Who doesn’t like to see their team winning? Does it matter why the team won? To the everyday fan, the ‘why’ doesn’t matter: Their team scored more goals than the opposition. End of story. Plus, wins leave you elated with the feeling that anything is possible. But if the fan wants to see a consistently competitive team, a team that is a regular playoff contender, how the team wins matters tremendously. When I focus on the question of why for the Oilers, I cannot help but conclude that the recent wins are deceiving.
With that in mind, I want to ease readers into my analysis of the Oilers’ progress compared to last season. Looking at the analysis one way, the results are promising, but in another way, the results can be slightly discouraging.
In a previous post, I explained how I computed my progress indices by comparing games to the respective 2014/15 season series. Why I used this comparative analysis is because simply looking at their current Weighted Shot (WghSh) or Shot Attempt (SAT) differentials, we don’t take into account the Oilers’ quality of competition. For instance, in their first 10 games, the Oilers played 7 games against top-5 teams (as measured by Weighted Shot differentials). We would not expect the Oilers’ shot metric differentials to be all that great against the best teams, so how would we know whether they’ve improved?
But if we compare their shot metrics from each game to their 2014/15 season series metrics (vs. the same team), we take care of quality of competition. For example, if last season the Oilers’ average SAT% against Los Angeles was 45%, but they achieved 48% this season, that would be a relative improvement, even though 48% is a poor differential in an absolute sense. If their SAT% against Buffalo was 54% last season, but only 51% this season, this would be a worse performance relatively speaking, even though 51% is a good differential in general.
Using data from war-on-ice, I compared 4 shot metrics from every game. For those unfamiliar to hockey analytics, the most confusing metric is Weighted Shots. I borrow from Matt Cane’s model that places 5 times more weight on goals versus shot attempts. That means both goals and shot attempts are part of Weighted Shots with goals getting 1 point and shot attempts receiving 0.2 points.
The 2nd metric is simply shot attempts (SAT), which includes blocked shots, missed shots, and shots on goal. SAT (a.k.a. Corsi) has been shown to be one the best predictors of a team’s future goal differential. Weighted Shots are also strong predictors, slightly better than shot attempts in fact, but to tease apart defense from goaltender performance, and offense from “puck luck,” (i.e., statistically high scoring rates) then shot metrics, like SAT, that exclude goals are useful. The 3rd metric is Scoring Chances (SC; defined by war-on-ice), which helps account for shot quality. The 4th metric is a proxy for the highest quality shot, which are scoring chances from the slot and are called High Danger Scoring Chances (HSC; defined by war-on-ice).
Shot Differentials
With our metrics defined, let’s have a look at their 2015/16-to-2014/15 comparative differentials using 5-game blocks.
To explain the graph a bit more, each differential is computed by dividing the ‘For’ metrics by the sum of the ‘For’ and ‘Against’ metrics. For example, SAT% = Shot Attempts For / (Shot Attempts For + Shot Attempts Against). For each 5-game block, I computed the average quality of competition using Weighted Shot differentials, which I put In parentheses. Teams greater than 50% are typically playoff-bound teams. So as you can see, the first 10 games were the toughest of the season.
Although little progress is evident in the first 5 games, albeit against some very tough competition, the progress indices are consistently positive up until game 25. That is, regardless of the quality of competition, the Oilers showed improvement in shot attempt and scoring chance differentials. One qualifier, though, is their inconsistency in High Danger Scoring Chances, which I’ll talk about more below when I look at defense and offense separately.
What happened after game 25? The Oilers’ winning streak of course! It makes perfect sense! Their differentials plunge below their levels of last season and yet, they win. For games 26-29, we can thank Anders “The Giant” Nilsson. Not much else to say there. Game 30, though, was nothing short of a shooting gallery for both the Oilers and New York Rangers. As a brief respite from the shellacking from the previous 4 games, the Oilers’ progress indices in the Rangers’ game were slightly positive. (This uptick is not seen in the graph because the 5-game average is negative.)
What’s going on after game 25? Their roster has been weakened by injuries and by diminished play of a few key players. Although the team maintained their improvement after McDavid broke his collarbone–you read that correctly, the Oilers were as good without McDavid–, the loss of McDavid’s linemates, Yakupov and Pouliot, appears to be straining the team’s offense.
Defensively, Fayne struggled mightily through this period. Prior to his re-assignment to the Condors, Fayne’s (score-adjusted) Relative SAT% for the previous 5 games was -11.7%. Even worse, with Fayne on the ice, the Oilers’ Scoring Chance differential was 26.2%, which contrasts with the team’s respectable 49.8% SC% without Fayne. Finally, his Relative High Danger SC% was -27.7%. No matter how we slice the metrics, Fayne’s performance had deteriorated considerably.
To add injury to injury, Klefbom was hurt in game 30 with a broken finger. Still not sure how long he’ll be out. Nikita Nikitin has been called up to replace him. I’m not optimistic that Nikitin will outperform Fayne, never mind Klefbom, but time will tell.
I wrote the above portion before watching and compiling data from games 31 and 32, which were against Boston and New York Rangers, respectively. In Boston, Talbot rescued the Oilers from being drubbed on the scoreboard with 47 saves. Different goalie, same story. The metrics indicated a worse performance compared to last season. However, against the Rangers, the metrics indicated a slight improvement over last season.
Offense
But back to the first 30 games, let me dig a little deeper and isolate offense from defense. The Oilers are scoring more this year, but is that because they generating more quality chances? Has their defense improved? First, I’ll examine the Oilers’ offense.
Improvement in offense has been inconsistent, and even though the indices (except for HSC), on average, are positive. The lack of improvement of High Danger Scoring Chances still suggests that, like previous seasons, the Oilers still struggle to create opportunities from the slot area, whether that’s by driving to the net, passing, or increased net traffic for deflections and rebounds. Games 21 to 25 were the height of their offensive improvement, which was followed by their weakest offense of the season from games 26 to 29.
Dissecting games 21-25, the Oilers did an excellent job of generating more shot attempts (+28/60 minutes compared to 14/15) and scoring chances (+18/60) against Washington, along with solid improvement in shot attempt generation (+5/60 on average) against Carolina, Detroit, & Pittsburgh. But this increased number of attempts didn’t necessarily result in boosting their scoring chance generation against Carolina (-2.8/60) and Pittsbrugh (-5/60). Against Detroit, though, there was strong improvement in scoring chance generation (+11/60).
Breaking down offense to the level of lines (Note: These numbers are not relative to 2014/15, but simply 2015/16 figures.), the Hall–Draisaitl–Purcell combination was the primary offense-generating machine with a score-adjusted SAT% of 59.2% and a SC% of 55.1%. In terms of individual performance, Hall had 18 and Draisaitl had 13 scoring chances to lead the way with Purcell firing 11 for a total of 42. That’s just over 8 scoring chances per game for this line. The Nugent-Hopkins–Eberle–Pouliot line combined for a score-adjusted SAT% of 51.1%, but only a 36% Scoring Chance differential. Ahead in the shot attempt game, but a lack of quality shots told a different story.
As to other forwards, Pakarinen was firing strong against Carolina & Detroit with 8 total scoring chances, but nothing at all in the other 3 games. Letestu and Korpikoski also had 8 each over 5 games. Yakupov was hurt by the 2nd game into this road-trip and had 2 scoring chances in one game. Lander recorded no scoring chances at all.
The Hall–Draisaitl–Purcell combination, then, appears to have been the primary catalyst to the improved offense during this 5-game stretch.
What changed on offense from games 26 to 29? The Oilers were without Yakupov and Pouliot due to injury. In effect, the Oilers 2nd forward line (McDavid–Pouliot–Yakupov) were all out. That meant more ice-time for marginal bottom-6 players, like Korpikoski and Gazdic, along with AHLers getting more ice-time (Pakarinen) and being called up (Juhar Khaira). From games 26 onward, the bottom-6 forward’s SAT% tells us the story: Gazdic – 44%, Lander – 43%, Pakarinen – 39%, Letestu – 34%, & Korpikoski – 32%.
To put this into perspective, the 20 worst SAT differentials (T.O.I. > 200 min) in the league range from 36% to 42%. For the record, Korpikoski happens to have the worst SAT% in the league. Ouch!
Defense
Next, let’s take a look at the comparative metrics for defense.
Defensive improvement, at least measured with Scoring Chances Against, had been consistent to game 25, but like the offense, dropped dramatically from game 26 onward. I already noted Fayne’s struggles, along with the entire loss of the 2L forwards to injury. These same roster deficiencies also help explain the team’s weakened defense. But is anyone else on defense that we can turn the analytical eye? This analysis won’t be popular, but here I go.
Nurse: Baptism by Fire
McLellan has been deploying Darnell Nurse as a top-2 defenceman. At first, things were going well. But as of late, Nurse’s game has soured a bit. Jonathan Willis gives his own take how Nurse’s play is not up to snuff. I would like to expand on that a bit. Willis’ article was not kindly received, if the comments section reflect the general fan-base.
With metrics, I noted how poorly the team is doing with when the bottom-6 forwards are on the ice. Nurse’s numbers are no better. From game 26 onward, the team’s SAT% with Nurse is 34%. The goal-focused fan may not have noticed though. With the Nurse on the ice, the team’s goal-differential is +1. It seems the gods of good fortune (referred to as PDO in hockey analytics) are smiling on Nurse and have blessed the team with a scoring rate of 24.2%! (The league-average scoring rate is 8%.)
With Sekera since game 26, Nurse’s numbers are not much better: Their combined SAT% is 35%. From the start of Nurse’s season, the team has allowed 28 Scoring Chances Against per hour. Relative to the other Oilers’ defense, only Fayne is worse at 29.6 per hour. In terms of Nurse’s overall Relative Scoring Chance differential (-5.3%), only Korpikoski (-9.6%) and Letestu (-6.4%) are worse. Not to imply that Nurse has been over his head the entire time. His SAT% prior to game 26 was a very respectable 48%.
G. Money, the blogger behind OilersNerdAlert.com, has developed an innovative (still tentative, but promising) metric to assess defenvemen that incorporates shot quality, namely, shot distance and shot type (e.g., slap shot, backhand, wrist shot). He calls the metric Dangerous Fenwick and it includes all unblocked shots. He has been tracking the Oilers’s metrics game-by-game. By his calculations, a Dangerous Fenwick Against (DFA) of under 39 (per hour) is “good” and conversely, greater than 39 is “bad.” From game 26 onward, Nurse’s DFA per hour is 49.9.
Finally, if we look at Nurse’s dCorsi score. Delta Corsi (dCorsi) is the difference between a player’s observed and expected SAT (a.k.a. Corsi). Expected SAT is the value an average defenceman would get in the same context (i.e., quality of competition, quality of teammates, and zone starts). Nurse’s his dCorsi is just over -8 per hour. Factor in his top-pairing ice-time, the impact on the team is a -50 shot attempt differential. In other words, he is performing well below what we would expect from an average defenceman with similar usage.
Because there is consistency in what all these metrics (Shot Attempts, Scoring Chances, Dangerous Fenwick, and dCorsi) tell us about Nurse, I believe this gives more weight to the conclusion that he is under-performing. I did not have intentions to discuss Nurse when I first started writing this article. But the more I wanted to explain why the Oilers’ woes after game 25, the more I had to break things down, which lead me to focus on a few individuals. Unfortunately, Nurse did not shine under the analytical spotlight.
The Oilers are in a defensive conundrum. Nurse is overwhelmed, and in my ideal world, should be sheltered by having him play against softer competition, if not returned to the AHL. Yet, Fayne has played so poorly that he was demoted to the AHL and now Klefbom is hurt. But if Nurse is sheltered or sent down, who takes his place? I don’t have an answer. Maybe McLellan doesn’t either. I can only assume that McLellan is deploying the best he can with who he has available to him. But the longer Nurse gets crushed in the shot counts, the more I would like to see him taken out of the limelight.
Summary
The downturn after game 25 has been substantial; wiping out improvement during first 25 games by over half. Still, shot attempt differential has improved by 2.2 per hour and scoring chance differential by 3.2 per hour (see table below). Despite the downswing, I am hopeful that first 25 games are a more reliable indicator of improvement. I believe that the Oilers’ recent struggles are mostly due to injuries rather than deteriorating play of individual players, although this is definitely part of it.
What do readers think? Are the wins enough for you? (For some, winning is enough.) Do you think the analysis above supports my conclusion, specifically, that the recent downturn is due more to injury than deteriorating play? What would you do with Nurse, if anything?
Data courtesy of war-on-ice and puckalytics.
