On McDavid Day, after drafting Connor McDavid first overall, the Oilers traded several of their draft picks. The specifics are a little messy, but one transaction involved trading away defenseman, Martin Marincin, to Toronto, which translated into acquiring defenseman, Eric Gryba, from Ottawa.

Quality of Competition Analysis

Not knowing a thing about Gryba, I decided to do analysis of the trade. First, though, some background on the players. Gryba was drafted in 2006 in the 3rd round. He is 27 years old and has played 3 seasons in the NHL. At 27, he is in his prime, and because of that, I don’t expect to see much improvement in his play, although it’s not impossible. Marincin was drafted in 2010 in the 2nd round. He is 23 years old and has played 2 seasons in the NHL. Marincin is not yet in his prime, but will hit them next year as an NHL player’s prime years tend to be from 24 to 29 years of age. In theory, at least, Marincin has a higher ceiling of improvement than Gryba.

For my analysis, I used even-strength (5v5) data from the 2014/15 season, unless otherwise stated. Based on David Johnson’s WOWY (With-or-Without-You) tables, I computed some “new” metrics I believed might be useful in comparing Gryba with Marincin. I often read the argument that runs something like, “But he plays tougher competition, which accounts for his lower Corsi.” This may be true, but it is an assumption that can be tested. Why not compare players at similar levels of competition? Using the WOWY opponent tables, that is what I did.

Introducing the “Break-Even Quality of Competition SAT%”

One of the first metrics I computed I refer to as the Break-Even Quality of Competition Shot-Attempt Differential. That’s a mouthful. The shortened form is Break-Even QualComp SAT%, or even shorter, Break-Even SAT%. (I use SAT% instead of the term, Corsi.) Using the opponent’s shot-attempt differential, this metric identifies the average level of competition in which a player achieves a neutral shot-attempt differential (i.e., SAT% = 50%). Put another way, it indicates the toughness of competition in which a player begins to hold their own. The sample size of this metric is about 50 opponents with a range of +/-2%. For example, if a player’s Break-Even QualComp SAT% is 52%, the range is from 50% to 54%.

Marincin’s Break-Even SAT% was 50.9%, which includes such opponents as Chicago’s defensemen, Mikal Rozsival and Brent Seabrook, along with forwards Milan Lucic (formerly of Boston; now of Los Angeles) and Alex Steen of St. Louis. This looks to be some decent competition that a top-4, or even top-pairing defenseman would be expected to handle. In contrast, Gryba’s Break-Even SAT% was 47%–4% less than that of Marincin–, which is substantial. Competition with a 47% shot-attempt differential includes defensemen Jack Johnson and David Savard (Columbus), as well as forwards Dominic Moore (New York Rangers) and Trevor Smith (Toronto). Gryba’s break-even quality of competition originated from very weak possession teams (e.g., Columbus & Toronto), or from moderately strong possession teams (e.g., New York Rangers), but who were lower on the depth chart.

On the graph below, I expand on this quality of competition (QualComp) comparison by showing each player’s shot-attempt differential at similar levels of competition. Starting from the left, the graph shows the players’ SAT% from the toughest competition (56% SAT%; e.g., Joe Thornton of San Jose, Anton Stralman of Tampa Bay) to opponents with a 46% shot attempt differential (e.g., top-line players on weak possession teams & bottom-6 players on stronger possession teams). At almost every level of competition, except for opposition with a 54% shot-attempt differential, Marincin’s SAT% exceeded Gryba’s differential. Also noteworthy, against the toughest levels of competition, both defensemen were thoroughly dominated. Against elite competition (QualComp SAT% = 56%), Gryba’s shot-attempt differential fell below 40%! This is Andrew Ference territory. (I’m not exaggerating. Against the toughest competition, Ference’s SAT% was indeed below 40%.)

High & Medium Probability Shots: Defense

Noting this discrepant performance that seems to show the Oilers gave up more than what they gained, I was disappointed. I carried on with the analysis, though. Shot attempt metrics are a useful, but very blunt measure. Fortunately, war-on-ice.com now provides detailed shot metric data for each player. Specifically, shots on goal are broken down by their location and translated into 3 primary scoring zones, which the war-on-ice authors identify as low, medium, and high-probability zones. For my analysis, I treat the medium and high-probability zones as the most informative. The high-probability zone is the slot area. From the slot, teams score at an average rate of about 20%. The medium-probability zone is the immediate area surrounding the slot in which teams score at an average rate of 8%.

With Marincin on the ice, the Oilers allowed 26% more high-probability shots relative to the league. (That’s pretty bad, if you hadn’t guessed.) When Gryba was on the ice, the Senators allowed 6% fewer shots relative to the league. However, the Senators were much better defensively than the Oilers. Relative to the league average, Ottawa allowed 4.2% fewer shots, whereas Edmonton allowed 23% more shots from the slot. Noting this huge difference of 27.2%, controlling for each team’s defense metrics was needed. This can be done by examining each player’s shots-against metrics in relation to their teams.

Relative to his team, Marincin (and his on-ice teammates) allowed 2% more shots from the slot, whereas Gryba and his teammates allowed 4.3% more shots from the slot. That is, both teams were worse at suppressing high-probability shots with Ottawa affected slightly more than Edmonton. However, Gryba’s impact did not result in the Senators suppression of high-probability shots to fall below league average. Marincin’s impact on the Oilers’ defense, though, made a bad defensive situation slightly worse.

In the medium-probability zone, Marincin and his on-ice company allowed 2% more shots than the team average. In comparison, Gryba and his teammates allowed 10.7% fewer shots relative to the team. Defending in this area, Gryba and crew seemed to be doing a better job, or did they? One could argue that there are fewer shots in medium-probability zone because more shots were allowed from the slot! If that’s true, what’s Marincin’s excuse? Edmonton’s defense was worse in both the medium and high-probability zones when he was on the ice.

All that being said, given that shots from the slot are 2.5 times more dangerous than medium-probability shots, I assigned high-probability shots more weight. To summarize, both Marincin and Gryba seemed to weaken their respective team’s suppression of very high quality shots.

Shot Attempt Metrics: Player’s Impact on Teammates

In addition to the quality of competition, a player’s shot-attempt differential depends heavily on the quality of his teammates. David Johnson’s site provides an overall shot-attempt differential metric, including for & against shot attempts, relative to teammates. This measure can suggest if a player drives possession. Maricin’s SAT% relative to his teammates was +1.1%, which is mostly derived from his influence in helping the team generate more shot attempts (+2.38 shot-attempts for/60 minutes). In contrast, Gryba’s SAT% relative to his teammates was -4.0%, which is mostly derived from his apparent inability to help the Senators generate more shot attempts (-7.2 shot-attempts for/60 minutes).

These values were useful as it seemed to show Marincin was significantly better than Gryba in helping his team generate offense. Still, I wanted to look at the underlying numbers of these metrics. so I unpacked the players’ WOWY teammate tables. I focused on teammates with whom the players shared at least 10% of their ice-time. Specifically, I examined how much Marincin and Gryba affected their teammates’ shot-attempt differentials.

Overall, Marincin’s impact on his teammates was, in a word, neutral. His teammates’ SAT% dropped slightly by -0.17% (-0.10 shot-attempts/60). Breaking it down by individual teammates, the players he appeared to help the most were Gordon (+4.2%), Nugent-Hopkins (+2.4%), and Yakupov (+3.7%). The teammates he negatively impacted were Hendricks (-4.5%) and Klinkhammer (-8.5%). All his other teammates’ SAT% changed very little. Gryba, on average, appeared to make his teammates’ SAT% worse, which is consistent with his SAT% relative-to-teammates metric.) In particular, Gryba lowered his teammates’ SAT% by 3.7%. Again, Gryba’s negative impact mostly involves lowering his teammates’ ability to generate offense with 7.1 fewer shot-attempts generated per 60 minutes.

On the defensive side of things, there wasn’t much of a difference. Both Marincin and Gryba appeared to be weaker compared to their teammates in suppressing shots, especially high quality shots. But in examining their impact on their teammates’ shot attempt differentials, a difference seems to be emerging. Marincin’s offense appears to be more generative than that of Gryba.

High & Medium Probability Shots: Offense

Examining shot generation in more depth, I focused on their contributions to high and medium-probability scoring chances. Here we find a massive difference in shots from the slot. With Marincin on the ice, he and his teammates are ability to generate 12% more high-probability shots, relative to the league, and 7.7%% more shots relative to the Oilers’ average. In contrast, with Gryba on the ice, he and his teammates generated 12.5% fewer high-probability shots, relative to the league, and 16.7% fewer shots relative to the Senators’ average. In other words, Marincin’s Oilers generated 24.5% more shots from the slot than Gryba’s Senators. What of medium-probability shots? Marincin and his teammates generated 7.1% fewer shots than league average, and Gryba and his teammates generated 10.6% shots below league average. In other words, similar to high-probability shots, Marincin’s Oilers generated more medium-probability shots than Gryba’s Senators.

Net Expected Goals

This shot quality information indicates that Marincin and his teammates generated not only more shot attempts, but substantially more quality shots than Gryba and his teammates. To make this more understandable–“All this talk about shots, but what about actual goals?”–these shot metrics can be converted into expected goals. Fortunately, using shot rates–for and against–in all three scoring zones, war-on-ice computed net expected goals for each player.

With Marincin on the ice, the Oilers’ net expected goals was +0.07 goals/60 minutes. Applying this to Marincin’s time-on-ice, this converts to +.78 goals over the season. Regrettably, then, whatever gains Marincin made in helping his line-mates generate more quality shots, much of this gain is canceled out by his limitations in suppressing quality shots.

With Gryba on the ice, the Senators’ net expected goals was -0.38 goals/60 minutes. Applying this to his time-on-ice, this converts to 6.07 goals against over the season. Gryba’s apparent weaknesses in both shot suppression and generation, especially the latter, ultimately result in 6 more expected goals.allowed than scored.

Shot Attempt Metrics: Impact of Primary Teammates

Returning to the shot-attempt differentials, I was curious to know what impact key teammates had Marincin’s and Gryba’s SAT%. Above, I noted their impact on their teammates’ SAT%, but what of the reciprocal impact of their teammates on their own SAT%? In particular, I wanted to know the influence of their primary defensive partner and first-line center, who is presumably their strongest possession center.

Primary Defensive Partner

Marincin’s primary defensive partner was Mark Fayne with whom he shared 50% of his ice-time. (Marincin also shared about 11% of his time with Schultz and 9% with Nikitin.) With Fayne, his SAT% dropped from 51.3% to 47.1%. When paired, they did play tougher competition. But as I noted above, WOWY tables allow us to compare players across similar levels of competition. I will be presenting a full analysis of all the Oilers in the next coming weeks, but what I can say here is that Fayne’s SAT% was less than Marincin’s SAT% at almost every level of competition. Also, Fayne’s lowered the SAT% of his most frequent teammates by -2.61%. What is one reasonable conclusion using this information? It appears that Fayne dragged down Marincin’s shot-attempt differential, which is not what we were expecting from the veteran. I’ll leave Fayne’s struggles for a future post.

When Gryba was with his partner, Mark Borowiecki, Gryba’s SAT% fell from 48.3% to 45.3% (-3%). Similar to Marincin, then, Gryba was worse off with his defensive partner than without, thus did not depend on his partner to carry him. Borowiecki’s SAT% also fell by about 3%. (48.2% to 45.3%). So either defenseman was not doing his partner any favours.

First-Line Center

When Marincin shared ice-time with Nugent-Hopkins–the Oilers #1 center–Marincin’s SAT% improved from 47.6% to 52.1% (+4.5%). Nugent-Hopkins’s 51.1% SAT% with Marincin was also an improvement from 49.7% (+2.4%). Thus, both players benefited from each other and improved to a respectable shot-attempt differential.

When Gryba was on the ice with Mika Zibanejad, the Senators #! center, his SAT% fell by almost 4% (47.7% to 43.8%). Thus, he was worse off with Zibanejad, as was Zibanejad, whose SAT% fell by a whopping 7.5% (51.3% to 43.8%). This is consistent with the overall pattern noted above: Gryba tends to drag down a a teammate’s shot-attempt differential, mostly in reduced shot generation. There is one center, though, who improved Gryba’s SAT% and that is Curtis Lazar. Lazar helped improved Gryba’s SAT% by 4.2%, but Zybra did not reciprocate: Lazar’s SAT% dropped by 1.3% when sharing time with Gryba. Overall, this pattern suggests that Gryba is not only not a possession driver, but that he drags down his teammates’ shot-attempt differentials.

Direct Contributions to Shot Attempts and Scoring Chances

Last season, Ryan Stimson and a host volunteers collected data for what Stimson refers to as the Passing Project. Some teams had their complete seasons analyzed (i.e., tracking events through video reviews of games), but most teams only had partial data. The Oilers had data from about a dozen games analyzed. A few of the metrics generated from this data are relevant here. One metric is referred to as a player’s Corsi Contribution (CC%). Relative to all shot attempts while he was on the ice, this metric counts what proportion originated from a player’s shot attempts, as well as his primary and secondary passes that lead to shot attempts. To account for direct contributions to quality shots (i.e., scoring chances), there is Scoring Contribution (SC%), which includes scoring chances (as defined by war-on-ice) and primary passes that lead to scoring chances. Finally, there is Shot Attempt Generation Efficiency (SAGE%). This metric computes what percentage of primary passes lead to shots on goals as a proportion unblocked and blocked shots. The SAGE metric answers the question: How often does a player’s primary passes toward shots on goals?

Below I compare the players on each metric and in parantheses, I note their league-wide rank among defensemen.

[table id=9 /]

Except for the SAGE metric, they appear to be quite similar to each other in their ability to generate shot attempts and scoring chances. The sample size is on the small end, especially for Marincin, (1/6th of his time-on-ice), which limits making any strong conclusions, but the metrics are informative.

Shot Metric Summary

Marincin has outperformed Gryba in various shot metrics. First, at almost every quality of competition level, Marincin’s shot-attempt differential exceeded that of Gryba. Second, when examining high quality shots–for and against–with Marincin on the ice, the Oilers net expected goals was about even. With Gryba, the Senators were expected to have 6 more goals against. Third, Marincin’s specific shot metric advantage over Gryba appears to be in helping his team generate more quality shots, which has some confirmation (limited by small sample size) from the Passing Project data. However, important to note that Gryba’s scoring chance contribution is equivalent to that of Marincin. Fourth, when Marincin shares ice-time with Nugent-Hopkins, both he and Nugent-Hopkins improve their shot-attempt differentials. When Zybra is on the ice with Zibanejad, both his and Zibanejad’s shot-attempt differentials drop considerably.

If most of these metrics point to Marincin as being the better defenseman, why did the Oilers trade him? To be honest, I don’t know. I would imagine a rationale would refer to Gryba’s greater physicality and that he is somehow a better stay-at-home defenseman. But if his physicality is not resulting in better shot metrics, offensively or defensively, than why does his physicality matter?

Point Production

Despite all this, there is one measurable area in which Gryba seems to have an advantage over Marincin and that is in actual point production. Although they seem fairly similar in their direction contributions to scoring chances, Gryba’s points/60 minutes of 0.69 (11 points; 6 primary assists & 5 secondary assists) exceeded Marincin’s production of 0.36 points/60 (4 points; 1 goal, 2 primary assists, 1 secondary assists). Perhaps Gryba produced more because Ottawa’s (5v5) shooting percentage exceeded Edmonton’s by 1% (8% vs. 7%) and generated more scoring chances (Scoring Chances/60 = 26.6) than Edmonton (SC60 = 24.6).

Proportionally, Gryba also contributed more to the team’s offense than Marincin. Gryba was involved in 32% of the team’s goals when on the ice. This metric is consistent across his first 3 seasons. Marincin, in comparison, was involved in 23.5% of his team’s goals when on the ice. Gryba’s and Marincin’s individual shot metrics were equivalent. Gryba’s shot rate was 3.77 shots/60 and Marincin’s was 3.14 shots/60.

Concluding Comments

I’m not a fan of this deal. I’m not sure of what it accomplished. Neither defenseman is particularly strong defensively, but as my analysis suggests, Marincin seemed to show more potential. Against almost every level of competition, he outperformed Gryba. Marincin’s Break-Even SAT% of 50.1% suggests he has an ability to hold his own against some reasonably strong competition, whereas held his own against much weaker competition. Gryba tended to negatively impact his teammates’ shot-attempt differentials. Although Marincin didn’t help his teammates’ SAT%, on average, neither did he worsen their SAT%. However, when Marincin played along side Edmonton’s top-line players, like Nugent-Hopkins, the improvement in shot-attempt differential was mutual. Breaking down the offensive side of Marincin’s shot metrics, his strength appeared to be his ability to directly contribute to scoring chances, especially in terms of his efficient passing. Gryba also shows some strength in creating scoring chances and moreover, out-produced Marincin. But his production advantage over Marincin is most likely due to the better offensive team around him. In fact, based on shot quality data (medium & high-probability shots), when Gryba was on the ice, Ottawa was expected to have 6 more goals against. Marincin, and his on-ice teammates, were expected to essentially break even in net goals.

If there is any fans of Gryba reading this article, I’d like to hear from you. What qualities does he bring to Oilers that are a measurable advantage over Marincin? To be honest, I’m not big on intangibles (e.g., grit, physicality), that is, unless a person can show how these qualities translate into measurable events that lead a team to score more goals than their opposition. I also welcome any other questions and feedback, especially on the new metrics and analysis. I hope I made the new metrics understandable to you.

Thanks for your patience  in reading this until the end. As I learn more about these metrics, I hope to make my future anyalyses on Oiler trades briefer.