puck on net – McKeen's Hockey

Kats Krunch: Tale of Two Power Plays

Gus Katsaros — Thu, 04 Dec 2014 21:24:13 +0000

Two teams, two power plays with two diverse results, while sharing a common affliction.

I had the opportunity to take in the Stars as they rolled in through the ACC to play the Maple Leafs on Dec 3, offering a glimpse into the abysmal Stars power play systems with some background through a live viewing to further decipher the reason for special teams futility. I had similar fortune with a live view of the Tampa Bay Lightning a couple of weeks back, offering the same opportunity to note the team’s power play systems – and potential flaws. The contrast in power play success here is striking.

They both have the same problem, a wide gap between shots on goal. Clearly, shot attempts are the best method of deciphering a power play and why it may not be firing on all cylinders, but what is presented below is a tale of two diverging successes despite the lack of shots.

This is the story of the potent Tampa Bay Lightning and futile Dallas Stars five-on-four power play.

I’m going to reserve the right to get into extensive video here, only because I’m trying to understand the power play in general on a more macro level. I’ll eventually get into the video breakdown but I’d like get it to another level with different teams playing different systems in zone – along with a different method of rushing through the neutral zone and gaining the blueline.

Below is a chart of the NHL power play as of Dec 2, 2014, after the Stars/Leafs game. Scales on either side represent power play efficiency and power play goals on the left and opportunities on the right.

Two bars are colored in red, two power plays with parallels despite two distinct styles. The Dallas Stars and Tampa Bay Lightning occupy a sizable gap in power play efficiency.

TampaBay has been humming along at a 23.6%, one standard deviation from the 18.57% NHL average and two standard deviations from Dallas sputtering at an abysmal 13.6%. both teams rank at the bottom of the league in power play shots with similar power play opportunities for each club. With a difference of eight minutes in 5v4 time has produced nine more goals for the Lightning than the Stars.

Team	GP	PP Opp	PP Shots	PPG	PP%	5v4 Time
TBL	26	89	91	21	23.6	131:50
DAL	25	88	79	12	13.6	140:02

Best illustrated in the image below, both teams take too long on average between shot attempts events on the 5v4 power play (for the purpose of this illustration I am only focusing on the one-man down situation, since a 5v3 carries a different dynamic). The Lightning and Stars take over 1:40 for a proper shot on goal.

In a direct comparison among the rest of the NHL, both teams are way above the pack in terms of time between shots on goal, comparing to the NHL averages as per the table below.

Tm	min/FF	min/CF	min/SF
TBL	1:08	0:50	1:43
DAL	1:15	0:48	1:48
NHL Av	0:53	0:38	1:12

The NHL average is 1:12 between a shot on goal, 53 seconds between Fenwick Events (unblocked shot attempts) and 38 for Corsi events. Historically, starting from the 2011-12 season, this is how the uniformity across the league looks year-to-year.

Fsh% and Csh% represent the percentage of shots that make up the underlying components of Fenwick and Corsi shot attempt metrics. I'll expand on the breakdown of Fenwick and Corsi events represented by shots for 2014-15 up to and including Dec 2, 2014 a little further down. For now this image is the comparison of time gap between shots on goal and the percentage of shots associated in Fenwick Events. Both Dallas and Tampa Bay are direct outliers here.

It’s different than a Pittsburgh that can set up and execute, or the Washington Capitals that are firing into a Corsi event every 28 seconds.

Some of the reasons I’ve observed to explain some of the team’s issues with the man-advantage are:

Tampa Bay Lightning

Lightning make it complicated to get into the zone from the breakout.
The setup for a breakout takes too long to get out of the defensive zone.
Options in the neutral zone aren’t always readily available and force the puckcarrier to gain the offensive zone as a result.
Wingers wait along the blueline to force back the defensemen, and may get a late pass to gain entry, otherwise set up as sentry posts without effect in the play.
Sometimes play with the puck at the top of the zone introducing a risk of penalty killers easily moving the puck out and forcing a regroup.
Less structure and more freelancing in the neutral zone on the regroup.
Once set up in the offensive zone, they pass it around too long looking for the perfect shot or the seam to send pucks through.
Too many blocked shots.

Dallas Stars

Dallas moves too slow through the neutral zone regardless of a formal breakout or a regroup from a dump out of the offensive zone by the penalty killing team.
Too many stationary players along the blueline waiting to get the zone instead of movement
Plays seem forced along the blueline.
Too much individualism, trying to gain the zone and not enough passing options.
Once gaining the zone, the tendency is to fire the puck around the boards to set up (seems a waste since they’ve already gained the zone.)
Stars also seem to switch the set up from new-wave traditional 1-3-1 into a modified 2-3 with two shooters at the top and three forwards a little lower in the zone (slot presence moves to the crease with the puck at the top of the zone and into the slot with the puck on either wing, with the weak side winger slightly edging towards the net as the set up to the slot man becomes the go-to play.)
Limiting the data to only shots on goal only limits the shot attempts function, so I wanted to see just what is happening with the shots in regards to overall shot attempts. The image below illustrates the percentage of shots comprising Fenwick (unblocked shot attempts) and Corsi (including blocked shots) events at 5v4. Both clubs fire less than 50% of their Corsi events as shots on goal. Tampa Bay has a slight edge over the Stars in terms of percentage of shots for their Corsi – although both are below the 50% threshold (Edmonton is the next closest bubble along the slope to the right of Tampa Bay).

Historically, within the BtN era (2007 - 2014), this is what the chart looks like up to and including 2014-15 data. Colorado has had the highest percentage of shots with Los Angeles, St. Louis and Dallas at the other end. Averages are in the blue box.

With the issue of gaining the zone eating up some valuable time off the clock during the 5v4, switching the focus from zone entry to in-zone time, both teams struggle to get shots on goal in comparison to the NHL average. Using data from BehindtheNet.ca sheds some light on what’s happening during shot attempts.

Dallas gets more shots on goal through to the net on a per-60 basis, yet fire more pucks that miss the net according to NHL average, but have half the rate of blocked shots at 5v4.

TampaBay, on the other hand, are almost three times as likely to have a shot blocked, pulling down the overall shots-for per 60 rate to almost half the NHL average. Not only are they taking a little too long to get shots through to the net, they’re struggling to get shots on goal.

Team	SF On/60	MF On/60	BF On/60
DAL	21.16	17.42	1.44
T.B	16.24	17.3	9.61
NHLAv	35.05	15.95	3.46

Taking it down one level from the team to position, the Lightning blueline seems to have a greater majority of shots blocked compared to the NHL average, while the forwards coast along the average, missing more shots than the average. It’s juxtaposition with the Stars and Lightning for missed shots, Stars forwards missing a lot more than defensemen and forwards firing past the net than Lightning defensemen.

Position	Team	SF On/60	MF On/60	BF On/60
Defense	DAL	18.69	21.71	1.35
Defense	T.B	13.69	14.78	12.53
Defense	NHLAv	36.65	15.98	2.47
Forward	DAL	22.48	15.13	1.49
Forward	T.B	21.58	21.78	3.43
Forward	NHLAv	34.18	15.94	3.99

There’s a possible shot quality argument that can be made at 5v4, exemplified by Tampa Bay, while Lightning writer, Kyle Alexander, makes a salient point of the Lightning’s power play.

@kalexanderRC I think shot quality matters more on the PP too but the data does suggest 5v4 SH% is almost all luck http://t.co/ZMlj1B2IHP

— Fear The Fin (@fearthefin) November 17, 2014

That chart in the link indicates that power play shooting percentages can be heavily influenced by ‘luck’ – a term I’m not apt to use without an explanation that ‘luck’ is more about the repeatability of a skillset to achieve similar results rather than pucks bouncing off of player’s asses into the gaping net.

More power play analysis is going to come down this pipeline.

Stats via Hockey Analysis and Behind The Net

************************

Follow the McKeen's team on Twitter:

@mckeenshockey
@KatsHockey

Kats Krunch: Q1 NHL Score-Adjusted Metrics & Situational TOI

Gus Katsaros — Tue, 02 Dec 2014 15:47:07 +0000

American Thanksgiving marked the unofficial end of the first quarter of the NHL season and time for some retrospection of some teams. Traditionally, I’d turn to typical Corsi and Fenwick along with derivative stats, but I've been enthusiastic by these newer score-adjusted metrics quickly establishing their value into the analytics discussion thus far around the quarter mark of the NHL season.

The original work by Eric Tulsky using the original data via behindthenet.ca has seen tweaks and enhancements to even suggestions that ‘close’ metrics may not necessarily be as predictive or originally though.

Steve Burtch expanded on this concept based off work by Micah Blake McCurdy and I’m intrigued to see where this is going. The nature seems to be to discard the limitations associated with ‘close’ parameters and supplement that with score, venue and schedule adjusted data.

There are even resources already doing the calculations for you, like Puck on Net, while and Puckalytics and original sister site Hockey Analysis already houses the raw data.

I like how the distinction here is between game-states, however, for the purposes of this writing, I’m not looking at the predictive value, only a snapshot of teams in different game states and keying in on some shorter team trends.

I like the fact we can break down specific game situations by time and events, when teams are tied, up or down by a goal, then by two or more goals. Let’s begin.

GAME TIED

Teams end up the majority of time with the game tied. There are outliers like the Pittsburgh Penguins who score first early and skew the time on ice by spending the majority of their time up a goal or two. The near past also shows a jump in the metrics based on time on ice with a deficit.

Minnesota, Carolina and Columbus also share this affinity, however that cluster of teams at the top of the chart, the bottom feeders of the NHL and the … Calgary Flames?

The chart (using data by Puck on Net) illustrates raw Corsi For and Corsi Against while the game is tied while bubble size represents time on ice. Teams in the lower left quadrant are playing less when the game is tied and more traveling along the x-axis. The cluster at the top is interesting due to the inclusion of the pesky Flames with that group. In short, while playing most of the game tied, the Flames along with the cluster of bottom feeding teams are allowing more Corsi Against events than Corsi For.

Time on ice is fairly consistent, the bubbles being very similar in size, albeit the ones in the lower left quadrant seem to be a bit smaller, mirrored by the low number of overall events. We will expand on the Hurricanes and Wild below.

GAME DOWN BY A GOAL

Here I’m more concerned with teams playing in traditional close situations and their individual components.

When the game starts to get away and teams start to lead by two or more goals, score effects kick in and we see teams with the lead press on a little less and teams playing without the lead apply more shooting pressure. Score effects are well documented and don’t need any expansion here.

The graph plots individual Corsi For events along the x-axis and the Corsi Against along the y-axis.

Most teams are clumped together in a range with outliers here are the Carolina Hurricanes and even the San Jose Sharks (while also playing at a high proportion of game tied minutes) with almost double the For events than Against. The Canes spend the majority of their game time down one goal and the cumulative effect over the season timeline has recently surpassed their time on ice while the game is tied (team charts are a rolling 3-game moving average).

The uptick with the team up by a goal and up by two goals, eerily corresponds to the return of team leader, Eric Staal from injury and even though the Hurricanes spend an inordinate amount of time playing from behind, the black line indicating being down by two or more goals has flatlined. Even with the Canes time on ice down by a goal their score-adjusted shot metrics are creeping up to over 50%.

The San Jose Sharks started off fairly hot, but signs over the near term are trending negative. Travis Yost, the analytics writer over at TSN does a good job expanding on the Sharks off season transactions after the disastrous playoff exit and just how well Joe Thornton has performed for the scrutiny he’s faced seemingly his entire career in the Bay Area.

This is how the Sharks situational season looks.

We can see early on how they played more with a wide lead and then (scoring 3.8 goals per game) and sputtered (down to 2.7 goals per game) – while even losing to Buffalo 2-1 at home. At the quarter point, the down one goal line begins to trend up – sharply, coinciding with a rise in time down two or more goals. Both metrics indicating playing with the lead are a lot flatter than a winning team desires.

A 3-game moving average of score adjusted Corsi, Fenwick and even shots, are all trending down after swooping upswing earlier on in the season. Something to watch for the California based club.

GAME UP BY A GOAL

Let's look at now at the game state of being up by a goal. There’s a little more separation here from the chart down one goal, with two main clusters.

It's no surprise Buffalo with a very small bubble is also shown your with very few Corsi For events, just over 50, and having almost four times the same amount of Corsi Against events. In fact, bubble size expands the further the bubble appears from the y-axis, with Winnipeg looking like Jupiter sized bubble compared to the Mars and Mercury-like size of the bubbles in the lower left quadrant.

Not pictured here, since these values are as of the American Thanksgiving, is the November 28 Buffalo game against Montreal featured the Sabres with a one goal lead for 30 minutes. Heading into the contest, having amassed an unassuming 118 minutes up one goal, the 30 minute increase represented an increase their time by almost 25% of the season total.

The real outlier here is the Winnipeg Jets, tracing the knife’s edge between winning and losing while only really being up a goal for the majority of their playing time (other than a tie game). Their situational season timeline is in the chart below.

The Jets were going down by more than two goals very early in the season and then between the fifth and 10th game something began to limit their defense affecting their down by two or more goals line to plateau, with a corresponding ascension in both game tied and up one goal situations.

Winnipeg started the season with a 1-4 record - scoring at a rate of 1.8 goals per game and allowing three - only to win six of their next eight games, allowing 11 goals (1.375 per game) and scoring only two per game – a very thin margin. They rallied along to a .500 record and a mark of (5-4-3) record in the last dozen games. To maintain any semblance of success the run after the break away from the pattern of being stuck in the zone between game tied in plus-1 situations. Their individual score-adjusted Corsi, Fenwick and Shots 3-game rolling average has seen the increase from the beginning of the season to peak just over 51% overall for all three metrics.

Let's move on to our final team, the Minnesota Wild. A hot start has become a staple for the Wild, only to tail off and struggle for a playoff spot.

Not so in 2014-15 where they started off with consecutive shutouts against last year’s surprising Colorado Avalanche. The Wild sustained some immediate success finding some scoring touch early on (averaging 3.4 goals in the first 10 games and only 2.3 since then up to American Thanksgiving) exemplified by their up two goals line in the chart below.

There’s a hiccup during a four game stand that saw them score only three goals, while allowing 14, indicated by the flat segment of the up two goals line and sharp upturn in the down two goals line. In the near past, that down two goals line has once again taken an immediate sharp upturn as the Wild defense hasn’t shown the same stinginess it did in the early part of the season.

When taking the individual Score-Adjusted Corsi, Fenwick and Shots moving average, the trend is negative from an unsustainable 60% clip early in the campaign. Even at this 55% clip, there is still some room for a negative correction with the result being a loss of standings points. It’s a good bet to keep a watch on the Wild.

************************

Follow the McKeen's team on Twitter:

@mckeenshockey
@KatsHockey