Friday, August 31, 2018

Power-Play Post-Mortem

Just past the media and the mouths of fans' fixture that was the tragi-comical penalty kill - vacuuming up the attention of both pitchfork-equipped corps of the Orange Crush and the chuckles of visiting broadcasters - what went miraculously much less noticed was a squandering of the man-advantage.

This was, in part, due to a puzzling lack of opportunities. In a setting where the go-to defending tactic against Connor McDavid was supposedly being cracked down upon and the preseason filled with penalties for the slashes to the hands at a rate that indicated the league was saying this was the new now in both actions and words, the overall penalty-drawing rate of the league did not increase and the Oilers ended up spending 34 minutes less on the powerplay than the year before. In a bewildering development, McDavid himself drew 12 less penalties at 5v5 year over year, despite playing more and being undeniably more dangerous. Those lost 12 whistles figures for a thirty-two percent drop in rate of penalties drawn during 5v5 play, jarring and confounding against the previously mentioned backdrop of cracking down on obstruction.

But after taking this into consideration by looking at the rate at which the Oilers scored in the time they were given, we find that the team was the least effective in the league at scoring goals during 5-on-4 play.

This is a drop from 6th in the league in 2016-17. This figures out for is a drop in goal differential from +46 to +25. 

Comparing the impact of this to the more infamous penalty killing, one also has to add in the in-season trending of both. From the first game of the season through to Christmas, the Oilers lead the league in 4-on-5 goals against per hour with 10.10. To end of the year from there, they were 16th in the league with 7.14 per hour.  If you shrink that second side performance to the last 25 games, they allowed the second lowest at 3.21.

Let's contrast this with the 5-on-4 performance, by the exact same measures: 

In the chosen first frame, their rate was 8th-lowest at 5.70.

In the chosen second, it drops to NHL-worst with 4.59.

Shrinking the sample to the last 25 raises it to 5.53, still worse than the pre-Christmas rate.

With the penalty kill, there's a lot you can point to for improvement. A change in personnel via players leaving at the deadline, and a change in tactics via Todd McLellan taking over the structuring from assistant Jim Johnson gives real, concrete reasons to believe that the change is more than just the good side of variance and goaltending.

Said assistant was also let go of by the organisation shortly following the close of the regular season.

It gives you a grounded belief that such failure is unlikely to be in the cards for next season.

But the powerplay is the opposite, and that's why I think it should be talked about more. The personnel is returning, there appeared to be no recourse for its floundering, and the man in charge of the unit is now coaching the AHL affiliate.

When Jim Johnson was relieved of the penalty kill, and Todd McLellan took over and implemented a basic structure, the problem was solved.

So then, in this event that there's new assistant coaches, you can simply say that if the penalty kill fails again, you can take the duties away from the assistants, give them to McLellan, and live happily ever after.

But by the same logic, there's no aiding a failing powerplay in the future. Yes, there is new assistants. Yes, Emmanuel Viveiros was a fine candidate and I'm glad they have him, but what we're talking about here is 'Plan B'.

They got to Plan Z at some point on the 5-on-4 last year, when they quite literally just rolled the 5v5 lines out there, which is somewhere between taking your toys and going home and saying that you're not even really trying, bro.

This makes performing the dissection of the powerplay infinitely intriguing in comparison. 

Random Number Generating

Let's check out the type of shots and chances recorded by the Oilers to contextualise the goals.

 (By the way, those following unless denoted otherwise and all those before this notice are via

Everything looks fine and dandy here, from a generation perspective the Oilers were above league average, but personally when I look at stuff like this on the powerplay what these rates tell me is more that the team was fine in transition - gaining the zone and setting up - rather than they were actually good at setting up real chances. This type of shot type/location context is useful in larger samples during 5v5 play, but in 5v4 the context is so wildly different team to team based on personnel and tactics that really, no two shots are the same.

For example, Patrick Laine:

Laine's recorded as registering a total 14 high danger chances on the powerplay in his first two years of NHL play. He's scored 29 goals. You can see where these shots come from, betraying distance they're dangerous because of the amount of right-shot one-time options the play has coming off of the right half-wall on WPG's powerplay:

So these are far-out shots from Laine that will result in a repeatable overperformance of shot location averages because of both his shot and the tactical depth of the unit he's on.

Now that we've established that you earn your shooting percentage much more at 5-on-4 than 5-on-5, let's look at how the Oilers executed:

And there it is. The overall shooting % looks bad by the bar, but I'll add the information that 9.09% was good for worst in the NHL - two teams shot better 5-on-5 than the Oilers did 5-on-4, and eight teams shot better even strength overall than Edmonton did 5-on-4.

These are remarkable statistics, let's check out who made up this average player-by-player

  • In 2017-18, 210 forwards played more than 100 minutes of 5-on-4 time,
  • The average of their shooting percentage was 14.81.
  • 1st on the team was Ryan Nugent-Hopkins; 70th/210; 18.18 SH%.
  • 2nd on the team was Leon Draisaitl; 103rd/210; 15.00 SH%.
  • 3rd on the team was Connor McDavid; 107th/210; 14.71 SH%.
  • 4th on the team was Milan Lucic; 118th/210; 13.04 SH%.
  • 5th on the team was Ryan Strome, 175th/210; 6.9 SH%.
  • 6th on the team was Mark Letestu, 179th/210; 6.45 SH%
  • 7th on the team was Patrick Maroon, 182nd/210, 6.25 SH%.

I'd like to address two relevant things as they've just come back to me. First, often when the Oilers' season was talked about in real-time, it was remarked that this is basically the same team as 2016-17.

At 5-on-5 and 4-on-5 this was obviously not true, but the Venn diagram of people who said this and the people who underrated players like Jordan Eberle, Tyler Pitlick, Benoit Pouliot, David Desharnais and Brandon Davidson would be a circle, so they could simply just mention their evaluation of those players as entirely dispensable, and continue on with their premise.

But on the powerplay, that sentiment holds much more true. This absolutely was an almost fully returning cast failing to deliver the same goods from the year before. Let's repeat the exercise for them.

  • In 2016-17. 200 forwards played more than 100 minutes on 5-0-4 time,
  • The average of their shooting percentage was 15.63.
  • 1st on the team was Milan Lucic, 25th/200; 25.00 SH%
  • 2nd on the team was Leon Draisaitl, 47th/200, 21.28 SH%
  • 3rd on the team was Ryan Nugent-Hopkins, 59th/200, 19.23 SH%.
  • 4th on the team was Mark Letestu, 72nd/200, 17.95 SH%.
  • 5th on the team was Patrick Maroon, 151st/200, 10.00 SH%.
  • 6th on the team was Connor McDavid, 156th/200, 10.00 SH%.
  • 7th on the team was Jordan Eberle, 161th/200, 8.82 SH%.

This is stark. The team had four top-half guys the year before, just barely two this past one. Only Ryan Nugent-Hopkins and Connor McDavid improved. Which, by the way, was the 2nd thing I wanted to mention that was talked about - multiple times I heard one of the problems with McDavid's powerplay unit was he 'wasn't a threat to shoot'. This is demonstrably false, he shot more and more potently year over year. We see the SH% increase, I'll also tell you now he shot more than he did the year before too. If he wasn't a threat to shoot in 17-18, then he also wasn't in 16-17 and that unit ran well, so it's evidently very low on the list of problems compared to how much it was talked about. I remember hearing it more than any other criticism of the team's powerplay by the media and fans, Oilers or otherwise.

Let's look at some of the big drop-offs:

Typically the ordering of a particular piece of discourse with regards to shooters shooting or players playing goes like this: Someone notes a players' high shooting percentage or point production rate, and says the player should shoot more, or the coaching sraff should play him more. A second someone will retort that if they shot more, their percentage would go down, and if they played more(harder) minutes, their production rate would go down.

The second someone is right quite often, there. The less you shoot the more often you're shooting in dangerous situations is the theory, and it plays out like that in practice a lot as well. The holders of the highest shooting percentage are often playmakers, who will look and look for someone to dish it to and have to have a very potent opportunity in front of them to convince them to shoot it themselves.

Here though, bafflingly, we see the opposite. Half the shot rate, and half the shooting percentage. That's how you go from 3.39 goals/60 to 1.06/60. It's such a cataclysmic shift that it suggests not a failure to perform a role, but a switch in role entirely. With that in our pocket, let's move to the shot map.

First thing to note - the total time on ice. We're looking at just raw counts, here, with a sizeable difference in amount of time to accrue them. For the answer to that question, you can refer back to the chart, where it shows Milan did indeed get much less chances at finishing from tips and deflections some feet out. The role change isn't clear from here, all we've been told is that the spots he's to shoot from stayed the same. What we do know from these two pictures is that the play went through Lucic less, there was a lower rate of executing net-front stuff-in and tip-in attempts. Is this from execution solely, or did they just try it less?

Everything that applied to Milan applies here too. This one we know was a role change. It's a very alarming phenomenon to learn this about these two players. 49 goals were scored 5-on-4 by forwards on Edmonton the year before last, these two players scored 22 of them. They were the primary options in close on a powerplay that looked like this:

Yeah. Good times. I remember watching this team draw a penalty while trailing and cackling as 97 came off the half wall and found 29 who found 27 who scored. Letestu's work at the right circle got more press(likely due to the novelty) but he scored just seven goals to the ten and twelve of the big netside men.

Aside from that aside, let's look at Leon's individual picture.

And does this ever match my eye, and memory.

This looks like a guy who went from having a job to having a dozen hats because folks are quitting the company. Notice especially that sparsity of the goals in 17-18 compared to the closely packed red marks of the year before. There's also much more similar TOI totals between the years compared to our look at Lucic's, which helps looking at the raw dots be a more fair method. But, I think a density chart would help communicate the total shot attempt disparity more.

Yeah, this is the result of moving a player around. What we'll have to look for in the video is what exactly preceded this - was the go-to play getting shut down too much, and a scramble followed after 'Plan A' didn't work? Or, did they come out the gate with a different setup than last year (likely) that saw Leon in different spots? Two places I'll draw your attention to is the top-side of both circles. There's a bunch of wristers on the left, what was the play there, looking for a rebound, or a low slot tip? Also, it looks like there's slapshots on his off-side circle, that's a royal road pass play, when during the year were they trying that, was that what they came into the season with, or was it when they plugged Strome onto the top unit later and had him make those cross-ice feeds? Tons of questions to consult the video with.

This is an equal drop in performance, but with a less confounding immediate observation; we have a distance problem.

Letestu's play eroded enormously in one summer, but that was due to his boots, not his shot. He actually shot better at 5-on-5 in last year's campaign than the once preceding, what he didn't lose was his shot. We've got something to look for in our spray chart, and something to look for in the video - shot type and distance and pre-shot puck movement, respectively.

A much simpler problem to solve is represented here, to be sure.

Given a 22% drop in total minutes, the density chart is key here for contextualising the location.

But what this one is particularly good for, is the exact types. Look at the amount of misses especially in 16-17. This speaks to a willingness to go to that cross-ice play over and over. This is a curio that drew me to check the numbers on it, and I found that Letestu went from about 30 shot attempts per hour to about 23. They just weren't running that play as often - one could say that perhaps the passes were being broken up repeatedly, or missing, but that would likely show up in the possession stats as it's hard to retain possession if a hard, cross-ice pass going to a side of the ice with only one player misses. Still though, we ought to find out in the video. The shot distance thing, especially clear in the area between the crease and the right dot is particularly barren by looking just at these raw attempts is troubling, but again we'll have to help our eyes with the density version.

As expected we see a bit less of a wholesale shift in location, but what should be noticed is those three pockets right dot and up - a lot of these are slapshot-squares and point to more cross-ice passes being made for shots at distance, of which an 11% accuracy drop can inform its effectiveness.

You can also see the amount of bumper-spot activity here was substantial - another video cue. Notice also that for all of the general activity up high, none of them resulted in first-shot goals. This is an awful result, but only if we can also rule out low rebounds, which the bumper player can often produce.

In checking for this, I've found that Letestu's rebounds created dropped from 1.80/60 to 0.93/60, and his primary assist per hour rate dropped from 1.08 to 0.46.

So, all in all, from the data we've gathered it certainly looks like when Letestu's role did change, it was almost entirely ineffective at generating goals and dangerous chances.

These three forwards - Mark Letestu, Milan Lucic, and Leon Draisaitl - accounted for 29 of the 51 goals scored by the top-5 converting 5-on-4 unit of 2016-17. All of them appeared to have significant changes in duties, and all of them dropped their goal-scoring by significant amount, combining for just 11 goals in 2017-18. (Though, it should be noted Letestu only played 60 games with the club last year before leaving in a deadline deal to Columbus.)

We will look at McDavid in video of course, but in terms of finishing the forwards I've featured in the data are by far the most important: In 2016-17, the other 17 goals scored by forwards were split among 6 other players including 5 from the 2nd unit's mastermind, Ryan Nugent-Hopkins.

The Defenders

Another large shift in the end-game of the Oilers powerplay is that they let many more plays go by shooting from the point.

In 2016-17, Andrei Sekera and Oscar Klefbom combined for 362 5-on-4 minutes of the 388 made available to the team.

On the first powerplay unit with Connor McDavid, Sekera clocked 10.23 SOG per hour, and 21.03 shot attempts per hour.

Oscar Klefbom in the same situation registered 9.06 SOG per hour, and 24.67 shot attempts per hour.

In 2017-18, the team had 351 such minutes, but with Sekera gone, the non-Klefbom minutes were divided amongst a number of blueliners. 63 to Matt Benning, 38 to Darnell Nurse, 35 to Andre Sekera, 31 to Ethan Bear, 17 to Kris Russell, and most perplexingly of all: 13 to Yohann Auvitu.

The ones we're (mostly, for now) examining are the top unit ones, and those belonged to Klefbom and nobody else of significant sample.

And in these minutes, Oscar Klefbom's SOG rate rocketed up to 16.67 per hour, and 33.31 shot attempts per hour.

That shots-on-goal rate ranked 6th among defencemen league-wide who played at least 100 minutes 5-on-4. He also sported the lowest shooting percentage until Keith Yandle's 0(zero!) at the 21st highest shot rate.

Put another way, this was funneling an inordinate amount of your possession value, your shot attempts through a defenceman with a shoulder injury who just isn't hitting it.

We know this is a development that certainly wasn't discouraged as during an interview before the 2017-18 season Oscar Klefbom mentioned that he and Jim Johnson had talked about "having a mindset and a goal to just shoot as much as I can".

Not entirely ineffective was the rebound-creating shot volume strategy - his 'rebounds created' per hour via went from 0.31 to 1.38 - but how much more potent does an extra rebound per 30 powerplays make your goal-scoring on the man-advantage?

Some of this is variance. You don't just post 1/5 of your shooting percentage year-over-year on merit (though being injured and trying to double your shots rate will certainly help) but what does the location say? Was he shooting from further out as well?

So it looks like the change wasn't a distance one but a degree one; Klefbom fired from everywhere up top and rotated to the left side often for same-side one-timers.

I noticed the centre-point shots, and I'm guessing all those left point releases had were aimed at a high slot tip.

It's easy to imagine a common response to what I'm writing in criticism of over-reliance on point releases is that I'm ignoring the danger that comes out of rebounds and other non-goal outcomes, but what I mean when I say point shots aren't dangerous is that they're not even dangerous when you include the rebounds you create or the tip opportunities.

That's to say, all of the more dangerous non-goal outcomes of releasing a puck off of the point happen more often and more consistently from a play that came from closer to the net.

By NaturalStatTrick's rebounds created, forwards with more than 100 minutes at 5-on-4 took 12854 shots, 7338 registered on goal, resulting in 978 rebounds.

Compare this to defencemen of the same TOI cutoff, who took 5008 shots, 2332 on goal, generating 352 rebounds.

So regular PP forwards take shots that are unblocked and hit the net on 57.1% of their attempts, 7.6% of their attempts generate rebounds and 13.3% of the SOG's do.

Comparatively, just 46.6% of defencemen shots go unblocked and force a save, 7.0% of their attempts generate rebounds and 15.0% of the SOG's do.

So, if it actually gets to the net there's a slight increase in the chance of a rebound, but once the low rate of attempt conversion is accounted for, you're no better off.

Considering how much less often these shots actually score, that leaves us with the conclusion that releasing plays from the point isn't exactly 'getting it to the net', because when your forward is shooting you're (slightly) more likely to get a rebound anyways.

The thing about this is, though. is a lot of the time I've heard and read traditional hockey mind's thoughts on it that it's a failure of puck-retrieval if you're losing ground in opportunity cost by opting to fire from the point. My response to this would be that, as a whole, the best puck-retrievers in the league's first powerplay units aren't far in net outcomes from the worst - the biggest factor is the bounce - because generally if you're on the top powerplay unit, you're a player with a good head on your shoulders, you work hard and you're smart. You can say on any given night a powerplay didn't work hard enough on puck retrieval, and it can be true. But when we're talking about pulling things back to the big picture, macro stuff over the whole season, I don't think you're making up the ground lost statistically and recovering pucks so much better than another team's 5 best offensive players to buck an NHL-wide trend.

Then, there's the disappearance of the point play among league-leading powerplays.

Here are the heatmaps for the top-5 5-on-4 GF/60 teams:

The theme? Green up high. Winnipeg's deviating a little, but we've talked about their options for puck movement and their elite shooting talent.

The Cutting Room Floor

I've gone back and gotten some video from the start of the season, for a few reasons:

  • Starting at the beginning of the season is good to see what the system was fresh off the conveyor belt, especially when it comes to personnel.
  • This was a critical point in the season - for both the unit and the team. Edmonton was -13 in all situations goal differential, and they had 5 5-on-4 goals while they best teams had 15. Having a top powerplay unit would have kept them afloat, much like Pittsburgh's unit saved their season.
  • I've cut out most of the transitional play where convenient. As I've said earlier in the piece, the zone entries and setting up wasn't a problem for the team last year. This is a conclusion that's come intuitively, if they have a near-league leading shot attempts rate, they're probably gaining and keeping the zone pretty well. Especially to my point earlier about puck retrieval rates,
    I don't think you can game your Corsi For rate by instantly shooting from anywhere without just losing the puck and having to re-enter, which in turn lowers your rate as you waste time getting back in shooting position..
  • This is a lot of video to annotate for my purposes requiring another long-form piece to fully review, and I'd also like to share the raw-ish material so you can come to your own conclusions without me drawing all over it like I usually do.

Here's what we're looking for:

  1. Does the play end up going back to the point more often than necessary?
  2. Is the puck movement directly prior to the shot cross-ice?
  3. Were attempted cross-ice passes broken up at a rate that point plays were the only option?
  4. Is Letestu in close as he could be?
  5. Is the play going to the net to Milan Lucic, or through Milan Lucic, very often?
  6. Does Leon Draisaitl have a concrete role?
  7. Does Oscar Klefbom shoot often with other plays available?

From what I can gather, the dream play is to get 97 the puck up high while the whole team spreads out in something of a reverse umbrella, and then as he comes in down the wall the rest of the unit closes in too, before making a play to the goal line and back in front.

Something I notice is the amount of times the shot comes from the same side of the ice as the last puck-movement - meaning no one has to move, not the goalie nor the defenders. This in part addresses question #2.

Speaking of the questions, #4's data and video both answer the question the same way. 55 is much further out than in 16-17, and it doesn't appear to have been forced by the opponent to be this way.

On the bright side, I don't think the lack of dangerous passes seems to be a result of the players not being skilled enough to complete them, but rather that they weren't being demanded by the system. This bodes well for next years' team.

In regards to Milan Lucic, it appears that plays through him switches from behind-the-net-to-front plays to high slot tips and deflections coming from the outside lanes in.

And on Klefbom, I wouldn't say that they're just funnelling attempts through him without thinking, but rather that the only pre-requisite for a satisfactory point shot was just a tiny-bit of puck movement like a pass from the half wall to the point. That's not nearly enough deception or movement.

You'll also notice Draisaitl's absence from the top unit from a coupe of these games. This is in addition to the different roles thrust upon him.


It's easy to look at last year's powerplay and say they weren't dangerous enough or didn't work hard enough.

It's also easy to say take less point shots and get plays from behind the net to the slot, or move the puck cross-ice more often.

A coaching staff would probably roll their eyes at me if I told them that, musing that misexecution can ruin their best laid plans.

The team wouldn't take any point shots at all if there was always an option to make a more dangerous play, right?

My answer is twofold. One, if the optimal scoring play is so often unable to be executed by two of the game's best playmakers, the plan might need revising, and two, if you're asking Oscar Klefbom to take 250 shots it seems it's not a big deal to release plays from the point.

The optimistic conclusion I have here, is from the answer to the following question, already put forth multiple times in this piece: are the players failing to engineer potent scoring opportunities, or are there not many in the blueprint?

It's a mix of both. The 'dream play' for this powerplay scheme was too unrealistic most of the time, so the team did indeed defer to the easier, safer(for their team as well as the opponent) play, but the raw amount of tries they gave to generate a surer goal was low enough that their percentage wouldn't be that of a unit we could expect continued failure from.

In short, given better tactics and a level of execution simply in line with the returning players' ability, one could expect a powerplay well away from the depths it sunk to last season.

This is, of course, a reasonable point of view even without having gone through the research I have, but the point of this blog is to drill down on the popular talking points among the media and the fanbase.

So when everyone says the Oilers powerplay can be expected to bounce back, I'll give that theory a pass. 

(Though, speaking of the public discourse, I think we've also made the discovery that the Oilers powerplay was even worse than what's talked about, and much more perplexing than the more famous penalty-kill!)


Friday, August 24, 2018

SSS Part 9: On the State of the Series + Mar 13 vs CGY (plus two more games later)

On the State of the Series:

I wanted to diversify the blog a bit from just one project so I went walkabout for a lil on this Leon deal. 

I also was a bit exasperated with my premise. I'm always quite a bit further ahead in games analysed than is published, yet don't want my present thoughts colouring the whatever current game I'm doing and it wasn't really working ignoring it so I'll address it:

A lot of the material covered has been about process and not necessarily just results. The goal share was lost by Leon, but instead of assuming it could have been better, could it not have been worse? Instead of the optimistic(sort of) view that the goaltending failing his unit and betraying the shot share numbers, was the goaltending actually stronger than we could hope for as average? That's to say, is Draisaitl not just not good defensively, but outright bad?

Where I'm at with that question is a hard no.

We've seen and we'll see stuff like defencemen blowing tires, making ridiculous pinches, soft goals and other oddities that rob Draisaitl of even the responsibility to bear the goals against number.

I've been sort of exasperated, I was expecting more smoke to the fire, where's the errors in processing or effort? Don't get me wrong, as a fan of the team, I'm thrilled that the problem seems to be less than it's spoken about, but again in the spirit of the project I'd hoped there'd be more to chew on.

So then, where do you go from there? I think you might have to just smoke out the goals.

When splitting game logs for other reasons including upcoming project work, I came across an interesting tidbit.

Leon's final tally way from McDavid was 28 goals for, 38 against, minus-10.

At Christmas, that number was 10 to 10.

So basically, the entirety of the GA problem occurred in the final 46 games.

In those splits, though, the score and venue adjusted possession and chance stats (from 5v5 play) also dive some:

(pre Christmas / / post Christmas)

CF%: 55.64% / / 47.06%

SCF%: 57.63% / / 45.23%

HDCF%: 53.97% / / 43.55%

This is pretty drastic stuff, but as I noted in my Yamamoto piece it's important to contextualise into the team environment. The entire (non-McDavid) team had similar woes, I'd suggest both tactical changes and individual player's level of play dropping off to varying degrees. Here's the non 97 or 29 follows.

CF%: 51.33% / / 45.89%

SCF% 51.11% / / 43.12%

HDCF%: 52.65% / / 37.94%

(Observe the second column's numbers, put them in your pocket for this coming fall when the heart, emotion, or any other type of feelings fortitude is called into question as a reason for the team's failings)

So a lot of those 28 goals we'll see on the way, but I want to compile them for the end of the series, that's what I'll do. I think one good look like that at the end will put this thing to bed.

March 13th vs CGY

Going in, we've got a goal here, and a few more saves forced.

Another thing to mention when it comes to the post-Christmas split, for Leon in particular, is quality of linemates. This is post-deadline but pre-demotion for Cammaleri and Lucic, so that's what Drai's saddled with here. They're not the worst of the bunch, but consider one of the wingers of the matchup is Johnny Gaudreau.

Also of note is the large minutes share with Russell-Bear.

Quick note: on the breakout I'll always prefer Leon to make a power move on his forechecker rather than throw the puck up for the tip-in, but tip-ins are a regular fixture in the NHL that I'm literally always against, so that's bias.

Bear getting his stick taken away makes a clean shot really easy, but Drai's backpressure ends the threat and the puck tumbles to the boards, where the Flame continues to mishandle and Leon grabs the puck, reverses it, and goes for a change. Russell never intends to make a clean breakout play here, so he's got to actually get that puck out but he doesn't, and I'm not sure if Drai's recorded as being on the bench yet so I left this part of the clip in.

There's an extended Benny Hill theme deserving breakout process here that Drai's up the ice during and it results in what is probably another marker on our map through no real fault of his own.

Which brings us again to the propensity of the coaching staff to have Drai play F2/3, with a veteran down low. (It was Cammaleri on this line and others, which is why I wasn't too surprised when he was 4C at some point down the stretch, the coach trusted him more than other options)

Going in I thought we'd see a lot more of actual F1/Centre duties being paid by Leon but we don't. It'd be really useful for the future of this team and the play at that position for him to have just played the position in the garbage time but I understand the coach feeling like his job was on the line because had he finished the season with closer to 70 points he'd have been gone by this point.

An opportunity for a clean entry gets turned into a nightmare here after a poor pass, eventually Russell puts up a wall really well and the play ends up outside to Leon who could have more carefully made the play up the boards, but in the end the Oilers are spared.

Again, there's a netfront attempt here. We've covered 3 of the 5 non-goal slot area tries.

There's probably nobody in the league you want to leave that alone with your goalie less than that guy.

It's his game, though, he's sneaky.

You can see him hanging out by his lonesome in the left corner earlier when the play goes around to the boards, when the play goes to the D originally I was about to ask myself why they were changing a forward at that time, before the camera pans and you notice Gaudreau was hiding off-screen.

It's a no win with him, he can do so much with so little space that you don't want to test what he can do in open ice, but then the other end of his game revolves around perfectly finding seams that open up when you pursue him.

 Darnell never knows he's there, and this is a combination of both the system playing harder to the man (no man in front of the net; not necessary to be in front of the net) and the breakout philosophy of having every player move up the ice as quick as possible during a possession change which punishes poor decision making by both teams harder.

Nice shot, nice goal, fuck you, let's move along.

Another slot shot on our map results from Bear making a play he can't make if he's gonna get out skated on the transition.

This showcases pretty perfectly two of the hurdles Ethan has to get over to become an NHL regular: decision making on stepping up, and turning/skating quicker.

To be fair, if he fixes one he fixes the other. You end up seeing a lot of Bears mistakes pouring through these minutes like I am, so from time to time I should mention I think he's good and I like that they gave him the opportunity.

However, like the biased critic I am, I'll turn that into the question of why the young wingers didn't get the same development-over-wins-column treatment.

This is actually a really funny clip, something about the way the Oilers played this as if they were winning 10-0, the gentleman's change by McLellan and Gulutzan, Lucic coming down because Bear was playing Bakersfield's DZ structure, there's just so much going on.

I think Milan barked at the kid to let him know what was up, but it didn't work so the Flame still got the shot off clean.

If you've been counting, that's all six down-low shots from the map. I didn't get to draw on these much because of Drai playing up high, but I'm marking this one down in the 'created more than he gave up' column, even if the bar wasn't too high.

March 14 vs SJS

The map's missing for this one, but we can still see the game's a microcosm of our study: an equal share of shots, but zero goals for and three against.


This is a play where, as fourth guy in Puljujarvi's got to be higher up, there's a seam between the two highest Sharks that he needs to hit for the pass to be any safe at all. Draisaitl can just choose not to take the risk, too. Really tough play. 98 almost gets back, which is impressive but even if he had that'd just be a mistake of pace by the Sharks.

This was a common type of GA against the Oilers 17-18, they held leads like a drunken gambler.

The first half of this clip is a great summary of the trouble having two same handed guys on a pairing gives you.

Sekera draws one forechecker right to him, and the second one to his side of the ice as well. If Russell's right handed, he can take a pass like this in stride and immediately skate or pass it out. Instead, he's gotta be turned around from the start, and he reverses back to find a guy who's cover has had plenty of time to get on him and the puck ends up back where it started.

2 draws time for 4 again, and again an easy pass for a right shot gets handled poorly and ends up going off the wall and out as if the forecheck outplayed the breakout, instead of the other way around.

The Sharks do the Shark thing and Burns throws it through layers at the net, dice comes up Oilers(it's heavily weighted that way with point shots, even Burns') and they clear.

This is an unnecessarily hard flip-in by a guy with Puljujarvi's feet, but Bear recovers the puck and does a nice little give and go play with Drai who absolutely loves sitting in that right corner in the OZ. I like the slapper here because he's in close enough and the pass is fast enough, and 98 or 27 will cash this a lot of the time in any year other than last. I left the last part of the clip in to show Drai staying in position on an assignment down low.

I saved and uploaded this clip solely because I love this entire shift by Pulujarvi. What a beauty.

This is a classic 'leading' play in the NZ, where the second layer of checking is so weak and so far back that it's a literal 4 on 2 by the blueline based on who's already skating. Drai has to cover the kick-out here so that his Dman can actually contest the carrier, but he doesn't follow and the first guy who gets a stick on this play is actually Puljujarvi which is pretty bad. Then later, a Shark comes in off the wall into three(3!) Oilers and walks everyone. Drai's in the right place for a rebound that probably should have just been in the net, and clears it. Very ugly shift.

Lucic gets walked here too, which is what's basically gonna happen no matter what - IF you're standing still with that much of a gap on a good player. It's a positioning error, not an effort/execution error, despite what this type of play usually gets called by fans and some media. It's why hitting a trailer is so dangerous, why it's attempted so often.

Later on, you really just need to get that Shark out of the net. Some people talk about Puljujarvi needing to discover a mean streak, but really he's just got to use his body here. He bodied this guy with less purpose I've seen him do by accident.

March 17 vs FLA

This is a game where they lost the goal battle undeservedly if you look at the raw shots, but on the map you'll notice the slot activity's pretty well the same. On the video you'll see reasons for this, and I want to use the opportunity to tackle some of the stuff I've got beef with in the attacking structure that I hope to see less of next year.

Boughner hard-matches, by the way, for anyone who noticed that. It's not just this game. He's gonna have a good team to do that with by the way next year, bet the over on the Cats.
(Also notice the completeness of the TOI Drai played this game, individually a fine game for him.)

Caggiula races to cover for Larsson stepping in, and as the play runs down the ice Matheson's gonna end up just walking in unchecked as it seems like both Slepyshev and Draisaitl want to play like they're last man back. Slepyshev wants to cover the guy he's facing, but just by ordering he should be F2 and go on Matheson. Drai sees that not happening and doesn't really react though, and as a result there's a three on two down low. Nurse stepping back all the way to the front of the net is important for later, too.

Darnell, here, should absolutely be playing to the other side of Larsson and his man. He ends up completely outside of any check, and there's almost no way for the play to go anywhere else considering the two Cats left in the clear

Leon makes a timely and necessary switch here in a play where a player could easily be fooled. Wanted to show this as a credit to him, could have been a tap-in. 

He makes an angry against replacement stop behind the net and they're free.

Here's a good play for looking at how the team with McDavid off the ice isn't very dangerous despite prolonged possession.
You can see an extended period of time with five Panthers down low, but the team's got the puck behind the net well clear of any checking but you can see Caggiula doesn't even look at the slot before going low to high. That's likely coached.
Then, when the puck gets to Bear note the sheer amount of space that's opened up. And then as Bear winds, note that the two net-front guys leave the screening immediately, signalling it's an intentional back boards play.
So then they've got all kinds of space in the middle high slot, but Drai doesn't even look and runs it back up high.
This play off the 'high cycle' was a common one for the Oilers last year. It's something of a give and go with options on both wings, and results this time in another flip down low. 

The puck gets recovered and thrown back up high.

Anyone else wondering what the end goal is here?

I understand they don't draw it up like "yeah let's use 40 seconds of OZ pressure of our second line to go low to high to low to high to low to high to point shot block and goal against" but if the way the team is utilizing the given tactics ends up with these results the process must change. Regardless of who's to blame or how much.

A winger misses on a clear by forcing it up the wall with a middle option available, then fumbles the outlet after Klefbom recovers. Draisaitl recognises the outnumbering on the re-entry, and we've got a bad angle for figuring if it was him or Oscar that breaks up the play

Caggiula looks off the easy pass to Slepyshev, and finds Draisaitl with a boards play.

This would result in a breakaway off the line-change, but either bounces or Matheson somehow take the play away before it starts.

Bjugstad's dumb-ass reach catches this play, miraculously (and not entirely legally, though the refs often the give the defender lots of leeway on catch-up plays like this). It's a waste of a gorgeous pass by Draisaitl.

Bear doesn't shoulder check very soon after hustling back towards the net, so he doesn't know the type of gap that's left to the third Panther in and resultantly there's all the time in the world to get a shot off, but the guy takes all of that and more and Bear recovers to stymie the attempt.

Pretty funny back and forth here, I'm not sure if they counted it as a shot attempt on the map. If it did, we've got a bit of context on another (coulda been) dangerous play where Leon's just kind of the last guy back.

There's a winger switch here because of some erratic energy-based(post-icings) stuff where the coach was pretty much forced to send different combinations out in order to have the freshest legs out to kill the clock.

I show this because it's another example of Leon playing high that kind of kills our premise of dangerous outshooting based on errors by him down low. Between the smaller amount of time spent on those duties in the first place and the perfectly fine ratio of mistakes when he is, I think I'm building myself a case here.

Friday, August 17, 2018

A short thought on my usage of models

Yesterday @Woodguy55 asked me a question in the comment section of Allan Mitchell's blog

"How do you feel about Corsica giving Larsson and Benning an almost identical Player Rating?"
I added a bunch of additional information(to the preamble directly before the first infograph) about how and why I use models, that should be posted here for reference.

Benning is a guy who we know has better macro shot share impacts, so this is a guy we know who is getting a boost.

Manny’s model we know rewards guys who offensively outperform their xGF, but doesn’t (or sometimes inversely) reward guys who defensively outperform xGA.

If a model makes you double-take, just use general rules about the strengths and weaknesses of it, and see if there’s consistency in players it underrates and overrates. I know Larsson and Benning are gonna have opposite-ended directional pulls just from what the model values before I even peruse their numbers, I’m not surprised by it.

I look at the methodology guys are using /then/ grab their data, every time.
Some of the language gets lost in translation, I do what I can.

Another example, Dom Luscycyzn’s game score loves point production. His model doesn’t like Larsson either. Guys who do more than produce points as play drivers get underrated, Nino-Hall at forward, Tanev-Hjalmarsson at D, to steal Ian Tulloch’s distillation.

You can see the weighting in action here

Player Game Score = (0.75 * G) + (0.7 * A1) + (0.55 * A2) + (0.075 * SOG) + (0.05 * BLK) + (0.15 * PD) – (0.15 * PT) + (0.01 * FOW) – (0.01 * FOL) + (0.05 * CF) – (0.05 * CA) + (0.15 * GF) – (0.15* GA)

Heavy on points.

Larsson’s two seasons in 15-16 and 16-17 tell me models like those two aren’t going to like him because he’s a usage freak who can not get murdered in roles 95% of Dmen would get murdered in.

The purpose of the models in the exercise isn’t really to dig deep on Edmonton’s D corps but to get a cursory glance at the other Pacific teams without having to do the level of research I feel was necessary to get the full scope of things.

I am reasonably comfortable in thinking I have a more detailed understanding of the individual Dmen on Edmonton than what the model can tell me, not entirely so with the other 40+ guys in the division.

The reason I’m not gonna give Larsson a bump in the numbers in the excercise or say that he should get one is because then you have to give that to Tanev and every other guy in that player type, and then you also have to ding Burns et ceter et cetera.

The other part is, especially with the GAR I was curious to see what the projected net of pairings were.

Like I said earlier in the thread, I’m not super confident in my ability to project a /pairing/ because of issues weighting stuff, especially in pairings where there’s one guy who’s good and one guy who’s bad. The GAR gives you a little picture like hey, the guy who’s good adds eight and the guy who’s bad only takes away two, so they might be alright.

That’s an interesting thing to note and come back on, to see if the model was right, wrong, and how often either. Or in what ways they were right and wrong. Then I take that and maybe I’m closer to projecting pairings than I am now.

Thursday, August 16, 2018

Sans Sekera

Most of my work here is further post-mortem than days following.

I'll deviate this morning, to address the reaction to Andrej Sekera.

The Oilers fanbase as with any other is often miscast as either a singular entity with a collectively homogenous viewpoint on any particular issue, or two tribal camps opposing one another.

Neither's true, save the second during certain times(typically trying) and I don't mean to play the snooty centrist believing himself above the partisans and star-bellied sneetches but I do believe that the middle ground is most reasonable here in regards to Sekera's absence for what will likely be an entire season.

I'm writing in real time during my research, so I have the unique ability to tell you that my initial reaction coming into this is that the main consequence for this development is the lack of cover for injuries.

As opposed to something like, say, the difference between the starting six prior to the loss of Sekera and after the loss of Sekera being the difference between projecting this team in the playoffs or not.

That might read as my projecting the Oilers to make the playoffs either way, it's not so.

I had the Oilers making it neither way, provided average injury and puck luck, and average goaltending variance.

Oilers Dcorps; up against the Pacific

I figure the most fair way to run this thing is look at what everyone else's got.

I'll make pretty pictures, with metrics I consider to have above-arbitrary value and then provide my own thoughts.

The numbers are projected Goals-Against-Replacement*; Blended Wins-Against-Replacement**; Corsica Player Rating***.

* via Chace McCallum @cmhockey66
** via Sean Tierney @ChartingHockey; Emmanuel Perry @manny_hockey; @DTMAboutHeart; @EvolvingWild

Many don't rhyme perfectly, I count that for and not against the value of putting them together here.

All of the details of their methodology can be found and best explained through their own resources in full thoroughness, but I'll add some especially relevant tidbits about (perceived and real) limitations here:

  • In general, what's been found by most data scientists around hockey is that offensive impacts(and forwards) are much easier to distill than defense, and even-strength play the same comparatively to special team situations, especially penalty kill. 
  • Certain player types are bearishly valued: Penalty differentials are one of the five big parts of these equations, and players who you may perceive to be rated too low often have a combination of the above point about lack of 5v5 offense, and a bad penalty differential. I believe the amount that penalty differential is punished and rewarded is entirely fair, and an understated(among mainstream media; casual fans) concern when evaluating players by traditional methods.
  • Even-strength shot share impacts are heavily weighted when it comes to defensive contributions; more than shot quality impacts This is due to the belief  that players cannot influence the save percentage of the goalies they play in front of. Many people part with the math crowd here.
  • Because of the self-professed (relative) limitations of evaluating penalty killing prowess in GAR, buoyed with the more precise ability to evaluate powerplay defencemen's ability, combined with the point above about shot quantity vs. quality means that most of the discrepancies between eye test and analytics is in the differing valuations of defensive defensemen and more gifted puck-movers and playmakers.
  • If you believe there to be limitations in the evaluation of defending, I would simply suggest you ding who you perceive to be the offensive defensemen who are poorer away from the puck, and boost the better sortie-stoppers as you see fit. Proceed with caution.
  • Each metric has trouble (obviously;understandably) with rookies. I'll fill in the gaps with my knowledge of what a player has to look like pre-NHL to look like something in the NHL.

Anaheim's gross. Though apparently they couldn't stop themselves from signing another couple bottom-pairing vets fully capable of negative-against-replacement play in Sustr and Schenn, indications I have from Ducks fans is that Petersson(#38, 2014) is knocking on the door and Carlyle's baked cookies. Pettersson's never been impressive from a points standpoint in Europe, but what we know about the SHL is it's a damn good league and if you're a good defender young of age over there you know a thing or two, couple that with the fact like similarly to the NHL but even more pronounced is the effect of getting almost no 5v5 points as a defenceman, regardless of ability. Petterson's 6'4", though not yet listed at 200lbs, and was the #7 ranked European by NHL CSS in their final rankings for the 2014 draft.

We also know things about Anaheim's defense drafting and honestly after getting what they did out of Josh Manson (converted forward out of the BCHL turned into one of the leagues premier shutdown guys, puke.) I just plain don't doubt any of their D selections anymore. He has very good relGF% numbers as a top-pairing guy in the AHL, so we're going to guess he's not going to be a negative player third pairing in his rookie year in the NHL. Truth be told that might not be flattering enough versus reality.

I don't have to write much about their top pairing and top four, that's why I haven't yet. Montour's unproven relative to the rest, but nothing seems unsustainable from cursory glances over his first full campaign.

The best D corps in the division, no contest - from myself.

These first two could have been in alphabetical order or in the order of quality and it'd come out the same.

What I mean is that you're looking at a Coyotes team with an actually enviable pile of defensemen.

There's a wrinkle here regarding Goligoski: He had a really, really bad half(or-so) season while miscast in deployment and generally floundering, and put up a negatively influential season, manifesting itself from a combination of negative penalty differential and negative defensive contribution. He'll be lower down the line-up and has had much better seasons before by the same metrics, so I'll say he'll wash out to replacement level on the third pairing.

The key to this configuration, for me, his Hjalmarsson has to not be done. Everything else is good, I think Chychrun may actually be underrated here, he's been talked about as sheltered but is in a spot where his level of success in that situation indicates he's well equipped for a promotion. The Coyotes really like him, and he'll get his shot.

The top pairing is a returning one that's not at the level of Anaheim's, but is sneaky close because of how much Ekmann-Larsson's dinged from the weighting on last season and Demers bouncing around last season. They'll probably end up closer to +20 as a sum in GAR than the 13 that's projected, by my estimation.

Key here is that there/s no-one negative unless Goligoski repeats results despite demotion, and that they have a workhorse top-pairing that can act as a true top pairing in a pinch as opposed to the usually top-four time on ice allocation scheme. This helps if one of the two second pairing guys go down, and the coach isn't comfortable with the spotted duo that results.

This is a very interesting situation to almost everyone following the league.

Dougie Hamilton is one of the very best defencemen in the league, full stop. He was 13 GAR last season(more than Giordano and Brodie's projection combined) that marks him as 99th percentile, he moves the puck extremely well out of his own zone as well as manning the point, puts up even strength offense not just by leeching points but by driving goal and shot rates, does well on the powerplay (even without being given the most PP! minutes!), and remarkably for a defenseman, draws more penalties than he takes, has an impressive arsenal of shots, list goes on.

It's always hard to parse the pairing when it comes to guys who play almost all their minutes together (model-builders admit this) however, and Giordano-Hamilton was like that. Teasing it out comes up Hamilton by GAR and WAR, we'll see. Giordano-Brodie was a little overly-famous pairing when Calgary ran a PDO heater all the way to the 2nd round of the playoffs, not sure if they'll be as good now, and I don't if they were as good as what was said then, considering Brodie's subsequent history away from Mark Giordano. Brodie certainly looks good by Corsica Player Rating, though.

Hamonic's been off for longer than he's been on as of late charitable consider him above water-level then consider his partner'll be a very young defenceman. Still, Hanifin is absolutely on track as a future top-four and we'll find out when exactly that future arrives(or alternative, if it ever does) in pretty short order.

Calgary shares with Edmonton a vulnerability to right shot maladies compounding, one guy goes down and it's top pairing Hamonic, 2nd pairing Stone. There's help on the farm, but they've made curious decisions there in the past in regards to floundering veterans staying in place over mathematically impressive solutions from Stockton.

Los Angeles has some real problems under the hood that they've been powering through with upgrades to their offensive engine both organic and artificial. That trend continues with Kovalchuk, but it's gonna be tough to outrun these negative guys, especially when they're veterans with ink to spare and you wonder if they know they're hurting and are waiting it out, or if reputations built during the cup years have hidden poor performance from their internal analysis. The team's top-heavy at forward too, and if you have one pairing adding value, one treading water and one hideously negative all fair in black and white, what you end up with is which pairing is on the ice swinging your forward line's results wildly and unless that's good, that's bad.

You end up with a situation where having Kopitar or Carter(subject to injury concerns) on the ice with the Doughty pairing is all world; either with the Muzzin pairing being just a hair above average(for a top six unit); the 3rd pairing sinks all boats regardless; and if you add that pairing to either bottom six line you hit the floor of the ocean.

I read something by Tyler Dellow where he mentioned that the Kings were the only team to have a better share of the shots when the opponent's top six was on the ice than they did when the opponent had their bottom six on the ice, and I bet that has a lot to do with that bottom pairing joining their bottom lines with Andreoff, Lewis, Mitchell et al on them and just getting crushed by other teams 3rd and 4th lines.

Now, individually, Oscar Fantenberg's a guy who's just too old to be given the benefit of the doubt on his North-American career based on his play so far in circumstances that should have been favourable. Phaneuf and (in a bordlerline sense) Martinez just seem done. That's who I was referring to in the leading paragraph. From there you've got Forbot playing the Methot role, and Muzzin trying his damndest with whatever he's gonna get. That second pairing sinks? They're a Doughty injury away from having a worse group than Vancouver.

This is what a situation that's almost identical with what happened in Winnipeg with Enstrom: Ryan entered just as veteran Martin aged out. He was bought out in spirit midway through the season, ans Joakim Ryan represented an example of the dividends of drafting and developing defensemen well. He's the most under-the-radar competent top-four guy, or at least Burns buoys him enough to make it look so. #198th overall in 2011. There's differencing defending from the AHL to the NHL, which is why you obviously wait until it happens but in terms of on-ice results in the AHL all signs were pointing to top-four potential with this guy draft position be damned. Again, I could be overstating this as there's the effect of the Burns pairing's deployment as well as Burns ability in general to consider, but if it works it works and it's working for them. Dillon & Demelo are just a few years removed from being utterly exposed by Pittsburgh in the playoffs; they were young then but not young enough to entirely excuse them but in terms of Pacific Division, regular ol' season patrolling they seem just fine at what they do.

Vlasic and Braun have both hit their 30's now, but they have an enduring style of play in both micro and macro time and it's bad business betting against them. Speaking of players the two-hundredth most favourite on draft day, Braun went three picks after Ryan's spot, four years earlier. Did anyone tell you life was fair?

Most everyone needs a mulligan on this whole team.

Engelland's carriable in the top four? And Theodore's that guy?

Yes and yes, apparently, and I'm more sure by eye of the latter (Theodore is a treat to watch) and as long as McNabb's brand of hockey sells that top pairing is in business.

I'm at a loss on these guys. Are they good enough to provide a boat to sail in for the non-Marshessault lines? The question marks are Merrill and again, Engelland, and I'm not a fan of the pairing concept employed here unless the superior is a Doughty-calibre guy, and even then it's a waste in my opinion.

Fleury mucked this whole thing up and I could so easily say so many things here that'd make me a fool in a matter of months. And I mean inordinately.

I'll ignore my own prescription: If McPhee wants to go all in with this core, he really needs to do that Karlsson dance or any other number that finds him a righty. Miller's the only true value-adder in my opinion here, and if Glass ruins the negotiation for you you're basically betting that he himself becomes the engine that makes a non-Marshessault line drive and you win that way because that's honestly what a pairing would do for you like say, Schmidt-Karlsson or McNabb-Karlsson. It's time sensitive too.

Bottom line, these guys work and they're about middle of the road for the division. It's good enough, but they're only gonna go as far as Marshessault-Karlsson-Smith takes them until a Suzuki or Glass bursts onto the scene. Again, time sensitive.

Also again, Fleury though. And I mean that both ways.

Below the spaghetti meteor, here lies the worst defense in the division.

Tanev's a victim, like Larsson of the offense-emphasis in these metrics.

They do fairly rate how poor their offensive contributions really are, though. That part's just the truth, but I saw him(Tanev) take some unfair criticism in the late '17 dust-up over Nylander trade ideas.

Stecher I like, he might be a 4 and I'm not trying to be funny with that, that's value. If you want to giggle at Vancouver I'll inform you that we're paying more for less.

Gudbranson is a guy that, obviously he's victim to his player-typing here but I'm honestly not certain he's been effective at what he's supposed to be. He's not Adam Larsson. What the Canucks have got here is a glowing-red weak-point third-pairing that's more typically of a team that's got enough substance somewhere higher up to be called top-heavy.

The real tragedy here? The fan aren't getting any Quinton Hughes viewings yet. That guy is special.

They need two more.

Edler's dormant explosives at this point, there's a chance he breaks and Tanev's not gonna carry the next one down the line because it's Del-Zotto and that guy is whole-foods Jack Johnson.

You know what I did here? I took Klefbom's 16-17 numbers and said that's the guy, that's who he is. That's my Oilers Optimism. @ me.

Know what else I did? I bet you do, because you can see it. For your information, Kevin Gravel was right around replacement but speaking of replacement I got us another Swede.

Enstrom's a guy who's all defensive contribution that still shows up good in the numbers. I took his 16-17 for the proj. GAR. To-be quite honest with you, I don't even know if he's a Chiarelli guy. I think, though, based on comments he is going to add but based on more comments prior it might be waiver wire. Gravel was definitely a 7D plan, I think Pete believes in him but in that role and not another.

I want to talk about Keegan Lowe. He did work on the Condors last year, the good kind. If I'm projecting Petersson on rels with Anaheim, Keegan's older but he's got those same credentials. This is a training camp thing but I'm not uncomfortable with the idea at all. Whoever you pick up, they're gonna want to play so maybe Lowe's there if waivers won't bring you anything and Toby doesn't pick up the phone.

I mentioned this all the way in the beginning but this is about depth. My plan would have been 77-6; 25-83 in the top four to start with but there's no cover for either of the lefties

I don't have the smartest touch when it comes to taking the temperature on the Oilers fanbase, though I do try and I feel comfortable forecasting that coming in, people are going to  be more sure of Darnell than they are of Oscar, and I'm opposite ways around on it. Far as I can tell, Nurse had one good half one bad half and so did Klefbom but there was no overlap and I think either people were tuned out or they had already made up their mind before Klefbom found redemption along the way.

Klefbom's healthy and that top pairing runs just fine, I'm sure of it. Through all of it, the bum shoulder, the shot demand, and the defensive scheme switch-up by the actual goals Klefbom closed okay.

I talked about all this in The Best Player Available but I'm strong on Benning in his spot, too. It's just the vulnerability to injuries that absolutely will happen. That's what we know.

But let's for a moment imagine that the probability of any player sustaining a long-term injury is equal.

I would say that, from there, there are the same number of season-sewering possibilities up front as there are on the back-end.

Klefbom or Larsson or Benning goes down, and everyone moves up a spot?

That's bad, but if Draisaitl, Nuge or McDavid goes down it's the same thing.

You had one piece of cover in one spot(LHD) and that was a good thing, but even if Sekera was 100% the vulnerability of the team to forward injury is still more dangerous overall. That's the real problem with this team, as of now there's three top six forwards. It was always going to be outside chance at the playoffs. The only way the non-McDavid minutes were even-ish in shot-share is by gaming the volume with point shots. Also, that discrepancy between the goals and the shots overall away from McDavid is going to be less the shots-against being defended poorly and resulting in relatively more goals, and more the shots-for lacking finishers behind them. I know this is a contradiction via the GF and GA rates relative to league average from last year, but I'm speaking from a place where I'm assuming the 5v5 defending will be different(we've been told this).

That being said, Talbot can save us all. But do we want him to? Do we want a first-round playoff exit that encourages status quo?

It seems infinitely greedy to ask for another difference-making forward from the draft, but honestly there's been more traded away than gained since 2015, assuming Puljujarvi doesn't defy the current developmental processes that appear to be inverse to what every other team is doing in their attempts to turn their top draft pick forwards into strong-links. For Kailer, they seem to like him so it's just up to chance whether or not he's Eberle or beyond.

I'm not sure the current organisation can put together what I believe to be a 'cup-guarantee' window, 5 years as a top-5 team, within McDavid's coming eight years with this kind of top talent deficit, given what's been done with a surplus.

I'm rambling, so I'll quit doomsday-preaching for now.

In conclusion, if Sekera's absence changes the year-long outlook of this team for you, we either differ in our evaluation of Sekera or our evaluation of the 2018-19 Edmonton Oilers as they stood before the news hit.


I had an exchange that added some clarity and information in the comments of, and thought I'd add it as a footnote here. I chopped up his original comment to respond to each facet individually:

Jordan said:

Hey Wilde,

Was just thinking about numbers from your projections for the D Corps in the Pacific and I have some questions:

Based on your understanding of the numbers and what they represent, could they be used to project either the value of a specific pairing, or of the defense corp(se) as a whole?

I was curious to see if the numbers correlated to what I thought I understood about the overall strength of a given team’s d-corps, so I aggregated them with the following results:

My response:

I’m glad you asked, I’ve mentioned this before but my main weakness in ‘predicting’ defense corps is putting individual players together into pairings.

For example, Shea Theodore + Deryk Engelland is not a pairing I thought would work.

I didn’t think Sekera + Russell would either.

I can rate defensemen fine, but when it comes to arbitrating who can carry who and who will sink who, weighting TOI shares to project the totality of the Dcorps I’m wrong a lot.

In terms of just the math of putting the numbers together, it’s tough, because of stuff like who plays with who and for how long and how they’re used.

For example, if a D corps’ highest rated guy ends up playing the least and with the worst partner, the net influence of his higher rating is lower

(that's simplified)

You can see in your document Anaheim is the best, and that was my take without putting the numbers together – they’re the easiest to do this with because there’s almost no ambiguity in year-to-year usage and context differences – only the bottom pairing is going to be different and there’s a rookie that’s decently promising and a veteran that’s a massive upgrade on last years’ Bieksa contribution.

You can also see this a bit with the difference between my evalutation of Arizona and your arithmetic: They’re held down by the difference between my personal projection of Ekmann-Larsson being much better and Goligoski being better sheltered and Chychrun continuing his trajectory in the GAR department.


Does the GAR, WAR or PR data take any circumstances of play (Common Linemates, PP/PK/EV TOI, or QualComp data into consideration, or are they “raw” numbers?

Yes and yes and yes, in the preamble about their limitations and discrepancies in evaluation versus traditional measures I talked a bit about disciplines and QoC and QoT, here’s the links to their methodology along with quick summary of components:

(The QoT and QoC and deployment are factored into each of the following)

Chace’s GAR:

3) Even Strength Offence
4) Even Strength Defense
5) Power-play Offence
6) Penalty Differential
7) Extras

Emmanuel Perry’s WAR:

Offensive shot rates

Defensive shot rates

Offensive shot quality

Defensive shot quality


Penalties taken

Penalties drawn

Zonal transitions

(Corsica Player Rating is a weighted stacking of per game WAR from the past 120 games, as I understand it)


Interested to hear feedback – does this provide insight, or does this provide noise?

No, this is useful and I should have included it, up until the ‘Overall D corps rank’. It’s good to see where their guys are by which model and how, as opposed to having to scroll around everywhere. For example, the only place the Oilers are good is the GAR, and you can bet it’s from me grabbing Oscar’s 2016-17 numbers.

Which then you can ask an interesting question: Is the most important facet of this Dcorps having a 2017 version of Oscar Klefbom? I’d say yes.

Tuesday, August 14, 2018

The Best Player Available (2/2) (Going long on the future of the Oilers Dmen)

The Oilers, the Oilers used to have a problem.

There's no Hedbergian twist here, they don't still do too.

The problem was the enormous imbalance between right and left shooting defencemen drafted by Edmonton, constructing what was dubbed by Allan Mitchell as the Leftorium:

Note not only the quantity, but the quality of the imbalance. Not only was there only one right-shooter selected in four drafts, the only two first-rounders are left shot guys and, adding up the values via Blue Bullet Report's Chart, the ratio of value spending between both sides is about sixty-eight to one.

The result? The NHL starting six's right side is a traded-for guy, a college free agent, and a left-handed veteran free agent playing his off-side.

Chiarelli walked the walk when it came to the lefty-righty paradigm, and as I write this, we stand with this:

Again, note the quality and quantity.

The only elite prospect, and the only one further than that who can be (reasonably, arbitrated by myself) expected to have a real shot at being a top-4 NHL defencemen, are both right shots. By the previous methodology using draft value, starting in 2015 the Oilers spent six picks on lefties, six on righties, and a value of ~8 on left, ~29 on right(the value is mostly Bouchard at 23.6, just as in the exercise prior the value was mostly Klef; Nurse).

Let's think about the NHL group of Edmonton defensemen as a series of probabilities.

Right now, on the right side there's Larsson who can complete a top four pairing, Benning who has done so in part in his rookie year, and then again in part of last year, and then Russell who can play the third pairing okay.

Larsson has a contract that runs through 2021, and since his value largely doesn't manifest itself in the box score, should actually not warrant that much of a raise. A comparable points-wise. age-wise and draft-position-wise is Karl Alzner, who hit unrestricted free-agency last summer as a 20-point, 28 year-old, former top-5 pick defenceman and signed a contract at 4.625MM for 5 years. The salary cap was 78 million at the time, making that contract roughly 6% of the cap. These financial details and Larsson's age make the future of the Oilers D corps about finding one more top-four capable guy on the right side. Outside of injury, the probability Larsson can be kept on board and contributing long-term as one of the top-four RHD is honestly near one-hundred percent.

Then comes the first relevant probability: Is Matt Benning going to be a top-four defenceman?

He's been one at times. The only thing flashy about Benning's play is his trademark open-ice hits, but good things generally happen when he's on the ice, particularly so in his opening campaign when he replaced Russell on a pairing with Sekera. His is a contribution that commonly falls under the radar, I'm often critical of Peter Chiarelli, and when his supporters come to the aid of his reputation they always(as they should) point to his good deals, typically Maroon and Talbot, but Matt Benning was a college free-agent that added value in a tough role in a rookie season, and that's something, and he'll be something for awhile and I'd count him as the best acquisition or tied with Talbot.

He's the 97th defenceman by Corsica player ratings, he's a projected +5.63 goals against replacement by Chace McCallum's model, he's a positive WAR guy by EvolvingWild's model and in general he just looks good by my parsing of the data as well. If he's 2RD on your team I wouldn't say that's a strength in the roster, but he doesn't appear to be a drag on anything there. There's still not a ton of track-record to him, though, and we don't have a solid idea what his 'normal' is because of how hard it was to evaluate defenders last year under the 5v5 defensive scheme deployed, and the concussion trouble the year before. I would put him as having an above 50% chance of contributing league average level play on the second pairing for some years to come.

Then comes the hole at the third pairing,  currently filled by off-handed Kris Russell. Truth be told there's more third pairing defencemen in the NHL than there is jobs, and most guys of Evan Bouchard's pedigree will make it at least that far. Ethan Bear has a number of things to work on to get to the NHL, then a number more to make it to second pairing level. The game here is you hope that the probability of Evan Bouchard becoming second pairing plus the probability of Ethan Bear becoming second pairing adds up to, at least, the probability of Matt Benning not being second pairing. That way the bet is hedged. Past that, I think where we stand right now the chance that either Ethan Bear or Evan Bouchard becomes a third pairing RHD is near certain or has a sum over 100%, and what this all adds up to is there's a lot that has to go wrong before there's a serious, debilitating vacancy at RHD in the Oilers system.

The age of all these guys being <26 means you can probably leave that alone.

Now, on the left side you've got something interesting going on. in 2016-17, Oscar Klefbom performed at what appeared to be a top pairing level. In 2017-18, Darnell Nurse performed at what appeared to be a top-four level. Add this up and hedge it with the chances Sekera recovers, and we're looking at those two top-four left shot spots being pretty reasonably filled.

What about that Sekera, though? He's got three more years at five and a half million per, and his health is in question. This year, if he recovers fully, the Oilers will have a miles-better third pairing lefty than most of the league.

Marcus Petersson, Alex Goligoski, Nathan Beaulieu, Ryan Murray, Brett Kulak, Erik Gustafsson, Mark Methot, Jon Ericsson, MacKenzie Weegar, Oscar Fantenberg, Nick Seeler, Jordie Benn, Egor Yakolev, Matt Irwin, Thomas Hickey, Ben Smith, Matt Borowiecki, Robert Hagg, Jack Johnson, Brendan Dillon, Jay Bouwmeetster, Braydon Coburn, Ben Hutton, Ben Chiarot.

Those are the names of the third-pairing left-side guys that I think, to varying degrees, can not reasonably be believed to deliver circa 2017 Andrei Sekera level of play in the upcoming NHL season.

There's 25 of them.

That means there's cover at this position. The Edmonton Oilers are not losing ground on the rest of the NHL by not upgrading here with some immediacy, and the on the other side of the same coin have expiring rights to that luxury via Sekera aging out, and the heaviness of his contract in this team's salary cap context.

This makes it a prime candidate for a well-timed upgrade via ELC.

Upgrading a position with an entry-level contract is extremely hard to pull off without either having lottery-level talent coming into the position, below replacement level talent pre-existing in the position, or both.

The Edmonton Oilers upgraded at 1C with an ELC, you've seen what that looks like.

In a less extreme example, look at the top lines of the Colorado Avalanche and the Boston Bruins: Two pre-existing duos with thirds that were upgraded via elite ELC talent. Typically rosters need to be balanced by having units rounded out by closer to replacement-level talent, but the financial power of the ELC allows for the opposite. When Mikko Rantanen has to sign a new contract, or the Avalanche in general get closer to the cap, he may eventually have to be moved off of that line in the talent crunch that comes with the cap crunch.

The Oilers can do this on a defense pairing: Whatever the sum of Matt Benning, Evan Bouchard and Ethan Bears' career arcs is  in a few years has one of them sitting on the third pairing, already being an above-average contributor at that roster spot.

This is where a best player available, say an incomplete but elite talent on left defense can morph the pairing into a terror when the spot in the order for them is considered. In the mold of a Sam Girard or a Ty Smith, a player who drops out of their talent tier at the draft simply because they shoot left and they're not big could be had by the Oilers at the mid-first-round (I believe the Oilers could finish anywhere from 10th-24h in the league this year) and be ready to step in at 21 years of age, and the defence corps iced could resemble something like this:

Klefbom - Larsson
Nurse - Bouchard/Bear/Benning
BPA - Bouchard/Bear/Benning

All this, in a league where a Stanley Cup Finalist can lose the entire series off of a weak-spot third pairing, while paradoxically getting there in spite of them. This configuration, however, ensures that there is would be no pairing that the coaching staff is afraid to send over the boards, no situation where the top-four is run ragged from many difficult hours in the first three rounds of the playoffs.

I don't believe the Oilers will be in a position to draft a real difference-maker forward for the remainder of McDavid's godsent contract. Scouting departments have forwards near nailed-down, with most of the river-pushers going in the very top end of the draft, even the mythical Johnny Gaudreau type third round selections are going extinct as teams wise up to offense predicting offense.

This means that if there's going to be a third line of forwards that push the goalshare in a significant way, it's going to have to involve the continued development of Jesse Puljujarvi into a force capable of levelling the opposition as an individual influence.

Where the efficiency still exists in the draft, where talent drops out of tiers most often is leftie D height optional, and the Oilers have the precise opportunity to ice a level of balance in their blue-line group that ensures expensive free agents and elusive trade targets need not apply.

When I was watching the Hlinka-Gretzky, truth be told none of the Dmen particularly stood out. As the year unfolds, you can be sure that I will be referencing this concept when narrowing down the draft options for 2019.

Names of interest, just based on the data:

Yegor Bryutov - light Russian, was 16 for most of an entire year that he put up crazy points for his age in the MHL. 3rd ranked D by Emmanuel Perry, seldom mentioned elsewhere

Bowen Byram - may go too high, Vancouver Giants guy, very good numbers for a late-birthday draft-1 defenceman last year

Cam York - part of the ridiculous US 2001 group, fits our profile near perfectly. Of all listed, you'll likely hear the most about him.

Henry Thrun - the biggest of this list, numbers roughly in line so far with McAvoy and Hanifin, another American 01.

Kim Nousianen - November guy, good numbers and he fits our profile, we'll find out really early from the math whether he's worth the first rounder because of his maturity relative to the other guys.

Jordan Spence - another guy that, despite not being older for the class will need to establish himself early as it's his rookie CHL year. 

I'll be updating on this phenomenon, and depending on how sour the year goes I'll be more literally applying the ELC upgrade concept to left wing instead, as noted before it's gated by draft position. Hypothetically duos of McDavid and Draisaitl with existing wingers like Yamamoto and Puljujarvi could be rounded out and turned up to ridiculous with an elite third talent.

From there I'll look at the before and afters of adding the Pastrnaks and Rantanens of the world to the Patrice Bergerons and Nathan MacKinnons, and project what happens when you do that except with McDavid or Draisaitl.

Which should be fun, though it would be presumably surrounded by a situation where the Oiler fans are talking draft in the middle of the year and there may be population cullings and public sacrifices, we'll have to play it by ear.

Until then, thank you for reading.