16	29.64	22.74	11.92	9.91	44.06	34.81	78.87	117	1.43	1st RD
2009-2010	51.10	27.67	2.91	2.35	29.19	28.69	12.64	11.77	44.74	42.81	87.55	113	1.38	WCF
2010-2011	54.44	30.37	2.48	2.10	30.33	26.19	13.01	10.07	45.82	38.36	84.18	105	1.28	WCF
2011-2012	52.19	32.14	2.39	2.25	29.51	26.50	12.00	11.45	43.90	40.20	84.10	96	1.17	1st RD
2012-2013	53.03	30.04	2.51	2.00	28.51	26.39	13.98	11.47	45.00	39.86	84.86	57	1.19	??

Fenwick and Playoffs

JpR-dc — Mon, 15 Apr 2013 18:44:17 -0000

Fenwick and Playoffs

I missed this. Of interest to ppl wanting to know association b/w Fenwick and Playoffs.

Response to PDOfrock: Comparing streaks (test case of shots in sets of 7)

JpR-dc — Wed, 10 Apr 2013 14:36:25 -0000

Recent Fanpost on compiled shot data from the homestand by PDOfrock led me to think about the idea of comparing compiled shot data. What looks the same and different? Is there value in comparing two sets of "win streaks"? (This is just a test case to see what comes up. A number of sets to compare are available. Feel free to apply elsewhere!)

5v5 close	GP	goals for	goals against	saved shots for	saved shots against	Shots%	missed shots for	missed shots against	Fenwick%	S.J Shots that were Blocked	blocked shots by SJ	Corsi%	On Ice EVsave%	On Ice EVshooting%	PDO
SJS 1-7	7	9	2	70	72	0.516	42	37	0.522	56	50	0.524	0.973	11.40%	1087
SJS 31-37	7	12	4	116	86	0.587	60	43	0.586	60	68	0.552	0.956	9.40%	1050

Yup, it's a table. Numbers from timeonice.com here and here. Games 1-7 and games 31-37.

So, here are the numbers from two win streaks, the first being games 1-7 (remember those? good times) and the second games 31-37 ("the homestand"). A number of things stand out.

1) Most significantly (probably): TON more shots toward the opposing net. Lots more SOGs and Missed Shots. If we see 70 SOGs in 5v5 close play, as we did in games 1-7, I think we should be worried. Note the 6.4% difference in Fenwick% and 2.8% difference in Corsi% in comparing the two sets. Note the Sharks had a -2 SOG differential in games 1-7, though Fenwick% and Corsi% favored them. It's still important (in my view) to look at all the shot statistics.

2) More goals and blocked more shots in 31-37, along with worse SV% and SH% (and PDO; so, probably closer to something we think of as "true" talent level).

Bottom line: first streak characterized by special teams, ridiculous goaltending and shooting%; second streak characterized by an enormous amount of shots. We should fear the first streak and embrace the second as perhaps the "ideal typical" win streak.

5v5 close	GP	goals for	goals against	saved shots for	saved shots against	Shots%	missed shots for	missed shots against	Fenwick%	S.J Shots that were Blocked	blocked shots by SJ	Corsi%	On Ice EVsave%	On Ice EVshooting%	PDO
SJS 1-7	7	9	2	70	72	0.516	42	37	0.522	56	50	0.524	0.973	11.40%	1087
SJS 31-37	7	12	4	116	86	0.587	60	43	0.586	60	68	0.552	0.956	9.40%	1050

Yup, it's a table. Numbers from timeonice.com here and here. Games 1-7 and games 31-37.

So, here are the numbers from two win streaks, the first being games 1-7 (remember those? good times) and the second games 31-37 ("the homestand"). A number of things stand out.

2) More goals and blocked more shots in 31-37, along with worse SV% and SH% (and PDO; so, probably closer to something we think of as "true" talent level).

Douglas Murray: regardez le nombres

JpR-dc — Fri, 22 Mar 2013 16:45:30 -0000

If there's one thing many people have agreed on it's that Douglas Murray 2012-2013 is killing our perception of Douglas Murray prior to 2012-2013. (We can let Douglas Murray 2011-2012 have his own place since, you know, puck broke his laryngeal prominence.) I don't have to look hard to see that there's a ton of criticism of his play.

Now, I'm going to take an agnostic position on all this for now. Let's look at some simple Shark defencemen numbers, and try to see what's going on.

I looked at two things I care about in Sharks defencemen. The first is simple +/-. Now, I know traditional +/- takes a beating, but I still think it tells us something about what happens on the ice 5v5. The second was to look at Fenwick +/-. I know that, again, some prefer different metrics, but I actually didn't want to include blocked shots in comparing Shark defencemen because a) Murray lets shots get off because he's such a good blocker, and b) I want to know about pucks actually getting toward/in the vicinity of the net. The numbers are drawn from this morning via behindthenet.ca.

NAME	GFON/60	GAON/60	G+/-	SFON/60	SAON/60	SHOTS +/-	MFON/60	MAON/60	FENWICK+/-
MATTHEWIRWIN	1.71	2.66	-1	32.7	21.5	11.2	11	9.3	12.9
DANBOYLE	1.72	2.25	-0.5	32.1	23.3	8.8	11.9	10.2	10.5
BRENTBURNS	2.29	3.44	-1.2	29.4	25.6	3.8	13.4	8.4	8.8
MARC-EDOUARDVLASIC	2.64	2.28	0.36	25.8	25.8	0	13	10.7	2.3
BRADSTUART	2.25	1.87	0.38	25.9	28.6	-2.7	11.4	11.5	-2.8
JASONDEMERS	0.75	2.24	-1.5	23.9	27.6	-3.7	12.7	12.3	-3.3
JUSTINBRAUN	1.74	2.13	-0.4	22.4	27.6	-5.2	13.5	12.6	-4.3
DOUGLASMURRAY	1.24	2.17	-0.9	23.3	31	-7.7	12.6	14.6	-9.7

Ja, beaucoup de nummern. Sorted by Fenwick +/-. Excluding Pelech and Petrecki.

Now, looking at this table, I want to point out a couple of things.

First, MATTHEWIRWIN. We should just call him MATTHEWIN.

Second, Murray is obviously giving up a ton of shots. Per 60 minutes, a difference of 5 more shots against than the next closest defenceman (Braun). In addition, Stuart, Demers, and Braun are all within 1.5 shots of each other. On the one hand, this little clumping at the bottom segment suggests Murray is a real outlier. Again, he gives up a shitton of SA/60. On the other hand, note that with the straight-up SOG +/- the gap is smaller. Presumably, Murray "produces" more missed shots because Murray is a human 2x4 Lego block (thanks, Nordic people). In SF/60, also note that Murray isn't even the bottom (Braun). He's in the bottom segment with Demers and Braun.

Third, Murray's traditional +/- rate is better than MATTHEWIN, Burns, and Demers. In terms of GF/60 when he is on the ice, Murray correlates better than Demers. In terms of GA/60 when he is on the ice, Murray correlates better than MATTHEWIN, Boyle, Burns, Vlasic, and Demers. What this means is that FOR SOME REASON, Murray on the ice equates with a very high SV% for Niemi (93.5%, only Stuart's is better at 93.9%). Since I'm not sure what this reason is (chance likely having some effect), I'm going to chalk this up to some kind of crazy Nordic telepathic shit. Watch how Murray interacts with Niemi. If you saw the last holiday video, you know what I'm talking about.

[Just for the shits: In the two previous seasons (10-11 and 11-2, i.e. when Niemi showed up), Murray's on-ice SV% was, respectively, 92.9% (behind Kent Huskins, 94.6% and Demers, 93.8%, both played at +50GP), and 94.4% (tops excluding Vandermeer, 25GP). This season Demers SV% is 92.5%, so, not quite the same crazy Nordic telepathic shit (CNTS) as Murray.]

Lastly, caveat: I'm leaving out his cap hit in this analysis. This is on-ice play 5v5. Also, I like all these guys and wish they would play better. Yup, I'm a fan.

Bottom line:

1) Because I care about shot differential, I'm least excited when Murray is on the ice. When he is on, there tends to be way more shots toward Niemi. I usually just hope for pleasegodnotanotheroddmanrush, saves, and a shot +/- of 0 at the end of the game. These are not particularly joyful feelings.

2) On the other hand, Murray's SF/60 rate suggests that scoring opportunities do occur when he is on the ice. In addition, Niemi seems to save a lot of shots when Murray is on.

Now let me add eyeballs here. The explanation for continued deployment seems to be: coaches like him. Maybe the reason why they like him over Demers and Braun (seems like they are the 'competition' here) is that coaches KNOW what they will get with him. Because of their youth (and the often attending fluctuations in play), perhaps this is just coaches leaning toward known quantities rather than more unpredictable play; even if those known quantities tend toward great shot differential. (I'm not sure how to measure unpredictable play, but this is what my eyeballs are telling me.) Lastly, I think that if Demers and Braun were playing like MATTHEWIN, we wouldn't be having this conversation. The numbers for those two, while generally better, aren't great.

Now, I'm going to take an agnostic position on all this for now. Let's look at some simple Shark defencemen numbers, and try to see what's going on.

NAME	GFON/60	GAON/60	G+/-	SFON/60	SAON/60	SHOTS +/-	MFON/60	MAON/60	FENWICK+/-
MATTHEWIRWIN	1.71	2.66	-1	32.7	21.5	11.2	11	9.3	12.9
DANBOYLE	1.72	2.25	-0.5	32.1	23.3	8.8	11.9	10.2	10.5
BRENTBURNS	2.29	3.44	-1.2	29.4	25.6	3.8	13.4	8.4	8.8
MARC-EDOUARDVLASIC	2.64	2.28	0.36	25.8	25.8	0	13	10.7	2.3
BRADSTUART	2.25	1.87	0.38	25.9	28.6	-2.7	11.4	11.5	-2.8
JASONDEMERS	0.75	2.24	-1.5	23.9	27.6	-3.7	12.7	12.3	-3.3
JUSTINBRAUN	1.74	2.13	-0.4	22.4	27.6	-5.2	13.5	12.6	-4.3
DOUGLASMURRAY	1.24	2.17	-0.9	23.3	31	-7.7	12.6	14.6	-9.7

Ja, beaucoup de nummern. Sorted by Fenwick +/-. Excluding Pelech and Petrecki.

Now, looking at this table, I want to point out a couple of things.

First, MATTHEWIRWIN. We should just call him MATTHEWIN.

Lastly, caveat: I'm leaving out his cap hit in this analysis. This is on-ice play 5v5. Also, I like all these guys and wish they would play better. Yup, I'm a fan.

Bottom line:

2) On the other hand, Murray's SF/60 rate suggests that scoring opportunities do occur when he is on the ice. In addition, Niemi seems to save a lot of shots when Murray is on.

For each FenwickClose% increase/decrease = what increase/decrease in Fenwick+?

JpR-dc — Wed, 20 Mar 2013 18:11:23 -0000

We need some distractions. Let’s get distracted for a minute...

The Sharks Fenwick numbers have taken a nose dive since around game 17, as noted here.

But, what does FenwickClose% mean in a more tangible way? What does the nose dive in FenwickClose% mean in a tangible way?

TANGIBILITY.

Let me say a couple basic things about FenClo% given the data typically used (5 seasons, 150 team-seasons, from behindthenet.ca).

First, remember all of this is 5v5 play. Second, we keep saying that FenClo% is an interesting statistic because it typifies puck possession as a percentage, not a raw number. In real terms, though, what does a FenClo of 50% mean? Alright, let’s do this.

A Fenwick event is a shot on goal or a missed shot (either for or against; let’s call this Fenwick+ and Fenwick-). The Fenwick percentage is the percentage of the total number of Fenwick+ for 1 team. FenwickClose% is this percentage is "close" situations (+1, tie, -1).

A FenClo% of 50 means that the teams had an even number of Fenwick+. A FenClo% of 100 means that all Fenwick+ events were for 1 team. A FenClo% of 0 means that this team has no Fenwick+ events. (Dang.)

Ok. First, we need to get the average number of total Fenwick events per game. Once we have that, we can compute what the percentages mean in real terms (Fenwick+, which, again, is shots on goal + missed shots on goal).

Per the data:

1) Close score situations (by minute) per game average = 28.49 (about ½ a game).

2) In those situations, the average number of shots for + missed for (per 60 minutes) AND the average number of shots against + missed against (per 60 minutes) = 77.48 (total Fenwick events per 60 minutes)

3) So, per minute, there are 1.291 Fenwick events in close situations.

4) Since we know time average, we can multiply and get this: on average there will be 36.79 total Fenwick events per game.

If a team has a FenClo% of 50, then they will have 18.39 Fenwick+ and 18.39 Fenwick- events. If we presume a linear increase (for the purpose of estimation, let’s accept this for the moment), a FenClo% of 51 means that we can expect 18.76 Fenwick+ and 18.03 Fenwick-. If FenClo% is 52? We can expect 19.13 Fenwick+ and 17.66 Fenwick-.

Remember that each 1% increase in FenClo% means a decrease of 1% for the opposition (50-50, 51-49, 52-48, etc.). In real terms, a 1% percent difference affects Fenwick differential by 0.74.

	A	B	A	B	A	B
FenClo%	50%	50%	51%	49%	52%	48%
Fenwick+	18.40	18.40	18.76	18.03	19.13	17.66

Yup, table of Fenwick+ increasing and decreasing.

By dropping from 53% to 50% in FenClo%, this means the Fenwick differential has increased by more than 2. That’s, in effect, one more shot to defend, and one less shot for. TANGIBLE.

Bottom line: The result here may not seem meaningful. Does one more shot to defend and one more shot toward the net matter? Consider though, all of the crap that has to happen for a shot to even get off. Some kind of possession gain (either face-off win and/or zone breakout). If not already in the offensive zone, some kind of positive zone entry (either a carry-in or a dump-in + recovery). This all takes work all going toward that one thing we know has to happen in order for the possibility of a goal: the shot. Remember the trail of crumbs (I mean, causal model) goes from: puck possession (preventing shots against, allowing shots for) → shots (pucks toward net) → goals (pucks in net). Let’s go shots.

We need some distractions. Let’s get distracted for a minute...

The Sharks Fenwick numbers have taken a nose dive since around game 17, as noted here.

But, what does FenwickClose% mean in a more tangible way? What does the nose dive in FenwickClose% mean in a tangible way?

TANGIBILITY.

Let me say a couple basic things about FenClo% given the data typically used (5 seasons, 150 team-seasons, from behindthenet.ca).

Per the data:

1) Close score situations (by minute) per game average = 28.49 (about ½ a game).

3) So, per minute, there are 1.291 Fenwick events in close situations.

4) Since we know time average, we can multiply and get this: on average there will be 36.79 total Fenwick events per game.

Remember that each 1% increase in FenClo% means a decrease of 1% for the opposition (50-50, 51-49, 52-48, etc.). In real terms, a 1% percent difference affects Fenwick differential by 0.74.

	A	B	A	B	A	B
FenClo%	50%	50%	51%	49%	52%	48%
Fenwick+	18.40	18.40	18.76	18.03	19.13	17.66

Yup, table of Fenwick+ increasing and decreasing.

By dropping from 53% to 50% in FenClo%, this means the Fenwick differential has increased by more than 2. That’s, in effect, one more shot to defend, and one less shot for. TANGIBLE.

Shot location search tool (yup)

JpR-dc — Mon, 18 Mar 2013 18:58:37 -0000

Shot location search tool (yup)

What the link title says... shots, goals, and, location. Don't know if you guys heard of it, but it was featured in a Japers article. I hadn't seen it until today.

Distractions: measuring individual player penalty-kill talent (UPDATED to suck less)

JpR-dc — Thu, 14 Mar 2013 17:43:30 -0000

… so I know Zus and Burish have been getting flak for their general inability to pick up points (combined 4 points in 50 games played... just in case you had forgotten, player points is like real important to scoring outcomes, like). But let’s look at some other stuff. Distractions, we need distractions. That’s key. That’s the ticket.

[THIS totally worked vs LA last night.]

So there were a couple pieces (here and here) that looked at penalty killing attributes using some behindthenet.ca statistics.

[For anybody using the behindthenet.ca statistics, I recommend using the "old" site interface. As it turns out, I'm ~~pretty~~ totally sure I screwed up my original table. Thus, the need for this corrective. God forbid we let errors lie. Never.]

Hmm. Yeah, but I don’t care about offense during the penalty kill. Most of the time getting a short-handed chance is either: a) somebody really fucking up a blue-line pass or b) a risk by a penalty kill player that may have reverse consequences.

Here’s an example. (This was so heart-warming, I did a breakdown here.)

So, here are the assumptions and the argument I want to offer:

1) I am placing value on the ability of PK forwards to prevent pucks from being shot toward the net.

2) Why? Teams on the PP want to get shots at the net. Power play shots (by the numbers) drives PPG. (Perhaps a post for another time.) Thus, I care about the prevention of shots toward the net.

3) We have a measurement for raw numbers of pucks toward your own net: Fenwick-. If anybody's seen others apply Corsi-/60 or Fenwick-/60 to PK, I'd love to see the links.

Caveat: Now I think that blocked shots are still an important PK ability. Indeed, there might be situations where a PK forward is goading a player to shoot so that he can block the shot. For the moment, I'm not buying that. The thought of the shooting player that precedes the blocked shot is, "hey, I can squeeze this shot through to the net." The ability to prevent the shooting player from thinking this is what I think is most important for the PK forward.

As it happens, I have a bit of evidence to support the caveat. I start here: we know that this year's Sharks PK is better than last year's PK. We know this because of the decline in PPGs allowed. If we look at shot statistics, what has happened with this decline?

I looked at last year's forward rates (GA On/60, SA-shots against-On/60, MA On/60, and BF On/60) and compared them to this year's forward rates (through 24 GP). The results:

2011-2012 (means)	2012-2013 (means)
GA On/60: 6.34	GA On/60: 2.52
SA On/60: 43.76	SA On/60: 25.56
MA On/60: 16.69	MA On/60: 9.83
BF On/60: 34.83	BF On/60: 18.36

Yup, table of nombres, part I. Bigger table below, you guys.

Obviously, the GA number stands out. In addition, we observe way fewer shots on goal and way fewer pucks getting through to the net (Fenwick-). Notice the corresponding decline in Goals Against with every other statistic. The 60% drop in GA On/60, corresponded with more than 40% drops in shots against, missed shots against, and blocked shots for.

Now, this is not entirely intuitive. Is it possible that a PPG drop can occur with an increase in blocked shots? Of course. But in this case, where we know the PK is better now than in 2011-2012, it looks like reducing shots actually getting off is more important than getting more blocked shots. (Yeah, obviously single case study, but it's still evidence.)

Now that I've totally justified Fenwick- as a measure of PK ability beyond the shadow of a doubt, let's check out how Shark forwards ranked last year and so far this year:

2011-2012
NAME	GP	TOI/60	SVPCT	GAON/60	SAON/60	MAON/60	BFON/60	Fenwick-/60
BENNFERRIERO	35	0.02	NaN	0.00	0.00	0.00	112.50	0.00
BRADWINCHESTER	67	0.02	1000	0.00	46.20	0.00	0.00	46.20
LOGANCOUTURE	80	1.06	836	6.39	32.60	15.60	31.90	48.20
DANIELWINNIK	84	2.42	877	4.42	31.50	25.30	26.50	56.80
PATRICKMARLEAU	82	1.46	867	6.03	39.20	19.60	31.60	58.80
ANDREWDESJARDINS	76	0.84	882	5.63	42.20	16.90	24.40	59.10
T.J.GALIARDI	69	0.95	860	7.29	44.70	14.60	16.40	59.30
TORREYMITCHELL	76	0.62	808	12.68	53.30	8.90	30.40	62.20
JOEPAVELSKI	82	1.52	830	8.20	40.10	24.10	31.40	64.20
DOMINICMOORE	79	1.61	822	8.96	41.50	27.30	26.90	68.80
JOETHORNTON	82	1.05	887	5.56	43.80	25.00	25.70	68.80
TOMMYWINGELS	33	0.05	1000	0.00	70.60	0.00	70.60	70.60
MICHALHANDZUS	67	1.34	864	8.03	50.80	21.40	31.40	72.20
ANDREWMURRAY	39	0.38	737	20.32	56.90	16.30	24.40	73.20
MARTINHAVLAT	39	0.58	864	7.90	50.00	23.70	44.80	73.70
RYANECLOWE	76	0.03	1000	0.00	56.70	28.30	28.30	85.00
MEANS				6.34	43.76	16.69	34.83	60.44

2012-2013
NAME	GP	TOI/60	SVPCT	GAON/60	SAON/60	MAON/60	BFON/60	Fenwick-/60
BRACKENKEARNS	1	1.20	NaN	0.00	0.00	0.00	0.00	0.00
SCOTTGOMEZ	18	0.01	NaN	0.00	0.00	0.00	0.00	0.00
RYANECLOWE	22	0.01	NaN	0.00	0.00	0.00	0.00	0.00
JAMESSHEPPARD	17	0.02	NaN	0.00	0.00	0.00	0.00	0.00
TIMKENNEDY	12	0.03	NaN	0.00	0.00	0.00	0.00	0.00
MARTINHAVLAT	20	0.40	1000	0.00	37.90	0.00	45.50	37.90
T.J.GALIARDI	15	1.17	818	6.86	30.90	10.30	20.60	41.20
TOMMYWINGELS	18	1.36	800	7.34	29.40	14.70	29.40	44.10
JOETHORNTON	24	0.59	1000	0.00	33.90	12.70	16.90	46.60
MICHALHANDZUS	24	1.83	871	5.47	36.90	13.70	27.30	50.60
PATRICKMARLEAU	24	1.58	875	4.73	33.10	18.90	15.80	52.00
ANDREWDESJARDINS	18	1.31	944	2.54	43.30	12.70	28.00	56.00
JOEPAVELSKI	24	1.88	892	5.32	43.90	22.60	17.30	66.50
LOGANCOUTURE	24	0.89	1000	0.00	53.30	14.00	44.90	67.30
ADAMBURISH	24	1.35	880	5.56	40.80	27.80	29.70	68.60
MEANS				2.52	25.56	9.83	18.36	35.39

Big-ass table, yo.

Based on this, we could argue that Joe Pavelski has struggled on the PK. Opposing players are getting lots of shots off (on goal and missed), and his on-ice block rate is relatively low. Pavelski's Fenwick- last year was 64.2, which suggests this might be his normal level of PK play given that this year it was at 66.5. (See below for what I think should happen.)

Adam Burish has a high Fenwick- as well, but his on-ice block rate is relatively high, so this may account for his perhaps only seemingly positive effect on the PK. The high Fenwick-, to me, means that he has not quite jelled with his PK unit.

The same maybe can be said for Andrew Desjardins. Lots of shots, lots of blocks. Notice that Desi's absurdly low GA On/60 is really an effect of a standout Save% by Antti Niemi. The converse is true for T.J. Galiardi and Tommy Wingels, who have the best Fenwick- numbers. Gotta give Wingels huge props as he was hardly deployed as a PK forward last year and has stepped up. With the numbers here, there's a case to be made that Galiardi and Wingels may have played the best PK minutes among the forwards so far this year. (So, forget my initial conclusions!)

One other interesting thing these Fenwick- numbers can tell us is in relation to last year. With the exception of Pavelski, every single PK forward improved in terms of Fenwick-. The biggest improvement from last year to this year has been Michal Handzus (-22). I think that's right. (I mean, he was kinda awful last year.) Galiardi and Wingels also saw significant double-digit jumps. Desi's TOI also saw a significant jump and his numbers are slightly better than last year's. Clearly, the forwards are doing a better job, and the young players are stepping up in positive ways.

As a result (and maybe this is where Zus and Burish might be worth it, though other younger cheaper options have actually played better), what we are not seeing is the offensively talented Joe Thornton, or last year's Fenwick- leader Logan Couture being deployed with the regularity they were last year (check out their reduced TOIs). In doing so, these guys can focus on, you know, scoring goals (and maybe Pavs could be deployed less in PK situations as well?).

[THIS totally worked vs LA last night.]

So there were a couple pieces (here and here) that looked at penalty killing attributes using some behindthenet.ca statistics.

So, here are the assumptions and the argument I want to offer:

2011-2012 (means)	2012-2013 (means)
GA On/60: 6.34	GA On/60: 2.52
SA On/60: 43.76	SA On/60: 25.56
MA On/60: 16.69	MA On/60: 9.83
BF On/60: 34.83	BF On/60: 18.36

Yup, table of nombres, part I. Bigger table below, you guys.

2011-2012
NAME	GP	TOI/60	SVPCT	GAON/60	SAON/60	MAON/60	BFON/60	Fenwick-/60
BENNFERRIERO	35	0.02	NaN	0.00	0.00	0.00	112.50	0.00
BRADWINCHESTER	67	0.02	1000	0.00	46.20	0.00	0.00	46.20
LOGANCOUTURE	80	1.06	836	6.39	32.60	15.60	31.90	48.20
DANIELWINNIK	84	2.42	877	4.42	31.50	25.30	26.50	56.80
PATRICKMARLEAU	82	1.46	867	6.03	39.20	19.60	31.60	58.80
ANDREWDESJARDINS	76	0.84	882	5.63	42.20	16.90	24.40	59.10
T.J.GALIARDI	69	0.95	860	7.29	44.70	14.60	16.40	59.30
TORREYMITCHELL	76	0.62	808	12.68	53.30	8.90	30.40	62.20
JOEPAVELSKI	82	1.52	830	8.20	40.10	24.10	31.40	64.20
DOMINICMOORE	79	1.61	822	8.96	41.50	27.30	26.90	68.80
JOETHORNTON	82	1.05	887	5.56	43.80	25.00	25.70	68.80
TOMMYWINGELS	33	0.05	1000	0.00	70.60	0.00	70.60	70.60
MICHALHANDZUS	67	1.34	864	8.03	50.80	21.40	31.40	72.20
ANDREWMURRAY	39	0.38	737	20.32	56.90	16.30	24.40	73.20
MARTINHAVLAT	39	0.58	864	7.90	50.00	23.70	44.80	73.70
RYANECLOWE	76	0.03	1000	0.00	56.70	28.30	28.30	85.00
MEANS				6.34	43.76	16.69	34.83	60.44

2012-2013
NAME	GP	TOI/60	SVPCT	GAON/60	SAON/60	MAON/60	BFON/60	Fenwick-/60
BRACKENKEARNS	1	1.20	NaN	0.00	0.00	0.00	0.00	0.00
SCOTTGOMEZ	18	0.01	NaN	0.00	0.00	0.00	0.00	0.00
RYANECLOWE	22	0.01	NaN	0.00	0.00	0.00	0.00	0.00
JAMESSHEPPARD	17	0.02	NaN	0.00	0.00	0.00	0.00	0.00
TIMKENNEDY	12	0.03	NaN	0.00	0.00	0.00	0.00	0.00
MARTINHAVLAT	20	0.40	1000	0.00	37.90	0.00	45.50	37.90
T.J.GALIARDI	15	1.17	818	6.86	30.90	10.30	20.60	41.20
TOMMYWINGELS	18	1.36	800	7.34	29.40	14.70	29.40	44.10
JOETHORNTON	24	0.59	1000	0.00	33.90	12.70	16.90	46.60
MICHALHANDZUS	24	1.83	871	5.47	36.90	13.70	27.30	50.60
PATRICKMARLEAU	24	1.58	875	4.73	33.10	18.90	15.80	52.00
ANDREWDESJARDINS	18	1.31	944	2.54	43.30	12.70	28.00	56.00
JOEPAVELSKI	24	1.88	892	5.32	43.90	22.60	17.30	66.50
LOGANCOUTURE	24	0.89	1000	0.00	53.30	14.00	44.90	67.30
ADAMBURISH	24	1.35	880	5.56	40.80	27.80	29.70	68.60
MEANS				2.52	25.56	9.83	18.36	35.39

Big-ass table, yo.

Link on NHL Dmen with TOI and Caphit variables

JpR-dc — Thu, 28 Feb 2013 18:26:20 -0000

Link on NHL Dmen with TOI and Caphit variables

Per a Japers Fanpost. If you want the Excel spreadsheet, which has the data he compiled, it's in the comments. Might of interest in terms of thinking about Dmen and value.

Will SJ 12-13 be NJ 07-08 (99 points) or NJ 10-11 (81 points)? (a response to ievans)

JpR-dc — Tue, 26 Feb 2013 16:34:54 -0000

ievans made a great suggestion in one of the comment threads:

"The frustration on my part here is that we’ll watch the PDO for, say, the Kings suffer for most of the year, then normalize and swing back in their favor, but the PDO for the Sharks last year didn’t. We rode the PDO train to 7 wins (yay!), and expected things to regress a bit, but instead it swung back even more violently, and now we have a losing record.

What I really want to know is if any other similar team has experienced this kind of extended poor luck as shown by the advanced stats. I’m not talking about bad teams performing badly. I’m talking about a team with solid underlying numbers vastly underperforming year to year. Because it feels like we’re the outlier here in the shitty direction, that probably indicates some underlying fact not captured in the stats themselves."

Here's a preliminary result that may help. I used the 150 team-year data drawn from behindthenet.ca (5 seasons, 30 teams).

1) I took the Sharks' current (through 17 games) FenClose, GF/60, and SF/60 (52.61%, 2.18, 25.36) as an indicator of 1) underlying puck possession ability and 2) shot-creating and goal-scoring ability (you know, offense). Remember, this is all 5v5 with Close scores because this is most hockey play and where puck possession and offense measures most usefully reflect team ability. I then did a spit-ball formula to get the teams with the closest figures to SJ 2012-2013 just by doing a sum % deviation from those statistics (FenClose, GF/60, and SF/60).

So I tried to do what ievans wanted. What team has comparable FenClose, GF/60, and SF/60 in these conditions? What was their corresponding PDO and point total? Can a team have solid "underlying numbers" and underperform? Does it happen often?

2) Result: two teams were really, really close. And they are from the same franchise. New Jersey Devils 2007-2008 and 2010-2011. Crap GF/60, low-mid range SF/60, good FenClose (top 10).

3) What were the PDOs and Season points for NJ 07-08 and 10-11? 07-08, 99 points, 996 PDO. 10-11, 81 points, 978 PDO.

4) This means what? While this does not answer ievans' question entirely, the numbers clearly show that if PDO stays around 1000, a low goal-scoring season, MAY NOT lead to a collapse in point total given a good FenClose. In fact, NJ 07-08 scored fewer goals/60 than SJ 12-13, and they even allowed more goals. Their PDO was 996 and ended with 99 points. NJ 10-11 ended with 81 points. The difference, in numbers, between NJ 07-08 and NJ 10-11 is PDO (in particular, SV%). If PDO remains stable (that is, if SH% stays in the gutter and SV% stays in the clouds OR a likely increase in SH% with a corresponding decrease in SV%), SJ 12-13 could very well be NJ 07-08 (which ain't so bad, man).

5) Does high FenClose happen often in conjunction with low GF/60, low-mid SF/60, PDO~1000, and low point total? I put a limit of +/-1% FenClose relative to SJ 12-13. The only team besides NJ 10-11 to have a point total below 90 was CMB 07-08. Their PDO? 985. So, again a PDO collapse (notably SV%).

6) So, all of this should be taken with salt as I haven't done all of the math required to conclude that if SJ 12-13 does end with a low point total, it would have company with two other teams out of 150 (er, it would be 3 out of 180, ~2%ish). However, so long as PDO does not collapse, I would say SJ 12-13 has a real good shot at being NJ 07-08 rather than NJ 10-11. I mean, that has to count for something, right?

Table with relevant numbers below. Have at it with context, history (any NJ perspective?), and critique (warranted or otherwise) in the comments! (I guess I'll have to write that FP on game theory another time...)

5v5 FenClo	YEAR	FENCLOSE	TOI	GF/60	GA/60	SF/60	SA/60	PTS	SUMPERCDEV	PDO
S.J	2012-2013	52.61	29.09	2.18	1.46	25.36	24.27	?	0.00%	1003
N.J	2007-2008	51.9	29.37	2.12	1.94	25.66	24.89	99	5.28%	996
N.J	2010-2011	52.16	29.04	2.12	2.39	26.08	23.68	81	6.45%	978
								FenClo, GF/60, SF/60
N.J	2011-2012	51.08	31.58	2.46	2.39	25.9	25.14	102	17.88%	994
N.J	2009-2010	51.49	29.84	2.58	2.16	27.59	25.73	103	29.27%	1005
N.J	2008-2009	53.04	26.92	2.72	1.85	29.47	26.78	106	41.79%	1009

CBJ	2007-2008	52.47	27.36	2.27	2.51	26.9	24.07	80	10.47%	985

Yup. Numbers above. (Would love to see a win tonight.)

ievans made a great suggestion in one of the comment threads:

"The frustration on my part here is that we’ll watch the PDO for, say, the Kings suffer for most of the year, then normalize and swing back in their favor, but the PDO for the Sharks last year didn’t. We rode the PDO train to 7 wins (yay!), and expected things to regress a bit, but instead it swung back even more violently, and now we have a losing record.

What I really want to know is if any other similar team has experienced this kind of extended poor luck as shown by the advanced stats. I’m not talking about bad teams performing badly. I’m talking about a team with solid underlying numbers vastly underperforming year to year. Because it feels like we’re the outlier here in the shitty direction, that probably indicates some underlying fact not captured in the stats themselves."

Here's a preliminary result that may help. I used the 150 team-year data drawn from behindthenet.ca (5 seasons, 30 teams).

2) Result: two teams were really, really close. And they are from the same franchise. New Jersey Devils 2007-2008 and 2010-2011. Crap GF/60, low-mid range SF/60, good FenClose (top 10).

3) What were the PDOs and Season points for NJ 07-08 and 10-11? 07-08, 99 points, 996 PDO. 10-11, 81 points, 978 PDO.

5v5 FenClo	YEAR	FENCLOSE	TOI	GF/60	GA/60	SF/60	SA/60	PTS	SUMPERCDEV	PDO
S.J	2012-2013	52.61	29.09	2.18	1.46	25.36	24.27	?	0.00%	1003
N.J	2007-2008	51.9	29.37	2.12	1.94	25.66	24.89	99	5.28%	996
N.J	2010-2011	52.16	29.04	2.12	2.39	26.08	23.68	81	6.45%	978
								FenClo, GF/60, SF/60
N.J	2011-2012	51.08	31.58	2.46	2.39	25.9	25.14	102	17.88%	994
N.J	2009-2010	51.49	29.84	2.58	2.16	27.59	25.73	103	29.27%	1005
N.J	2008-2009	53.04	26.92	2.72	1.85	29.47	26.78	106	41.79%	1009

CBJ	2007-2008	52.47	27.36	2.27	2.51	26.9	24.07	80	10.47%	985

Yup. Numbers above. (Would love to see a win tonight.)

Game theory Fanpost over at Japers

JpR-dc — Fri, 22 Feb 2013 21:15:10 -0000

Game theory Fanpost over at Japers

A great Fanpost on dump-ins using game theory. So good I had to comment on it. ... Since I live in DC, I also watch Caps and read JP... it happens.

Comparing the underlying numbers between the current Tampa Bay and San Jose losing streaks, or: Ok, dude, when do we press the panic button?

JpR-dc — Sun, 17 Feb 2013 05:43:37 -0000

(Note: This was written before TB's game vs. FLA.)

Oh man. Was sitting around last night. Doing a lot of sitting. Sittin’ and starin’. Starin’ at NHL standings. . Even though the Sharks picked up a couple points along the way, man, *7* losses. Starin’ at streaks. . Friends in commiseration? Oh, look at you, Tampa Bay...

In one of the previews during the losing streak, SBN's Raw Charge cited two reasons for the streak and one reason to keep calm. Reasons for streak: crap puck possession, not great conditioning/stamina. Reason to keep calm: long-term view (read: big picture).

I wondered if any of this was true. Did TB have crap puck possession so far in 2012-2013? How do they compare with San Jose’s? If folks over at Raw Charge are telling people to take a deep breath given these figures, should we? Are the numbers giving delusions of grandeur? And if so, why... are... they... DOING... THIS?!

. Ok, numbers. Talk to me.

Now, first some context. Remember 2010-2011, when TB pushed Boston to 7 games in the Eastern Conference Finals? No. Well, you should. They were good! Here's (via behindthenet) their Fenwick numbers for that regular season (103 points, 5th in the East):

TEAM	TIED	UP2	UP1	DOWN2	DOWN1	CLOSE
T.B	52.89	48.49	50.27	58.75	59.66	53.68

Pretty dang solid. FenwickClose% was 3rd in the league (behind... that’s right, you guessed ‘em: SJS and CHI). 2010-2011 5v5 numbers? Yup, here dey at:

TOI	GF	GA	GF/60	SF/60	SH%	GA/60	SA/60	SV%
3799.1	166	160	2.6	31.5	9.17	2.5	27.3	90.8

About right. So this means that in most hockey play (non-special teams), TB was *good* in 2010-2011. Over the course of that season, TB had good puck possession and not driven by a hot goalie or inflated SH%. It means that if there is a difference between the numbers then and the numbers now, this is evidence that something is fundamentally wrong with the team (given the similar cast of characters).

So, what’s 2012-2013 look like for TB, and how do they compare to the Sharks’ numbers?

TEAM	TIED	UP2	UP1	DOWN2	DOWN1	CLOSE
S.J	53.93	29.41	49.43	48.20	52.50	52.50
T.B	43.36	37.10	48.41	47.22	41.03	42.99

Oh, man. FenwickClose, FenwickDown1, FenwickTied. 10% difference between the teams, and a severe drop from 2010-2011 figures. Oh, man. You guys. Look at them. Look at those numbers. Man. If San Jose has these numbers, I’d defini--

(Dodging stuff from Tampa Bay.)

Ok, ok. How about 5v5 play?

TEAM	TOI	GF	GA	GF/60	SF/60	SH%	GA/60	SA/60	SV%
S.J	623.3	20	20	1.9	27.8	9.31	1.9	28.6	93.3
T.B	594.5	33	25	3.3	28.0	8.81	2.5	31.6	92.0

This doesn’t look so bad. Shots For rate/60 minutes are about even. Goals For rate/60 minutes is great! Indeed, TB is +0.8 on goal differential. Problem is, they are giving up a lot of shots, and consequently more goals. Fenwick agrees. The real question here is something only someone following TB would be able to answer: has the identity (and strategy) of the team shifted from puck possession-cycle to firewagon-style hockey?

(Note: It is confusing that TB appears to have an even number of Shots For rate, a higher Goals For rate, and yet a lower Shooting%. I would speculate that this is because bedhindthenet Shooting% is Goals / (SOG + Missed Shots), and Shots For counts SOGs only. Indeed, this would mean that not only are they getting out-SOGged, they are giving up a lot of pucks toward the net in general, which Fenwick numbers would corroborate. But someone correct me if I’m doing it wrong. Which... probably.)

Bottom line: two points. First, let me assure Bolts fans that I’m not trollin’. I mean, you guys have a Cup for crissake. But even though you have scored the most goals overall in the league (via espn), if San Jose had your underlying figures, well, I’d be worried. This is probably because the Sharks aren’t built to be a firewagon hockey team. Their identity is to push when possible, but usually the objective is to play good entries, possess the puck in the offensive zone, and get puck at the net/into the slot from the point or in deep. Now, if you’re cool with firewagon hockey, as opposed to possess-and-cycle hockey, then the TB numbers might be fine with you. It just seems weird to be cool with that since in 2010-2011, TB was possessing the shit out of the puck.

Second. Now I’m not trying to hide behind numbers (Panic Button stirs from slumber). SJ’s lack of goals (Panic Button creeps out of closet) in 5v5 play unfort(Panic Button saunters over to the desk)unately has led to some significant worrying (Panic Button looks up at me) given the two tough road games coming up (press me, press--

Me: Listen, PB. I’m not ready for you yet. Please go away.

PB: Ok, man. Fair enough. . Does looking at the numbers help?

Me: Sortof. I think if I were TB I might be more into you, but I’m not just not there yet. Fenwick numbers still 10% above TB’s, you know. If those guys are still hopeful, why shouldn’t we? I still think everything is going to be ok once the D-men play together more and things get in sync.

PB: Hmm. Interesting. . Ever think the numbers are just deluding you?

Me: What? No. No way. That’s just crazy talk. You’re talking crazy. I’m writing this fanpost, you know, to get some of this out there, you know--and then people will agree, you’ll see. I mean c’mon, look at Tampa’s numbers. We got better numbers, we’re good. It’s all good. Really. I’m not even going to look at FTF until Tuesday. Seriously. Really. Really seriously.

PB: Ok, well, just remember, I’ll be right here by your desk. Don’t let me distract you. . Maybe see you Tuesday night, then?

Me: You just shut your mouth.

PB: You seem wound up. Maybe you should lay off the Fanpost-ing?

Me: I will drop-kick you into next week--

PB: Ok, Friday. Friday it is.

TEAM	TIED	UP2	UP1	DOWN2	DOWN1	CLOSE
T.B	52.89	48.49	50.27	58.75	59.66	53.68

Pretty dang solid. FenwickClose% was 3rd in the league (behind... that’s right, you guessed ‘em: SJS and CHI). 2010-2011 5v5 numbers? Yup, here dey at:

TOI	GF	GA	GF/60	SF/60	SH%	GA/60	SA/60	SV%
3799.1	166	160	2.6	31.5	9.17	2.5	27.3	90.8

TEAM	TIED	UP2	UP1	DOWN2	DOWN1	CLOSE
S.J	53.93	29.41	49.43	48.20	52.50	52.50
T.B	43.36	37.10	48.41	47.22	41.03	42.99

TEAM	TOI	GF	GA	GF/60	SF/60	SH%	GA/60	SA/60	SV%
S.J	623.3	20	20	1.9	27.8	9.31	1.9	28.6	93.3
T.B	594.5	33	25	3.3	28.0	8.81	2.5	31.6	92.0

Me: You just shut your mouth.

PB: You seem wound up. Maybe you should lay off the Fanpost-ing?

Me: I will drop-kick you into next week--

PB: Ok, Friday. Friday it is.

Awcrapgottooinvolved: Corsi+/scoring+ ratio and a possibly stupid prediction

JpR-dc — Thu, 14 Feb 2013 19:54:43 -0000

A couple ideas from the recent posts (Mike Chen, Preds numbers) have spurred some thoughts. (I started writing two comments and then decided oh god sunk costs so heavy do fan post.) One of the big questions that hit me is a simple one: what’s a streak? How do we know when a team is on a streak? Statistics (by my eye!) helps us organize and describe (and potentially explain, though not always) what we are seeing. At first, I thought that we only really know a streak in retrospect, not while it’s happening. But maybe I’m wrong about this.

This led me to wonder further about Corsi for (Corsi+) and scoring chances (scoring+). Perhaps it is obvious to say that these two statistics are highly correlated (seems to hold, see here), but I wondered about what the “happy” ratio would be for a winning team, and for the Sharks, in particular. The aforementioned link looks at the comparative utility of tracking shot differential versus scoring chance differential for game outcomes (bottom line: no statistically significant difference, shots tell us enough), but I haven't seen anything that uses a ratio of shots toward the net (Corsi for events) and scoring chances as a measure of efficiency (one could also calculate the ratio of Corsi for and SOG, but I’m not sure this is as good).

(Note: Underlying this proposition is that Corsi counts a lot of events, some of which are more helpful in leading to a goal than others. Scoring chances is the closest thing we have to a goal beside an actual goal. So, I’m assuming this to be a way to estimate Corsi event quality in hockey practice. Now, there is a live debate over “shot quality” (see here and here), but these are typically over whether or not this is a “repeatable” team skill and looking at shooting percentages. I’d be happy to know if someone else has already profitably tilled this rather arid patch of soil/panned this barely flowing creek for golden (but not really) flakes; or, if this really is an effect of some other force. Also, this exercise does not control for possible “score effects,” which is the tendency/behavio(u)ral pattern of leading teams to allow more shots and losing teams to take more shots. Statistics drawn when teams are tied can be used to manage this, but I haven’t run the numbers. Ok. Enough qualifying.)

So why calculate this ratio of Corsi+/scoring+ (where a “1” is ultimate efficiency, as each Corsi+ event is a scoring chance, and where “INSERT big number here” is crap)? Well, here’s a 2x2 table to guide us:

	Efficient (low number)	Inefficient (high number)
Win	Bif-urcated into two scenarios: 1) if team wins by wide margin, then team is awesome, go all-in/bet the farm for the Cup, or 2) if team wins by low margin, ohshitohshit not really as good as we thought	Ha! Poopin’ gold, suckers! Just wait ‘til we all don’t have the flu!
Lose	Team not talented enough. . . Scout WHL, watch Beanpot, etc.	Ohmanohmanohman. Team sucks real bad. Rant on FTF and demand: 1) DW be fired, 2) Larry Robinson to replace TMac, and 3) trade Marleau.

Figure 1. 2x2 = science. Yup.

Awesome. So, what data and what results? To do this properly, we need to know game outcomes and calculate corsi+/scoring+ (via the req’d reading). Would be best to do this for all teams (always in need of bigger N), but for the purpose of preliminary analysis (spit-balling) let’s just look at the Sharks on their two streaks. In addition (for the sake of total bull-shitting), let’s look only at regulation wins and losses. This also helps us understand if there’s something really wrong with the team or not (see lower right-hand box of the 2x2). That results in a very small sample (n=8!), but maybe, just maybe, indicative of some pattern.

opponent	result	ratio
EDM	W 6-3	3.33
PHX	W 5-3	3.40
ANA	L 1-2	3.62
CHI	L 3-5	3.92
CMB	L 2-6	4.57
VAN	W 4-1	4.8
CAL	W 4-1	5.5
COL	W 4-0	5.5

Descriptives for the ratio:

Mean	4.33
Median	4.25
Mode	5.5 (unhelpful in this instance)
Min/Max	3.33 - 5.5
Standard deviation	0.89

(Note: Assuming normal distribution with this small-N sample, 68% of the total population will be within 0.89 of the mean, or between 5.22 and 3.44; i.e. can typify most results to fall within this range(...ish) and perhaps can draw the high-low efficiency boundary along these two numbers.)

Right. Check out the bottom two ratios. These, perhaps in being more than one standard deviation from the mean, suggest the Sharks were straight-up gold-poopin’. Check out the top two. These, again more than one standard deviation from the mean, suggest we should bet the farm on the Sharks. Middle four ratios give mixed results.

Bottom line: ahm... no real pattern. When they win, sometimes it looks like the Sharks are talented. Other times, perhaps more luck than skill. One notable point, though, is that their regulation losses do not appear to be the result of inability to convert pucks toward the net into actual scoring chances. This is evidence (assuming this is a repeatable team skill) that there is probably nothing fundamentally flawed with this team (other than the obvious... you know... too streaky, too soft, etc.).

Final note: just for fun, one may predict that when Corsi+/scoring+ < 3.44 we’ll see a WIN, and when it is >5.22, we’ll see a LOSS. I am willing to admit that this could be stupid. Thoughts?

	Efficient (low number)	Inefficient (high number)
Win	Bif-urcated into two scenarios: 1) if team wins by wide margin, then team is awesome, go all-in/bet the farm for the Cup, or 2) if team wins by low margin, ohshitohshit not really as good as we thought	Ha! Poopin’ gold, suckers! Just wait ‘til we all don’t have the flu!
Lose	Team not talented enough. . . Scout WHL, watch Beanpot, etc.	Ohmanohmanohman. Team sucks real bad. Rant on FTF and demand: 1) DW be fired, 2) Larry Robinson to replace TMac, and 3) trade Marleau.

opponent	result	ratio
EDM	W 6-3	3.33
PHX	W 5-3	3.40
ANA	L 1-2	3.62
CHI	L 3-5	3.92
CMB	L 2-6	4.57
VAN	W 4-1	4.8
CAL	W 4-1	5.5
COL	W 4-0	5.5

Descriptives for the ratio:

Mean	4.33
Median	4.25
Mode	5.5 (unhelpful in this instance)
Min/Max	3.33 - 5.5
Standard deviation	0.89

A kind of experiment in extremely small sample sizes (n=1)

JpR-dc — Mon, 11 Feb 2013 05:12:46 -0000

Suppose you wanted to compare two players in terms of their value toward winning games. Suppose you wanted to see what one game (one game!) could tell you. In addition, suppose you wanted to compare what you saw by eye, and the numbers from the RTSS sheets (which are great, but you have to get into a rhythm to read them to draw any meaning out; otherwise it's a headache-inducing mish-mash of numbers and names).

Well, for this Phoenix game, I thought I'd try to focus on Boyd Gordon and Michal Handzus and see if this quasi-experiment could turn up anything. There's some element of control here, in that both are face-off winning centers that play somewhat similar roles (bottom-6, PK). I was able to watch only part of the game, as I had to multi-task with a trivia meet-up (which is daft I know, there's a so much stuff going on in hockey--SO MUCH STUFF).

By the eye, Gordon seemed to be on the ice a lot on the PK. He didn't seem dominant in the circle. He seemed to be busting his ass pretty well (which, as my wife tells me, is what these Saskatchewan dudes are all about). Handzus seemed engaged. That's about all I can say.

Looking at the numbers provided by nhl.com's RTSS sheets:

NAME	G	A	P	+/-	PN	PIM	TOT	SHF	AVG	PP	SH	EV	S	A/B	MS	HT	GV	TK	BS	FW	FL	F%
GORDON	0	0	0	0	1	2	16:39	28	00:35	00:23	02:41	13:35	0	0	2	0	0	2	0	13	12	52
HANDZUS	0	0	0	0	1	2	14:24	25	00:34	00:27	00:45	13:12	3	1	1	1	1	0	0	9	4	59

They look fairly even, except for the shots and missed shots (positive Corsi events) in favor of Handzus and giveaway/takeaway in favor of Gordon. Also, Gordon's SH time is significant. Both players played relatively short shifts compared to teammates. I took a look at the face-off numbers and I didn't expect this:

	off	def	neutral	overall
vs. 15	3-3 / 100%	1-1 / 100%	0-1 / 0%	4-5 / 80%

Against Gordon, Handzus won 4 out of 5 face-offs including 3 in the offensive zone. Didn't remember this by eye, but it's certainly meaningful. The FTF compiled Corsi numbers (here) also bear out quite positively for Handzus. Now, obviously this is all cursory, but I think Handzus definitely demonstrated more value, though perhaps not as much by the eye (and maybe not at his cap hit; Gordon's is 1.325 vs. 2.5).

Surprising thing I found was the way that my eye didn't remember the face-off differential even though I was watching. In that sense, the experiment suggests that while eye evaluation is helpful, so are the tracking numbers, even the most simple ones. This might be completely obvious to some of you, but with hockey still resisting analysis beyond "didn't keep his stick on the ice" and "goalies these days just go down too soon," etc., it remains an important point.

Before I forget, an interesting read: Caps coach Adam Oates insinuates that they track rushes and whether or not there are dump-ins... and that dump-ins are bad (here).

Looking at the numbers provided by nhl.com's RTSS sheets:

NAME	G	A	P	+/-	PN	PIM	TOT	SHF	AVG	PP	SH	EV	S	A/B	MS	HT	GV	TK	BS	FW	FL	F%
GORDON	0	0	0	0	1	2	16:39	28	00:35	00:23	02:41	13:35	0	0	2	0	0	2	0	13	12	52
HANDZUS	0	0	0	0	1	2	14:24	25	00:34	00:27	00:45	13:12	3	1	1	1	1	0	0	9	4	59

	off	def	neutral	overall
vs. 15	3-3 / 100%	1-1 / 100%	0-1 / 0%	4-5 / 80%

Before I forget, an interesting read: Caps coach Adam Oates insinuates that they track rushes and whether or not there are dump-ins... and that dump-ins are bad (here).

The paradox of believing in face-off win value and that Handzus is terrible at hockey

JpR-dc — Fri, 08 Feb 2013 21:20:17 -0000

First fanpost, long-time Shark fan (i.e. when people talk about the Ray Whitney goal in the double OT Game 7 against the Flames, I add, "and Wade Flaherty had 56 saves!").

Something about all the talk regarding Handzus struck me.

(Note: when critiquing players in the By the Numbers posts, Handzus appears most frequently. For those of you counting, he shows up 10 out of 25 total cases of player critique (out of 59 total bullet points) and the closest others were "vlasic-stuart" and "thornton/topline"; not surprising given Neutral's rather open--and not completely irrational--bias.)

So, hear me out on this.

1) We still think there is some value with face-off wins. Indeed, this is at least part of the reason why the Sharks pursued and signed Handzus in the first place (here). Data is jumbly for sure, but one (pretty magisterial) post here and elsewhere suggests that face-off wins are still useful in some respect for explaining and predicting puck possession. At least, both seemed unwilling to give up on face-off wins as part of accounting for possession.

2) Note, in addition, that other players we think are good, have good face-off numbers (per ESPN; FO, W, L, %):

Thornton: 115, 66, 49, 57.4
Couture: 105, 60, 45, 57.1
Pavelski: 106, 59, 47, 55.7

Also, they have good Corsis--and I believe that we think this relationship holds true.

3) Yet Handzus has very comparable face-off numbers (140, 79, 61, 56.4), and his Corsi is crap.

It should be noted that for players who take this many face-offs (let's say >100), Handzus is 15th in the league in %. (And much better than his numbers last year.) So, relative to other draw-takers, he should have value, but we don't have a statistic for this, do we?

(Note: In running some individual correlations between Handzus' Corsi+, Corsi +/- (as per FTF) and face-off percentages (off, def, neutral, total)... that's right, all negative/weak with one exception: Corsi+ and off%... and though this was positive, it was weak.)

4) Therefore, what gives? Do we reject the thought that face-off wins have no impact on Corsi? Seems unlikely. If we don't reject this, are there other explanations for Handzus' Corsi other than "dude sucks"? My intuition is that it's got something to do with zone entry, but I haven't seen numbers on this. Disclaimer: I don't think Handzus is particularly good, but I'd like to be able to quantify his value in some way.

Bottom line claim: it's paradoxical to think that Handzus sucks and also believe there to be a positive relationship between face-off wins and Corsi+. Thoughts?