passing project – McKeen's Hockey

Kats Krunch: Karlsson Wins Norris

Gus Katsaros — Sat, 19 Mar 2016 03:58:55 +0000

Erik Karlsson should win the Norris Trophy in 2016

I wanted to analyze the Norris race using different criteria unexamined outside of traditional box-score or shot based metrics, to highlight the importance of Erik Karlsson, Drew Doughty and Brent Burns.

I never thought I would see another defenseman come close to scoring 30 goals since Mike Green did it in 2008-09 but here we are. The Senators and Sharks defensemen are on the brink of achieving an NHL feat not accomplished since 1973-74.

Five kilometers south from Burns in San Jose resides another Norris contender, Drew Doughty, Karlsson’s fiercest competitor.

I’m going to use some Passing Project data (who have introduced a shiny new little visualization) to offer an unexplored alternative.

Before that, I want to touch on the defensive debate. Erik Karlsson doesn’t play penalty killing minutes, while analysts laying more importance to a structural part of the game, rather than individual skill and smarts.

Let’s start with the penalty kill. If you were the owner of the Senators, would you want to see Erik Karlsson taking multiple shots to the ankles in an effort to save a power play goal? How about the GM? Coach? How about a fan? How much is a blocked shot worth in comparison to Karlsson generated scoring chances?

Karlsson is so effective playing 5v5 and 5v4 minutes, it only makes sense to limit the penalty kill time in an effort to capitalize on prime assets. It’s what a good coach would do. Shorthanded minutes would be detrimental, not fundamental. A traditional two-forward/two-defenseman penalty killing unit doesn’t even need to be structured in that way if players can fit the roles. Theoretically, assuming adequate backwards skating and they’re taught to flawlessly slot into the defensive zone setup as a blueliner, any forward could do the job. Interchangeable parts should be the fundamental basis even if that means using untraditional roles for players. Clearly, using a slew of forwards as a penalty kill unit is extreme, but it can be done.

Any defenseman can play the penalty kill. The fact Karlsson doesn’t appear at 4v5 could very well be due to the influence of the analysis I’m identifying here today. The Senators must already know the value and have built strategy around the central idea.

Karlsson’s offensive skillset can carry the play individually so effectively, whether rushing the puck or manning the point. Defensively, Karlsson plays the support role, moving the puck in transition with his partner (or backside pressure, or some other contributing factor) providing engage.

In Karlsson’s breakout season, Filip Kuba was instrumental in achieving this support/engage dynamic, the Czech doing the heavy defensive lifting, the primary engage to jostle pucks loose, while Karlsson provided the transition and offensive spark.

To use him in situations only deemed appropriate by tradition or common assumption is to misuse his talent. This basic concept: if I have the puck, your team is playing defense.

Let’s take a closer look at some data to see what I mean.

Small Sample Alert

Data is assembled via War-On-Ice.com and the passing project – albeit, there is a very small sample size here to work with from all three teams that have been tracked. LA only has 10 games tracked, while Ottawa has 13. San Jose has 21 games, which is actually above average for the NHL but there’s a distinct theme here supported by some shot-based metrics, for merit.

Passing Project data tracks passes ending up in shot attempts, with three passes recorded (along with originating zone and lane) prior to the end event. We could isolate each defenseman’s tracked record as the tertiary, secondary or primary passer – as well as shooter which could benefit Burns most, when tallying all shots at the team level.

All data is Score Adjusted 5v5 to limit the effects of special teams. Using war on ice data I took the players stats for tracked games, with results in the table. There’s lots happening here, but TOI is the sum of time on ice. iCF is the average individual Corsi For.

Player	TOI	iCF	SF%Rel	CF%Rel
Burns	431.10	5.86	-2.26	-0.08
Doughty	181.40	4.00	1.72	0.57
Karlsson	264.90	3.92	10.95	8.49

Take notice of that gap among the Norris contenders in Shots For relative to team (SF%Rel) and corresponding relative Corsi For% (CF%Rel). Both Karlsson and Doughty averaged similar individual Corsi Events, with Burns the shooter edging them both out.

At the team level, the table for games tracked. The column %Shots By Passes is the representative ratio of shots on goal generated from a pass event prior to the shot.

Team	CF%	GF%	SF	%Shots by Passes
L.A	57.28	56.34	237.43	69.07
OTT	46.49	38.19	272.79	75.52
S.J	49.34	49.02	507.22	70.78

To no one’s surprise, the Kings dominate the Corsi battle, with Ottawa lagging in shots and goals for ratios. Summing the passing project data, The Senators top the percentage of shots via at least a pass.

Once again, sample size, but my guess would lean towards either less passes leading up to a shot event for Ottawa – almost as if they overly rely on singular players to generate shot attempts – while the Kings and Sharks have other scoring options and system implementations to spread the responsibility around.

The table below filters for passes ending as a shot on goal. Each player’s individual contributions as the shooter (sh), primary (A1), secondary (A2) and tertiary (A3) passers are listed in columns. The Percent column represents percentage of pass contributions to total team shots. The Shot% column is individual player shots as a percentage of total team shots. No surprise Burns leads here with almost 12% of the Sharks shots from a pass.

Erik Karlsson was a contributor to 55% of the shots on goal in the games tracked. Over half of the Senators shot generation involved Karlsson at some level. Half.

Shots	Team Shots from Pass	Sh	A1	A2	A3	Percent	Shot%
Doughty	164	11	11	5	1	17.07	6.70
Karlsson	206	10	56	29	20	55.83	4.85
Burns	359	42	37	32	14	34.82	11.70

Let’s change it to shot attempts, that didn’t end as a shot on goal, represented by the table below and whoa … Erik Karlsson.

non SOG attempts	Team Shots from Pass	Sh	A1	A2	A3	Percent	Shot%
Doughty	175	19	19	7	4	28.00	10.86
Karlsson	173	46	55	28	14	82.66	26.59
Burns	354	104	42	18	19	51.69	29.38

We see two different things here. First, the incredible 82.66 contribution rate (if only Ottawa had finishers) and the gap between Shot% closing with Burns. Doughty must be playing too much defense, or penalty killing minutes to make almost one-third of the impact Karlsson does.

The Kings play a ‘heavy’ team game, with scoring/shooting opportunities created from cycles, zone time and quick strikes. Postulating removing Doughty and the Kings offensive game doesn’t suffer as much as removing Burns or Karlsson.

A summation of shot attempts values Karlsson’s contribution to the Senators 5v5 shot generation three times that of Doughty’s contributions.

Karlsson contributed to 73% of his passes turning into a shot attempt at the team level. How can that not spell N-O-R-R-I-S?

Total Shots	Team Shots from Pass	Sh	A1	A2	A3	Percent	Shot%
Doughty	339	30	30	12	5	22.71	8.85
Karlsson	379	75	111	57	34	73.09	19.79
Burns	713	174	79	50	33	47.12	24.40

Once more, with feeling. Small sample. But there is something there and I’m dying to see this race with a full season’s data that would include less results disparity.

So there you have it. Go ahead and cite the ‘doesn’t play defense’ or ‘lacking penalty killing minutes argument’ but you should ask yourself first: are those things equivalent to driving three of every four team shot attempts at even strength toward the opposition net?

If so, I’ll change my mind.

Erik Karlsson deserves and will win the Norris Trophy.

Follow the McKeen's team on Twitter:************************

@mckeenshockey
@KatsHockey

Kats Krunch: 5v4 Offensive Zone Time

Gus Katsaros — Fri, 11 Mar 2016 21:03:34 +0000

The Tampa Bay Lightning spend a lot of time in the offensive zone during 5v4 man advantages. The Chicago Blackhawks spend a lot of time in the offensive zone during 5v4 power plays. this isn’t new, in fact traces are evident from earlier in 2015-16, and even as far back as early 2014-15.

Tampa Bay ranks among the bottom five in power play efficiency; Chicago is battling Washington and Anaheim for the league’s best.

Randomness abides (the Lightning ranked in the middle for GF60 in 2014-15, and dead last in SF60, while GF60 is and thrashing around the bottom five in ’15-16) while SF60 ranks at the bottom of the NHL. Chicago ranked 20th in GF60 in ’14-15 and 14th in SF60 in ’15-16. In ’15-16 the Blackhawks ranked 26th in SF60 and sixth overall in GF60.

Offensive zone time clearly matters, but only with the constructive ability to create offensive chances. In other words, there has to be progression to scoring chances, not just endless cycling or keep away with the puck. I took a recent look at the effects of multiple passes in the offense of zone at 5v5 on shooting percentage. Intuitively, the greater the propensity of sustained zone time with multiple passes greatly increases the chance of one of the shots ending up as a goal.

Unless you’re the Tampa Bay Lightning.

In that study, the Lightning are shooting a meager 2% from multiple passes in the offensive zone prior to the shot event. At 5v4, they don’t fare much better as their 26th ranked power play suggests. Chicago on the other hand, have a fairly consistent approach to zone time across a variety of situations, and the 5v4 power play has benefitted this season.

I dipped back into the Passing Project data, where a bunch of people are tracking passing events for NHL teams. I’ve spoken about and written about this project ad nauseum and I’m bullish with their latest data released the day after the NHL trade deadline. The data has 270 odd games tracked with some teams significantly represented more than others, so there are limitations (and sample sizes even with these limitations are clearly small). Some actionable data exists for first glimpses, but we need to remain conscious of the fact that this is a very small data set. For this writing, workable data within the passing project is restricted to teams with above average games tracked, instead of the entire NHL where two-thirds of the league teams lack significant results.

The study for shooting percentage was based at even strength (5v5), my curiosity led to effects of offensive zone time at 5v4. I didn't want the focus to be strictly on shooting percentage here, however, instead I wanted to investigate teams offensive zone time at 5v4. A focused special teams blog headed up by Arik Parnass is examining various scenarios of the power-play – Arik is showing how the 20% of the game matters – however sustained zone time doesn't seem to be one area that's been touched yet.

For this purpose, nine teams from the passing project, almost a full one third of the NHL has a greater than average games tracked, the list including Boston, Chicago, Dallas, Edmonton, New Jersey, San Jose, Tampa Bay, Toronto and Washington.

Passing project data fortunately tracks up to three passes prior to any shooting event, be it shot on goal or missed/blocked shots. The project also tracks where the passes originated from, including the side of the ice in addition to zone. In the table below all the per/60 data is courtesy of War-On-Ice specifically isolated for only the gains that have been tracked – so keep in mind the sample.

Using the sequencing feature, we could start to isolate components of the power-play, such as the rate at which teams make one to three or more passes prior to shooting events. With more data, this could eventually be used as a proxy for zone time, or even cycling based on the sequence of passing.

The table looks like it does below.

							Raw			Per Game
Team	GP	CF60	SF60	HSCF60	Sh%	TOI	3 OZ Pass	2 OZ pass	1 OZ pass	3 OZ Pass	2 OZ pass	1 OZ pass
BOS	21	122.06	56.26	22.56	13.86	4.86	32	24	13	1.52	1.14	0.62
CHI	49	84.42	47.83	19.07	20.83	4.63	75	21	21	1.53	0.43	0.43
DAL	27	103.25	57.11	20.34	13.22	4.97	37	40	26	1.37	1.48	0.96
EDM	20	98.00	56.40	27.91	13.46	5.12	28	12	12	1.40	0.60	0.60
N.J	50	88.87	45.62	16.90	14.18	4.98	72	37	41	1.44	0.74	0.82
S.J	21	101.56	53.80	23.91	16.08	5.46	42	23	20	2.00	1.10	0.95
T.B	33	84.52	42.57	16.99	14.82	5.68	68	23	11	2.06	0.70	0.33
TOR	24	111.54	55.90	35.29	11.74	5.53	42	26	23	1.75	1.08	0.96
WSH	29	109.17	61.38	18.20	13.35	5.19	66	29	26	2.28	1.00	0.90

Right off the bat we see is separation from some teams in all the per 60 category such as CF60 and SF60, but it's also very impressive to see how the passing data potentially correlates to the shooting metrics expressed in the per/60 data.

For instance, Tampa Bay generates 84.52 Corsi events per 60 at 5v4. Incorporating passing data, we can see the Lightning get a lot of zone time that includes plenty of passes, but generate very few actual shooting events. They had a (data-set sample size) low of 42.57 shots for every 60 min, highlighted by a 16.99 high danger scoring chances per 60, meaning there aren’t a lot of shooting events from high danger scoring areas. From the sample that we've been using, 67% of Tampa Bay’s shot events originate from three or more offense of zone passes, while a meager 11% are derived by one solitary pass in the offense of zone. That’s not necessarily a bad thing after all. Teams shouldn’t be trying to gain the zone and fire aimlessly for the sake of getting shots on goal, but sustained zone time is about creating lanes for shooting opportunities.

We can further surmise teams ability to gain the zone and keep it. In the case of the Edmonton Oilers, they seem to want to gain the zone and retain possession, but they only seem to be able to generate about 9.33 events of three or more passes for each 60 minutes. Contrast that to a shot event happening once almost every minute for the Blackhawks and Lightning, Edmonton doesn’t seem capable of extended zone time. The table below breaks down the per60 for events by each primary event.

Toronto leads the sample with 35.29 HD scoring chances per 60, with a split among all three pass categories. The Leafs like to get the shot on net after setting up and crash hard to look for rebounds and second chances.

Team	GP	3 OZ/60	2 OZ/60	1 OZ/60
BOS	21	11.20	8.40	4.55
CHI	49	61.25	17.15	17.15
DAL	27	16.65	18.00	11.70
EDM	20	9.33	4.00	4.00
N.J	50	60.00	30.83	34.17
S.J	21	14.70	8.05	7.00
T.B	33	37.40	12.65	6.05
TOR	24	16.80	10.40	9.20
WSH	29	31.90	14.02	12.57

Similarly the Stanley Cup champions Chicago Blackhawks, generate 64% of their power plays with zone time based on three or more offensive zone passes, while lacking the shooting frequency of some other teams in the data set. With an 84.42 Corsi For per 60 rate, they rank lower than the Lightning among teams here, and only 19.07 highh danger shot events per 60 minutes. This itself can be further studied for effect, but its out of scope of the theme of this blog post.

It's clear, more shots will lead to eventual scoring chances and then goals, however Chicago leads the league in power-play efficiency, and they do it based on sustained zone time rather than just a barrage of shots in ’15-16.

Once again, with more data across the board for more teams, the 5v4 offensive zone time study coupled with zone entry data could prove vital to analyze difficulties on the power play. In the absence of RFID technology and resultant data, this is likely the best tool available to analysts.

Follow the McKeen's team on Twitter:************************

@mckeenshockey
@KatsHockey

Kats Krunch: Examining the Stretch Pass

Gus Katsaros — Sat, 05 Mar 2016 14:40:45 +0000

In a post about drop passes to enter the zone on the power play in Arik Parnass special teams project , he commented about how a stretch pass leading to more quick strikes than sustained zone time.

The quote is here, but I would recommend reading that post for the effects of a drop pass in the neutral zone, personal no-no if I was coaching any team, regardless of the effectiveness.

Here’s Arik’s definition of a stretch pass.

My definition of a stretch pass was a little bit subjective, but I think hockey guys would agree that you kind of know one when you see one. Generally speaking, a stretch pass is a pass that comes before the defensive blue line that tries to stretch the defense, generally hitting a guy with considerable speed in motion at the offensive blue line.

Predictably, stretch passes lead to the fewest average seconds in the offensive zone. They are the highest risk plays and strive for shot quality on the rush over zone time.

I would include passes from the neutral zone, because of the nature of hitting the man on the fly while the speed element backs off the opposition in anticipation of breaking up a streaking man into the defensive zone. Fortunately, subjective as it may be, we have some data for that.

We can explore the stretch pass, considering there’s a small(ish) set of workable data via the passing project. Now we can isolate stretch passes and investigate their impact … if any.

Quick strikes from both teams leads to firewagon hockey, fun to watch but a nightmare to coach, and reliant on both teams abandoning defensive structure, looking for quick rushes the other way, or trying to catch the opposition off guard.

The 2013-14 Toronto Maple Leafs were the epitome of quick strike hockey, trying to take advantage of speed and shots from the rush, forsaking sustained zone time and pressure. Needless to say, the strategy, while indicating some early success, failed miserably.

Clearly, the style is not a sustainable model.

But do stretch passes make any difference? Let’s get into the passing project data for aggregation and context.

The data fortunately allows for event sequencing (labelled as A1, A2, or A3) along with originating zone (A1 Zone, A2 Zone, A3 Zone), beginning with three passes from an eventual shot event, a shot on goal, or shot directed to the net (missed/blocked) and a goal.

The project, to a detriment, doesn’t record passes that don’t lead to an eventual shot event, a void to analysis, since there is value in determining why or how plays broke down into non-events. The non-event could provide quite the bit of value, but with the amount of work involved and all done on a volunteer basis places a great onus on trackers to be focused.

Using the entire dataset from the project’s latest release, I looked at two types of stretch passes at 5v5:

a primary pass preceding an eventual shot event designated as a stretch pass as the from the defensive or neutral zone, stamped with a ‘s’ in the zone data
a primary pass prior to a shot event originating from the offensive zone, after a stretch pass allowed for entry into the zone either from the defensive or neutral zone.
If applicable, the percentage of rebounds generated from both above categories.

Starting with the table below, I’ve separated stretch passes into two categories we can calculate the shooting percentage or efficiency via each event.

Getting a stretch pass and fring a shot on goal has led to an 8.35% shooting percentage. When there’s a pass in the offensive zone after the initial stretch pass, shooting efficiency drops to 5.62%.

Situation	# of Events	SOG sh%	Stretch Pass sh%
Stretch Pass	587	8.35%	5.62%
Stretch + 1 OZ Pass	265	7.75%	4.41%
Rebounds	21/12	3.58%	4.53%

The problem is unfortunately is one of small samples. The Passing project has tracked approximately 270 games, or about 22% of 2015-16 games. With 587 events, there’s approximately 2.2 stretch pass events that leads to a shot on goal.

Intuitively, the results make sense. There’s a greater chance at scoring a goal off a solo rush and quick strike via a stretch pass. When it comes to rebounds, quick strikes lead to less rebounds, by a small percentage difference from a rebound off an offensive zone pass before the shot. An interesting note, of the 21 rebound events recorded - 11 shots on goal - with 10 players getting their own rebound, while 10 times a trailer, or another player, stealthily, or via a speed burst got to the net to take advantage of any loose pucks – without generating a shot on goal.

With the limitations in the amount of tracked games, it’s difficult to attain uniformity among all NHL teams, but we can do an estimate of the amount of stretch passes allowed per game at 5v5. The table below calculates stretch passes per game taken and allowed by all NHL clubs. The amount of individual games tracked is included for context.

From the project data, it seems like San Jose is the biggest culprit on both ends, taking advantage of stretch passes while allowing the most per game. Giving teams chances to score, even if it’s a couple of times per game at 5v5 could be detrimental.

Sticking to California, Los Angeles seems to allow a couple of stretch passes per game in comparison to the amount of passing events they generate (we know of the grinding style LA utilizes, featuring effective zone time and shot generation), while an hour down the way, Anaheim is among league leaders in generating scoring chances from stretch passing.

Tm	GP	Pass/GM Taken	SOG/Gm Taken	Pass/Gm Allowed	SOG/Gm Allowed
ANA	9	1.89	1.44	1.11	0.67
ARI	8	1.38	0.88	1.63	1.38
BOS	21	1.19	0.86	1.52	1.00
BUF	14	1.36	0.86	1.29	0.86
CAR	15	1.20	0.87	0.67	0.47
CBJ	14	1.36	0.93	1.50	1.21
CGY	15	1.07	0.67	2.00	1.20
CHI	49	1.27	0.69	0.76	0.53
COL	15	0.33	0.20	0.93	0.53
DAL	27	1.63	0.96	1.04	0.59
DET	17	1.29	0.82	1.06	0.76
EDM	20	1.15	0.80	1.85	1.15
FLA	17	0.35	0.29	0.59	0.29
L.A	9	0.78	0.56	2.00	1.56
MIN	12	1.58	0.83	1.50	0.75

Elliotte Friedman outlined in a 30 Thoughts blog about the Montreal Canadiens propensity to throw pucks into the neutral zone and fight for the puck outside of the defensive zone. It’s not exactly a stretch pass, but this is a similarity to the quick strike notion outlined in the stretch pass definition.

Tm	GP	Pass/GM Taken	SOG/Gm Taken	Pass/Gm Allowed	SOG/Gm Allowed
MTL	15	1.60	1.40	0.80	0.53
N.J	50	0.44	0.32	0.84	0.64
NSH	12	1.25	0.75	1.00	0.58
NYI	6	0.83	0.67	1.00	0.67
NYR	15	1.60	1.13	0.67	0.53
OTT	13	1.15	1.00	1.31	1.00
PHI	10	0.50	0.40	0.20	0.20
PIT	10	0.90	0.60	1.60	1.10
S.J	21	2.33	1.76	2.62	1.86
STL	14	1.57	1.07	1.07	0.71
T.B	33	0.58	0.30	0.61	0.33
TOR	24	0.79	0.54	0.83	0.63
VAN	16	0.81	0.69	1.06	0.63
WPG	14	0.71	0.43	1.43	0.93
WSH	29	0.79	0.48	0.31	0.21

This, along with a bunch of other interesting items should become a lot more clearer with a greater data set.

Get tracking.

Follow the McKeen's team on Twitter:************************

@mckeenshockey
@KatsHockey